title
stringlengths 1
827
⌀ | uuid
stringlengths 36
36
| pmc_id
stringlengths 6
8
| search_term
stringclasses 18
values | text
stringlengths 0
6.94M
|
|---|---|---|---|---|
Integration and Potential Applications of Cardiovascular Computed Tomography in Cardio-Oncology | ae3d38d1-cb3b-4733-b4c8-3d9e3b8be4c7 | 11814013 | Internal Medicine[mh] | The field of cardio-oncology continues to grow in its significance as cardiovascular disease (CVD) is the top cause of morbidity and non-cancer mortality in the growing population of cancer survivors . Increased cardiovascular risk is likely due to a combination of the effect of shared risk factors, cancer itself, and cancer therapy-related adverse effects. There are many underlying shared risk factors between CVD and cancer including hypertension, hyperlipidemia, obesity, physical inactivity, poor diet, diabetes, and smoking. Different therapies and certain cancer types may be more associated with atherosclerotic cardiovascular disease (ASCVD); however, traditional ASCVD risk models do not take into account this heterogeneity of cardiovascular (CV) risk . In addition, The cardio-oncology population encompasses different groups of individuals, including those with many shared cardiovascular and oncologic risk factors, patients undergoing pre-treatment assessment before initiating cancer therapeutics, those actively receiving cancer treatment, and individuals who have completed therapy. This latter group includes patients who may still have cancer, as well as cancer survivors without current evidence of disease or recurrence, thus no longer classified as having active cancer. Computed tomography (CT) is readily used for cancer staging but can also reveal underlying CVD . The use of cardiovascular imaging techniques, such as cardiovascular computed tomography (CCT), has emerged as a valuable tool for the evaluation and management of cardiovascular conditions in patients with cancer. CCT offers several advantages that make it a valuable imaging modality in cardio-oncology. These include non-invasive and high-resolution imaging of coronary arteries with measurement of coronary artery calcium (CAC) scores, evaluating for obstructive and nonobstructive plaque, differentiating between ischemic and nonischemic etiologists of cardiomyopathy etiology, and diagnosing pericardial disease, pulmonary embolism, and calcific valvular heart diseases. Our review aims to highlight the current indications, advantages, and challenges of the use of CCT in the field of cardio-oncology. Cardio-oncology guidelines recommend the use of CCT for chest pain evaluation and measurement of CAC for CV risk assessment (Table ). The European Society of Cardiology (ESC) recommends the use of coronary computed tomography angiography (CCTA) to exclude acute coronary syndrome (ACS) in cancer-related Takotsubo syndrome as a Class 1 indication with C level of evidence (LOE) . The ESC Cardio-oncology guidelines also recommend the use of CAC scoring to reclassify baseline CV risk in addition to traditional risk factors. (Class 1, LOE C) Furthermore, beginning at 5 years after chest radiotherapy, CCTA screening can be considered for high-risk patients to detect radiation-induced coronary artery disease (CAD) and valvular calcifications, and it can be used to guide the management of ischemia as a Class 1 indication with C LOE. It is important to note, that despite the guideline’s endorsement of these CCTA indications, the LOE is categorized as C (driven by expert opinion and/or low-level of evidence specifically in patients with cancer), highlighting the need for further research, such as randomized-controlled trials and prospective studies in the cardio-oncology population. A Society of Cardiovascular Computed Tomography (SCCT) expert consensus endorsed by IC-OS (International Cardio-Oncology Society) has published a statement that provides recommendations of applications of CCTA among patients with cancer, which include using readily available non-cardiac chest CT scans to report CAC absence or presence, and estimation of CAC extent in asymptomatic patients with cancer. Moreover, non-contrast gated CAC score CT is recommended for baseline CVD risk factor evaluation as a way to further refine ASCVD risk stratification to help guide decision-making to start lipid-lowering therapy, and prior to planned valvular interventions. Table lists Class I and/or strong recommendations by the ESC and SCCT. (8) ACC CV Imaging and Cardio-oncology Councils have released a joint statement about the significance of using multimodal imaging in patients with cancer . CCTA can assess CAD and cardiac masses as well as help with the preplanning of transcatheter valve repair procedures. Additionally, CCTA can evaluate for cardiotoxicity-caused ACS-like symptoms. For cancer survivors, CCTA can assess traditional ASCVD risk. The American College of Cardiology/American Heart Association (ACC/AHA) has not released an official joint expert consensus document, yet, due to the need for more rigorous evidence in the cardio-oncology population to guide recommendations. Patients with cancer have an increased risk of CAD due to shared risk factors, radiation therapy, and chemotherapy-induced cardiotoxicity . Using contrast, CCTA can identify obstructive and nonobstructive coronary artery disease, enabling early intervention, optimization of CV risk, and optimal medical or interventional management strategies (Table ). This section discusses the utility of CCT in several cardiovascular disease states (Fig. ). Coronary Artery Disease In addition to its use in cancer diagnosis and tumor staging, CT can also be used for the detection of coronary atherosclerosis. CCT has emerged as a valuable tool to visualize coronary anatomy, including bypass grafts and stents, to detect coronary artery plaques, including plaque burden, degree of stenosis, calcification, and other characteristics. CCT particularly plays a crucial role in detecting subclinical atherosclerosis and non-obstructive/non-calcified plaque by quantifying CAC. Traditional clinical ASCVD risk scores do not take into account cancer-related risk factors such as specific chemoradiation therapies and the presence of somatic mutations defined as clonal hematopoiesis of indeterminate potential (CHIP), which has been associated with a higher degree of CVD in an older population without a history of cancer . Readily available, routinely obtained non-cardiac chest CT for cancer evaluation can provide the chance to estimate the extent of CAC for ASCVD risk stratification, in addition to traditional ASCVD 10-year risk stratification . Non-cardiac CT scans can be integrated to measure CAC scores after necessary reconstructions such as evaluation with slice thickness of 2–3 mm . CAC DRS (data reporting system) helps risk classification based on Agatston or visual CAC scores . The Agatston score category identifies CAC = 0 as very low risk, CAC 1–99 as mildly increased, CAC 100–299 as moderately increased and CAC > 300 as moderately to severely increased risk (Table ). CAC is a robust predictor of CVD and increasing CAC scores are associated with higher all-cause mortality in patients with a history of cancer . Enhancing traditional ASCVD risk estimations with CAC scores, CV risk groups can be reclassified to help control CV risk by implementing primary prevention strategies such as statins . In addition to these applications, CCTA is an effective tool with high negative predictive value to rule out obstructive CAD. This is particularly useful for patients with cancer, who often face physical constraints that limit their ability to undergo exercise-based stress testing and have increased risk of complications associated with invasive coronary angiography (i.e., increased risk of bleeding due to low platelet counts). In this population, CT - Fractional flow reserve (CT-FFR) enables a noninvasive assessment of hemodynamic significance of stenoses in the coronary arteries . CT-FFR represents a relatively recent advancement that enhances the functional assessment capability of CCT, addressing its limitation of stenosis specificity . CCT is also useful in identifying nonobstructive lesions and high-risk plaque features such as thin fibrous cap, lipid core in plaques with high risk of rupture and positive remodeling . Numerous outcome studies provide a strong correlation between CAC scores and cardiovascular mortality. Patients with cancer with a higher CAC score are at a greater risk of CV events . By identifying nonobstructive plaque characteristics, CCT assists in initiating appropriate primary/secondary preventive interventions such as statins, antiplatelet therapy, and/or, ezetimibe, proprotein convertase subtilisin/Kexin type 9 (PCSK9) inhibitors for optimal lipid-lowering strategies . Implementing early interventions to mitigate CV risk profiles of patients with cancer may lead to improved outcomes and overall survival rates; however, current data is still lacking on outcomes of early interventions of ASCVD risk factor modification in the cancer population . Increased risk of CVD in people with cancer is a well-recognized complication that may arise early in cancer treatment or later during survivorship care; proactively monitoring these patients to lower CV risk at the earliest opportunity is of utmost importance . Some cardiotoxic agents can mimic acute coronary syndrome (ACS), complicating diagnosis and management. Mimicking ACS can happen through various mechanisms. For example, 5-fluorouracil (5-FU) can cause coronary vasospasm, immune checkpoint inhibitors (ICI) can cause myocarditis, trastuzumab cardiotoxicity can cause left bundle branch block, and cardiac dysfunction during treatment . In these cases, CCTA plays a pivotal role in excluding or diagnosing ACS by providing detailed imaging of the coronary arteries. CCTA, in patients with a lower pretest probability of ASCVD, can help determine if invasive therapy is necessary, which is especially important for patients with a higher risk of procedural complications, such as those with hematologic abnormalities leading to increased bleeding risk. Valvular Diseases Valvular dysfunction is more prevalent among patients with cancer and progresses during cancer treatment in more than 30% of the patients reported in several studies . CCT can assess valve morphology, annular size, severity of calcification, and/or severity of aortic or mitral valve stenosis. This is a critical tool in planning for surgical or percutaneous structural interventions, including transcatheter aortic/mitral interventions (TAVR / TMVR). 5,26 Lastly, CCT can be used to evaluate aortic arch and/or ascending aorta calcification (i.e. porcelain aorta) which can arise from certain cancer treatments, such as mediastinal radiation, and can be associated with a higher risk of perioperative strokes during cardiac surgeries or transcatheter valvular procedures . Cardiomyopathies Cancer treatment related cardiac dysfunction (CTRCD) is a serious and prevalent short and long term sequalae of cancer treatment that can cause both diastolic and systolic dysfunction. Young individuals with a history of cancer are at 15 times higher long-term risk of developing heart failure due to CTRCD . Traditionally, CTRCD has been most closely associated with anthracyclines and anti-HER2 (human epidermal growth factor receptor 2) therapies . CCT has an important role in ruling out CAD and ischemic causes of cardiomyopathy during the evaluation of new suspected CTRCD. There are emerging causes of treatment-related cardiomyopathy, including ICI-myocarditis. Particularly in the case of ICI myocarditis where troponin is elevated, cardiac dysfunction must be distinguished from ischemic cardiomyopathy . CCT provides similar accurate left ventricular ejection fraction (LVEF) estimations compared to cardiac magnetic resonance imaging (cMRI) . With contemporary dose modulation acquisition techniques, CT-derived LVEF measurement can be conducted with very low-radiation doses . The excellent spatial resolution in CCT enables visualization of coronary arteries and thus identifying cardiomyopathy etiology . Nonetheless, guidelines recommend use of cMRI and echocardiography in measuring LVEF; CCT requires iodinated contrast and ionizing radiation, making it a less attractive option to assess cardiac function as a standalone indication . cMRI is the preferred imaging modality for differentiating cardiomyopathies due to the additional soft tissue characterization, particularly the presence of fibrosis, and the lack of ionizing radiation and iodinated contrast required. It is important to recognize that in the functional assessment of heart failure, other imaging modalities such as echocardiography or cMRI are favored in the current guidelines . Pericardial Disease CCT helps identify the heterogeneous spectrum of pericardial diseases, including pericardial effusions that may arise from active malignancy or a consequence of the cancer treatment, pericardial thickness, and pericardial calcifications from chronic pericardial inflammation and/or treatments (i.e. radiation) . Readily available chest CT scans among patients with cancer can also raise a suspicion of pericardial diseases, especially pericardial effusion. While cMRI offers high-resolution imaging of pericardial and cardiac anatomy, Hounsfield unit (HU) measurement in CCT can be helpful in the discrimination of exudative or transudative effusion. Exudative effusions yield a higher HU due to higher content of pericardial fluid albumin, lactate dehydrogenase (LDH), and white blood cells . The high resolution of CCT, compared to transthoracic echocardiography (TTE) or cMRI, can be necessary to evaluate the thin pericardium and particularly to detect the presence of calcification in constrictive pericarditis. Radiation-Induced Cardiovascular Diseases Radiation-induced cardiovascular disease (RI-CVD) refers to any cardiovascular compromise in patients receiving radiation therapy . Chest radiation therapy is a cornerstone of treatment for certain cancers including lung and breast cancers. RI-CVD leads to multiple cardiovascular complications that can be seen on a single CCT, such as coronary artery disease, valvular dysfunction, myocardial dysfunction, cardiomyopathies, and pericardial . CCT provides detailed information about cardiac anatomy, coronary arteries, valves, pericardium, and extracardiac cardiac structures. Valvular dysfunction, particularly of the aortic and mitral valves, occurs due to accelerated valvular calcification and can be well seen using CCT imaging. CCT, as explained above, is useful in the assessment of obstructive or nonobstructive CAD . Furthermore, CCT may help radiation oncologists to accurately define cancerous target volume and spare critical cardiac structures, minimizing the risk of cardiotoxicity while optimizing tumor control . Pulmonary Embolism PE is a potentially life-threatening condition that can occur more prevalently during cancer treatment. CCT involves thoracic imaging which visualizes pulmonary arteries and thus incidentally, if the contrast bolus is appropriately timed, CCT can diagnose or rule out pulmonary embolism (PE) . Furthermore, it can be used to assess hemodynamic consequences of PE, including right heart strain by evaluating the relative sizes of the right and left ventricles. To utilize CCTA for PE evaluation, special consideration must be paid to the timing of contrast bolus to ensure full opacification of both the coronary arteries and the pulmonary arteries. Cardiac Masses Cardiac masses encompass various entities such as thrombi, vegetations, benign tumors like myxomas and papillary fibroelastomas, as well as rare malignant primary or metastatic tumors. CCT surpasses cMRI with its high spatial resolution. CCT can assess for tumor vascularity using contrast enhancement, calcification extent, the presence of adipose tissue, and simultaneous extracardiac cancer staging. Particularly for masses adjacent to prosthetic valves, CCT is the preferred choice and over cMRI in detecting calcified masses . Additionally, CT’s enhanced spatial resolution aids in 3D reconstruction, and may assist in radiation treatment planning for metastatic or primary malignancies involving the heart . However, due to its ability to distinguish tissue characteristics, cMRI is the preferred modality for distinguishing cardiac thrombus from malignancies, and for detailed characterization of cardiac tumor types . Preplanning for Transcatheter Procedures CCT has an important role in preplanning for transcatheter procedures, particularly for TAVR. CCT encompasses a comprehensive, noninvasive evaluation of the sequelae of radiation exposure in the heart and adjacent structures such as the lungs and peripheral vasculature. The holistic view of the aortic root and valvular anatomy is crucial and appropriate for patient selection as emphasized by SCCT Expert Consensus statement . In addition to its use in cancer diagnosis and tumor staging, CT can also be used for the detection of coronary atherosclerosis. CCT has emerged as a valuable tool to visualize coronary anatomy, including bypass grafts and stents, to detect coronary artery plaques, including plaque burden, degree of stenosis, calcification, and other characteristics. CCT particularly plays a crucial role in detecting subclinical atherosclerosis and non-obstructive/non-calcified plaque by quantifying CAC. Traditional clinical ASCVD risk scores do not take into account cancer-related risk factors such as specific chemoradiation therapies and the presence of somatic mutations defined as clonal hematopoiesis of indeterminate potential (CHIP), which has been associated with a higher degree of CVD in an older population without a history of cancer . Readily available, routinely obtained non-cardiac chest CT for cancer evaluation can provide the chance to estimate the extent of CAC for ASCVD risk stratification, in addition to traditional ASCVD 10-year risk stratification . Non-cardiac CT scans can be integrated to measure CAC scores after necessary reconstructions such as evaluation with slice thickness of 2–3 mm . CAC DRS (data reporting system) helps risk classification based on Agatston or visual CAC scores . The Agatston score category identifies CAC = 0 as very low risk, CAC 1–99 as mildly increased, CAC 100–299 as moderately increased and CAC > 300 as moderately to severely increased risk (Table ). CAC is a robust predictor of CVD and increasing CAC scores are associated with higher all-cause mortality in patients with a history of cancer . Enhancing traditional ASCVD risk estimations with CAC scores, CV risk groups can be reclassified to help control CV risk by implementing primary prevention strategies such as statins . In addition to these applications, CCTA is an effective tool with high negative predictive value to rule out obstructive CAD. This is particularly useful for patients with cancer, who often face physical constraints that limit their ability to undergo exercise-based stress testing and have increased risk of complications associated with invasive coronary angiography (i.e., increased risk of bleeding due to low platelet counts). In this population, CT - Fractional flow reserve (CT-FFR) enables a noninvasive assessment of hemodynamic significance of stenoses in the coronary arteries . CT-FFR represents a relatively recent advancement that enhances the functional assessment capability of CCT, addressing its limitation of stenosis specificity . CCT is also useful in identifying nonobstructive lesions and high-risk plaque features such as thin fibrous cap, lipid core in plaques with high risk of rupture and positive remodeling . Numerous outcome studies provide a strong correlation between CAC scores and cardiovascular mortality. Patients with cancer with a higher CAC score are at a greater risk of CV events . By identifying nonobstructive plaque characteristics, CCT assists in initiating appropriate primary/secondary preventive interventions such as statins, antiplatelet therapy, and/or, ezetimibe, proprotein convertase subtilisin/Kexin type 9 (PCSK9) inhibitors for optimal lipid-lowering strategies . Implementing early interventions to mitigate CV risk profiles of patients with cancer may lead to improved outcomes and overall survival rates; however, current data is still lacking on outcomes of early interventions of ASCVD risk factor modification in the cancer population . Increased risk of CVD in people with cancer is a well-recognized complication that may arise early in cancer treatment or later during survivorship care; proactively monitoring these patients to lower CV risk at the earliest opportunity is of utmost importance . Some cardiotoxic agents can mimic acute coronary syndrome (ACS), complicating diagnosis and management. Mimicking ACS can happen through various mechanisms. For example, 5-fluorouracil (5-FU) can cause coronary vasospasm, immune checkpoint inhibitors (ICI) can cause myocarditis, trastuzumab cardiotoxicity can cause left bundle branch block, and cardiac dysfunction during treatment . In these cases, CCTA plays a pivotal role in excluding or diagnosing ACS by providing detailed imaging of the coronary arteries. CCTA, in patients with a lower pretest probability of ASCVD, can help determine if invasive therapy is necessary, which is especially important for patients with a higher risk of procedural complications, such as those with hematologic abnormalities leading to increased bleeding risk. Valvular dysfunction is more prevalent among patients with cancer and progresses during cancer treatment in more than 30% of the patients reported in several studies . CCT can assess valve morphology, annular size, severity of calcification, and/or severity of aortic or mitral valve stenosis. This is a critical tool in planning for surgical or percutaneous structural interventions, including transcatheter aortic/mitral interventions (TAVR / TMVR). 5,26 Lastly, CCT can be used to evaluate aortic arch and/or ascending aorta calcification (i.e. porcelain aorta) which can arise from certain cancer treatments, such as mediastinal radiation, and can be associated with a higher risk of perioperative strokes during cardiac surgeries or transcatheter valvular procedures . Cancer treatment related cardiac dysfunction (CTRCD) is a serious and prevalent short and long term sequalae of cancer treatment that can cause both diastolic and systolic dysfunction. Young individuals with a history of cancer are at 15 times higher long-term risk of developing heart failure due to CTRCD . Traditionally, CTRCD has been most closely associated with anthracyclines and anti-HER2 (human epidermal growth factor receptor 2) therapies . CCT has an important role in ruling out CAD and ischemic causes of cardiomyopathy during the evaluation of new suspected CTRCD. There are emerging causes of treatment-related cardiomyopathy, including ICI-myocarditis. Particularly in the case of ICI myocarditis where troponin is elevated, cardiac dysfunction must be distinguished from ischemic cardiomyopathy . CCT provides similar accurate left ventricular ejection fraction (LVEF) estimations compared to cardiac magnetic resonance imaging (cMRI) . With contemporary dose modulation acquisition techniques, CT-derived LVEF measurement can be conducted with very low-radiation doses . The excellent spatial resolution in CCT enables visualization of coronary arteries and thus identifying cardiomyopathy etiology . Nonetheless, guidelines recommend use of cMRI and echocardiography in measuring LVEF; CCT requires iodinated contrast and ionizing radiation, making it a less attractive option to assess cardiac function as a standalone indication . cMRI is the preferred imaging modality for differentiating cardiomyopathies due to the additional soft tissue characterization, particularly the presence of fibrosis, and the lack of ionizing radiation and iodinated contrast required. It is important to recognize that in the functional assessment of heart failure, other imaging modalities such as echocardiography or cMRI are favored in the current guidelines . CCT helps identify the heterogeneous spectrum of pericardial diseases, including pericardial effusions that may arise from active malignancy or a consequence of the cancer treatment, pericardial thickness, and pericardial calcifications from chronic pericardial inflammation and/or treatments (i.e. radiation) . Readily available chest CT scans among patients with cancer can also raise a suspicion of pericardial diseases, especially pericardial effusion. While cMRI offers high-resolution imaging of pericardial and cardiac anatomy, Hounsfield unit (HU) measurement in CCT can be helpful in the discrimination of exudative or transudative effusion. Exudative effusions yield a higher HU due to higher content of pericardial fluid albumin, lactate dehydrogenase (LDH), and white blood cells . The high resolution of CCT, compared to transthoracic echocardiography (TTE) or cMRI, can be necessary to evaluate the thin pericardium and particularly to detect the presence of calcification in constrictive pericarditis. Radiation-induced cardiovascular disease (RI-CVD) refers to any cardiovascular compromise in patients receiving radiation therapy . Chest radiation therapy is a cornerstone of treatment for certain cancers including lung and breast cancers. RI-CVD leads to multiple cardiovascular complications that can be seen on a single CCT, such as coronary artery disease, valvular dysfunction, myocardial dysfunction, cardiomyopathies, and pericardial . CCT provides detailed information about cardiac anatomy, coronary arteries, valves, pericardium, and extracardiac cardiac structures. Valvular dysfunction, particularly of the aortic and mitral valves, occurs due to accelerated valvular calcification and can be well seen using CCT imaging. CCT, as explained above, is useful in the assessment of obstructive or nonobstructive CAD . Furthermore, CCT may help radiation oncologists to accurately define cancerous target volume and spare critical cardiac structures, minimizing the risk of cardiotoxicity while optimizing tumor control . PE is a potentially life-threatening condition that can occur more prevalently during cancer treatment. CCT involves thoracic imaging which visualizes pulmonary arteries and thus incidentally, if the contrast bolus is appropriately timed, CCT can diagnose or rule out pulmonary embolism (PE) . Furthermore, it can be used to assess hemodynamic consequences of PE, including right heart strain by evaluating the relative sizes of the right and left ventricles. To utilize CCTA for PE evaluation, special consideration must be paid to the timing of contrast bolus to ensure full opacification of both the coronary arteries and the pulmonary arteries. Cardiac masses encompass various entities such as thrombi, vegetations, benign tumors like myxomas and papillary fibroelastomas, as well as rare malignant primary or metastatic tumors. CCT surpasses cMRI with its high spatial resolution. CCT can assess for tumor vascularity using contrast enhancement, calcification extent, the presence of adipose tissue, and simultaneous extracardiac cancer staging. Particularly for masses adjacent to prosthetic valves, CCT is the preferred choice and over cMRI in detecting calcified masses . Additionally, CT’s enhanced spatial resolution aids in 3D reconstruction, and may assist in radiation treatment planning for metastatic or primary malignancies involving the heart . However, due to its ability to distinguish tissue characteristics, cMRI is the preferred modality for distinguishing cardiac thrombus from malignancies, and for detailed characterization of cardiac tumor types . CCT has an important role in preplanning for transcatheter procedures, particularly for TAVR. CCT encompasses a comprehensive, noninvasive evaluation of the sequelae of radiation exposure in the heart and adjacent structures such as the lungs and peripheral vasculature. The holistic view of the aortic root and valvular anatomy is crucial and appropriate for patient selection as emphasized by SCCT Expert Consensus statement . CCT offers several advantages over other imaging techniques in cardio-oncology. CCT provides superior spatial resolution and the ability to evaluate the entire coronary tree and extracardiac structures. In comparison to cMRI, CCT is less susceptible to motion artifacts, making it more suitable for patients who have difficulty remaining still during the imaging process . Additionally, CT imaging is faster and more readily available than MRI, which can be important in the timely evaluation of cardio-oncology patients. Lastly, CCT plays an important role in the assessment of cardiac masses, with the ability to also evaluate calcified elements, within the heart as mentioned above . Overall, CCT combines excellent spatial resolution, comprehensive cardiac assessment, and accessibility, making it a valuable imaging modality in cardio-oncology for the detection, monitoring, and management of cardiovascular complications associated with cancer and its treatments (Fig. ). While CCT offers numerous advantages in cardio-oncology, it is essential to acknowledge its limitations. Understanding these limitations is crucial for healthcare providers to make informed decisions regarding the appropriate use of CCT in cardio-oncology patients. Risk of Ionizing Radiation CCT involves the use of ionizing radiation, which poses a potential risk to patients who often require repeated imaging studies for staging during their cancer treatment course . It is important to balance the potential benefits of CCT with the radiation risk, especially in younger patients with breast cancer, those undergoing radiotherapy, and those with genetic predispositions to developing malignancies. Radiation exposure should be decreased with the As Low As Reasonably Achievable (ALARA) approach which constitutes three components: lowering time, maximizing the distance, and the appropriate shielding . Although retrospectively gated heliacal CCT has a high amount of radiation exposure as high as 9–32 mSv, several dose reduction strategies have been developed such as ECG-correlated tube current modulation resulting in 37% radiation dose reduction in CCT . Additionally, prospective axial gating protocol offers up to 77% reduction in radiation dose. Thus, the modern CCT procedure typically results in low amounts of radiation, outweighing its risks (Table ). Because prospective gating does not capture during systole, CCT may have limited application for serial monitoring of LVEF due to the risk of radiation . Radiation dose reduction strategies, including appropriate patient selection and optimization of scanning protocols, should be employed to minimize radiation exposure while maintaining diagnostic image quality . Use of Iodinated Contrast CCT commonly requires the use of iodinated contrast agents, which can pose risks for patients with impaired kidney function . Contrast-induced nephropathy (CIN) is defined as an elevation of serum creatinine of more than 25% or ≥ 0.5 mg/dl (44 µmol/l) from baseline within 48 h of exposure. Also, even though rare, patients with a known allergy to iodinated contrast agents may not be suitable candidates for CCT, and alternative imaging modalities or contrast agent protocols may need to be considered . Artifacts Motion artifacts can significantly impact the accuracy and reliability of CCT images, leading to high false positive rates and potential diagnostic uncertainty . Techniques such as breath-holding instructions and heart rate control, with the administration of medications such as beta-blockers when needed, can help mitigate motion artifacts; however, challenges may persist, especially in patients who struggle with breath-holding or have an irregular heart rate. Blooming artifacts arise due to high-density structures, such as calcium or stents, making them appear larger than their actual size. This can be due to partial volume averaging, motion, or beam hardening . Blooming artifacts can compromise the accuracy of CT images, leading to difficulties in accurately assessing nearby anatomical structures and potentially leading to false-positive findings . Beam hardening artifacts can compromise CCT images by creating shadings mimicking myocardial ischemia. Cone-beam artifact occurs when the cone-beam geometry is inappropriate, shadings occur near the spine and ribs. Banding artifacts caused by irregular heartbeats or suboptimal gating scheme can lead to non-diagnostic images. β-blocker use can reduce heart rate variation and more robust gating schemes can solve these issues . Limited Application in Patients with Certain Conditions While functional assessment of intermediate coronary stenosis is enhanced with the addition of FFR, alternative imaging modalities such as MRI, nuclear, and echocardiography stress imaging offer a more comprehensive evaluation of cardiac function and ischemia and should be considered when CCTA will likely not be diagnostic. Alternative imaging options, such as cMRI, should be explored in these situations to ensure patient safety and diagnostic accuracy. Additionally, CT has limited soft tissue contrast and evaluation of some diagnoses may be better suited to echocardiography or cMRI, including infiltrative cardiomyopathies, fibrosis, or myocardial edema. Lastly, a major limitation of CCTA is the need to have a controlled heart rate for optimal imaging, usually a heart rate < 60 bpm, which often requires administration of B-blockers. Additionally, nitrates are required for standard clinical CCTA exams to allow for accurate assessment of coronary stenoses. Given that patients with cancer often have sinus tachycardia and borderline low blood pressure, there may be clinical limitations to obtaining CCTA in some circumstances . CCT involves the use of ionizing radiation, which poses a potential risk to patients who often require repeated imaging studies for staging during their cancer treatment course . It is important to balance the potential benefits of CCT with the radiation risk, especially in younger patients with breast cancer, those undergoing radiotherapy, and those with genetic predispositions to developing malignancies. Radiation exposure should be decreased with the As Low As Reasonably Achievable (ALARA) approach which constitutes three components: lowering time, maximizing the distance, and the appropriate shielding . Although retrospectively gated heliacal CCT has a high amount of radiation exposure as high as 9–32 mSv, several dose reduction strategies have been developed such as ECG-correlated tube current modulation resulting in 37% radiation dose reduction in CCT . Additionally, prospective axial gating protocol offers up to 77% reduction in radiation dose. Thus, the modern CCT procedure typically results in low amounts of radiation, outweighing its risks (Table ). Because prospective gating does not capture during systole, CCT may have limited application for serial monitoring of LVEF due to the risk of radiation . Radiation dose reduction strategies, including appropriate patient selection and optimization of scanning protocols, should be employed to minimize radiation exposure while maintaining diagnostic image quality . CCT commonly requires the use of iodinated contrast agents, which can pose risks for patients with impaired kidney function . Contrast-induced nephropathy (CIN) is defined as an elevation of serum creatinine of more than 25% or ≥ 0.5 mg/dl (44 µmol/l) from baseline within 48 h of exposure. Also, even though rare, patients with a known allergy to iodinated contrast agents may not be suitable candidates for CCT, and alternative imaging modalities or contrast agent protocols may need to be considered . Motion artifacts can significantly impact the accuracy and reliability of CCT images, leading to high false positive rates and potential diagnostic uncertainty . Techniques such as breath-holding instructions and heart rate control, with the administration of medications such as beta-blockers when needed, can help mitigate motion artifacts; however, challenges may persist, especially in patients who struggle with breath-holding or have an irregular heart rate. Blooming artifacts arise due to high-density structures, such as calcium or stents, making them appear larger than their actual size. This can be due to partial volume averaging, motion, or beam hardening . Blooming artifacts can compromise the accuracy of CT images, leading to difficulties in accurately assessing nearby anatomical structures and potentially leading to false-positive findings . Beam hardening artifacts can compromise CCT images by creating shadings mimicking myocardial ischemia. Cone-beam artifact occurs when the cone-beam geometry is inappropriate, shadings occur near the spine and ribs. Banding artifacts caused by irregular heartbeats or suboptimal gating scheme can lead to non-diagnostic images. β-blocker use can reduce heart rate variation and more robust gating schemes can solve these issues . While functional assessment of intermediate coronary stenosis is enhanced with the addition of FFR, alternative imaging modalities such as MRI, nuclear, and echocardiography stress imaging offer a more comprehensive evaluation of cardiac function and ischemia and should be considered when CCTA will likely not be diagnostic. Alternative imaging options, such as cMRI, should be explored in these situations to ensure patient safety and diagnostic accuracy. Additionally, CT has limited soft tissue contrast and evaluation of some diagnoses may be better suited to echocardiography or cMRI, including infiltrative cardiomyopathies, fibrosis, or myocardial edema. Lastly, a major limitation of CCTA is the need to have a controlled heart rate for optimal imaging, usually a heart rate < 60 bpm, which often requires administration of B-blockers. Additionally, nitrates are required for standard clinical CCTA exams to allow for accurate assessment of coronary stenoses. Given that patients with cancer often have sinus tachycardia and borderline low blood pressure, there may be clinical limitations to obtaining CCTA in some circumstances . Integration of Artificial Intelligence Algorithms As technology and research continue to advance, there are promising future directions for the use of CCT in cardio-oncology. The integration of artificial intelligence (AI); machine learning (ML) and deep learning (DL) algorithms into CCT analysis holds the potential for automated image interpretation, improved precision, personalized care, and enhanced ASCVD risk stratification in cardio-oncology patients . Current ML algorithms can accurately predict the stenoses grade and ischemia as shown in a CT-FFR study . In this study, an ML algorithm was trained on 581 vessels from the prospective PACIFIC trial to develop an ML score for ischemia prediction. The ML score was then applied to predict myocardial blood flow from corresponding cardiac PET scans and ML score performance was compared with CCTA reads and noninvasive CT-FFR. The study showed that ML algorithm have a higher area under the receiver-operating characteristic curve (AUC) compared to FFR-defined ischemia and impaired blood flow prediction. A study from CAC Consortium developed an ML model including 77 variables and is trained with data from 66,636 asymptomatic subjects. The model is evaluated using a cross-validation framework from the available data and predictive value of the proposed model is compared to ASCVD and CAC scores based on their performance in AUC . AUC in CVD and coronary heart disease (CHD) death prediction were superior to ASCVD and CAC scores. [CVD prediction: 0.845 (ML) 0.821 (ASCVD) 0.781 (CAC) / CHD prediction: 0.86 (ML) / 0.835(ASCVD) 0.816 (CAC); p < 0.0001 for all]. Deep learning (DL) is a subset of ML that uses neural networks with multiple hidden layers for capturing complex patterns and image recognition. It’s primarily used for large datasets and focuses on deeper interactions. Several studies using DL algorithms that are externally validated, meaning that is validated by a different cohort than its training cohort for minimizing the overfitting and maximizing generalizability, have reported that automated CAC score prediction is noninferior to expert-annotated CAC scores . Another study highlights the use of DL algorithm on non-ECG gated chest CTs to detect incidental CAC > 100, as this score is associated with a worse CVD and mortality outcomes independent from traditional risk factors . Thus, DL and ML algorithms are promising tools to allow for opportunities for earlier intervention and CVD prevention. Future work in the field of preventive cardiology should focus on supporting implementation of AI algorithms, identifying subclinical CVD in patients with a history of cancer and further personalizing the CVD prevention in people with cancer . As technology and research continue to advance, there are promising future directions for the use of CCT in cardio-oncology. The integration of artificial intelligence (AI); machine learning (ML) and deep learning (DL) algorithms into CCT analysis holds the potential for automated image interpretation, improved precision, personalized care, and enhanced ASCVD risk stratification in cardio-oncology patients . Current ML algorithms can accurately predict the stenoses grade and ischemia as shown in a CT-FFR study . In this study, an ML algorithm was trained on 581 vessels from the prospective PACIFIC trial to develop an ML score for ischemia prediction. The ML score was then applied to predict myocardial blood flow from corresponding cardiac PET scans and ML score performance was compared with CCTA reads and noninvasive CT-FFR. The study showed that ML algorithm have a higher area under the receiver-operating characteristic curve (AUC) compared to FFR-defined ischemia and impaired blood flow prediction. A study from CAC Consortium developed an ML model including 77 variables and is trained with data from 66,636 asymptomatic subjects. The model is evaluated using a cross-validation framework from the available data and predictive value of the proposed model is compared to ASCVD and CAC scores based on their performance in AUC . AUC in CVD and coronary heart disease (CHD) death prediction were superior to ASCVD and CAC scores. [CVD prediction: 0.845 (ML) 0.821 (ASCVD) 0.781 (CAC) / CHD prediction: 0.86 (ML) / 0.835(ASCVD) 0.816 (CAC); p < 0.0001 for all]. Deep learning (DL) is a subset of ML that uses neural networks with multiple hidden layers for capturing complex patterns and image recognition. It’s primarily used for large datasets and focuses on deeper interactions. Several studies using DL algorithms that are externally validated, meaning that is validated by a different cohort than its training cohort for minimizing the overfitting and maximizing generalizability, have reported that automated CAC score prediction is noninferior to expert-annotated CAC scores . Another study highlights the use of DL algorithm on non-ECG gated chest CTs to detect incidental CAC > 100, as this score is associated with a worse CVD and mortality outcomes independent from traditional risk factors . Thus, DL and ML algorithms are promising tools to allow for opportunities for earlier intervention and CVD prevention. Future work in the field of preventive cardiology should focus on supporting implementation of AI algorithms, identifying subclinical CVD in patients with a history of cancer and further personalizing the CVD prevention in people with cancer . In conclusion, CCT plays a role in risk stratification through the detection of coronary artery disease in both cardiac and non-cardiac scans as a pivotal step in preventive cardiovascular event management. By accurately assessing CAD risk, clinicians can implement tailored preventive measures, further reducing the incidence of cardiovascular events. Moreover, CCTA is an invaluable imaging modality for patents presenting with CAD symptoms, whether stable or acute. In evaluation of cardiomyopathy, CCT aids in distinguishing between ischemic cardiomyopathy or chemotherapy related cardiotoxicity. The role of CCT extends beyond CAD assessment, encompassing the evaluation of valves, pericardium, and cardiac masses, offering a holistic perspective on cardiac health and contributing to informed clinical decision-making. As advancements in cancer treatment leads to an increasing number of cancer survivors, CCT can be an invaluable tool in providing information on cardiac anatomy including the presence of preexisting or acquired cardiovascular disease through the continuum of the patient’s cancer journey. Lopez-Mattei J, Yang EH, Baldassarre LA, et al. Cardiac computed tomographic imaging in cardio-oncology: An expert consensus document of the Society of Cardiovascular Computed Tomography (SCCT). Endorsed by the International Cardio-Oncology Society (ICOS). J Cardiovasc Comput Tomogr . 2023;17(1). 10.1016/j.jcct.2022.09.002. This paper includes several recommendations for use of CCT in cardio-oncology population. Baldassarre LA, Ganatra S, Lopez-Mattei J, et al. Advances in Multimodality Imaging in Cardio-Oncology: JACC State-of-the-Art Review. J Am Coll Cardiol . 2022;80(16). 10.1016/j.jacc.2022.08.743. This review provides valuable and holistic approach to cardiovascular imaging in populations of patients with cancer. Miller RJH, Mamas MA, Tamarappoo B, et al. Extensive coronary artery calcification is associated with all-cause mortality patients with a history of cancer. J Cardiovasc Comput Tomogr . Published online 2023. 10.1016/j.jcct.2023.04.001. This review explores that higher coronary artery calcification increases the all-cause mortality in patients with a history of cancer. Using nongated scans, CAC scanning can be used for risk stratification in this population. |
Affinity Ultrafiltration Mass Spectrometry for Screening Active Ingredients in Traditional Chinese Medicine: A Review of the Past Decade (2014–2024) | f15049c7-8bcb-459c-9d35-9eb9da12ab29 | 11820328 | Surgical Procedures, Operative[mh] | Traditional Chinese medicine (TCM), a remarkable medical resource with a long history, has unique advantages in preventing and treating various diseases, particularly in the control of major epidemics and clinical treatment. Approximately 35% of the global pharmaceutical market annually derives directly or indirectly from natural products, predominantly from plant sources (25%), with microbial sources (13%) and animal sources (3%) following . The active ingredients of TCM form the material basis for its therapeutic effects and serve as an important source of biologically active compounds. However, TCM and its formulations often contain numerous chemical components, and the complexity of these mixtures makes the evaluation and identification of active ingredients highly challenging. Thus, identifying the active ingredients in TCM is a critical scientific challenge in its modernization and a significant bottleneck in its global development. The traditional strategy for researching active ingredients in TCM involves “chemical extraction and separation, molecular structure identification, and pharmacological activity evaluation” . Although effective, this strategy is cumbersome and time-consuming, making it challenging to efficiently screen active structures. Modern pharmacological research indicates that a drug’s affinity for biological macromolecules is the first step in its mechanism of action, and the drug target is the critical starting point for its therapeutic effects in vivo . Small molecules in TCM regulate biological processes and exert medicinal effects by interacting with target proteins in organisms. Consequently, molecular targeting methods for drug screening, based on disease-related biomacromolecules, have emerged. Affinity ultrafiltration (AUF)–liquid chromatography (LC)–mass spectrometry (MS) is a solution affinity selection platform that separates target–ligand complexes in solution via ultrafiltration. It serves as a powerful tool for identifying active molecules within complex natural products. Compared with traditional methods, AUF is simple to operate, and it significantly reduces screening time and lowers the consumption of samples and reagents. The technology enables the online integration of various detection instruments, allowing for an accurate reflection of the interaction between the natural conformation of active substances and receptors. Due to its high sensitivity and strong selectivity, AUF-LC-MS holds unique value in small-molecule drug discovery and has garnered widespread attention from the pharmaceutical community. Before that, Chen et al. also provided an overview, summary, and outlook on AUF-LC-MS technology. On this basis, this review provides a more comprehensive review of the basic principles, characteristics, and influencing factors of AUF-LC-MS technology, and summarizes its application in the screening of bioactive components of medicinal plants in the past ten years. For example, Panax ginseng has many functions such as enhancing immunity, anti-fatigue, and antioxidants. Panax ginseng is rich in saponins, which have a wide range of benefits for the human body. Modern pharmacological research shows that the most important ones are ginsenosides Rg1, Re, and Rb1 . In recent years, researchers have used α-glucosidase, acetylcholinesterase, Monoamine oxidase type-B, and N -methyl-D-aspartic acid as targets, and adopted AUF-LC-MS technology to screen out 24, 16, 7, and 3 active ingredients, respectively . In addition, 5, 12, and 32 active ingredients were also screened from Coptis chinensis Franch, Salvia miltiorrhiza Bge., Curcuma longa , etc. . Please refer to for specific contents, which provide a certain scientific basis for rapid targeted screening of active ingredients in medicinal plants. This review also discusses the adaptability of this technology to a wider range of natural products and its combination with other analytical techniques, and prospects for its development, so that AUF technology can be widely used internationally. AUF combines affinity capture with ultrafiltration, facilitating high-throughput compound screening . Developed in 1981, this technique initially evaluated ultrafiltration’s theoretical and experimental applications in clinical serum binding assays . Discovering drug target proteins is crucial for drug research . In the late 1990s, AUF became widely used in targeted drug discovery and an indispensable tool for many pharmaceutical companies. Over the past decade, significant advancements have been made in AUF in terms of membrane materials, separation properties, and system optimization. Many new affinity membrane materials have been developed recently to enhance the selectivity and performance of the membranes. The separation capabilities of these membranes are enhanced by introducing various affinity ligands or through surface modifications. For instance, the use of hydrophilic polymers, nanomaterials, or composite materials enhances the affinity and anti-fouling properties of these membranes . To enhance their separation performance, researchers have improved the separation abilities of AUF membranes by combining various affinity ligands, such as different antibodies, proteins, or small molecules. Particularly in complex biological systems, this multi-level separation significantly enhances the purity and efficiency of target molecule separation . Additionally, AUF technology has increasingly adopted automation and intelligent control systems to enhance operational efficiency. For example, the use of real-time sensors to monitor membrane status, in combination with machine learning algorithms for automatic adjustments, enhances both the performance and the operational ease of the membrane system . The ultrafiltration screening method is based on ligand–receptor-specific binding, with screening potential active ligands binding to the target protein by disease-specific characteristics . First, the ligand mixture is combined with the receptor. After ultrafiltration, the ligand dissociates from the receptor, or the binding part is directly observed. Finally, the potential active ingredients are analyzed by LC-MS. AUF-MS is mainly divided into centrifugal ultrafiltration-MS (CU-MS) and pulsed ultrafiltration-MS (PU-MS). In both methods, the basic principle of small-molecule screening is the same: ligand enrichment is achieved through the selectivity of a semi-permeable membrane. The CU-MS by ultrafiltration chamber and LC-MS platform operate independently, necessitating manual injection of ultrafiltration samples into the LC-MS system, hence the term “off-line ultrafiltration”. CU-MS employs commercial ultrafiltration centrifuge tubes to screen compounds, offering straightforward procedures and good reproducibility. Chen et al. developed an off-line ultrafiltration-LC-MS platform to screen for inhibitors of α-glucosidase and pancreatic lipase. Fifteen potential ligands, including glucomoringin, 3-caffeoylquinic acid, and quinic acid, were quickly screened and identified from Moringa oleifera leaf extracts. The study identified 14 potential α-glucosidase ligands and 10 potential pancreatic lipase ligands. Feng et al. captured 12 phytochemicals with varying affinities for topoisomerase I, topoisomerase II, COX-2, and ACE2 from Dysosma versipellis root and stem extracts by using an off-line ultrafiltration-LC-electrospray ionization (ESI)-MS/MS model. In vitro antiproliferation tests demonstrated that podophyllotoxin and quercetin had the strongest inhibition rates on A549 and HT-29 cells, whereas kaempferol exhibited a significant dose-dependent effect on COX-2. Additionally, quercetin exhibited a strong inhibitory effect on ACE2 . PU-MS consists of a flow chamber, a magnetic stirrer, and an ultrafiltration membrane . It is an online combination of PU and electrospray MS. After the test sample and target protein are added to the flow chamber, the ligand–receptor complex and inactive components can be separated by applying pressure. Unlike CU-MS, this technology is an online affinity MS screening method. Hence, it is also referred to as online ultrafiltration. PU-MS was first proposed by van Breemen et al. to screen potential compounds binding to target receptors from complex systems. Adenosine deaminase inhibitors were successfully identified from a combinatory chemical library of 20 adenosine analogs by using this method. Beverly et al. utilized PU-MS to evaluate a 35 μL binding chamber’s ability to screen ligands forming noncovalent complexes with protein targets. They found that the platform quickly screened and enriched the carbonic anhydrase inhibitor acetazolamide from bacterial fermentation broth extracts, completing the process in 5 min only. Compared with PU, CU cannot be integrated with MS online. Additionally, the concentration polarization during centrifugal ultrafiltration can reduce the filtration speed, and in severe cases, cause protein adsorption and deposition on the membrane surface, affecting free drug transport. Thus, CU is primarily used for screening small-molecule active compounds within a limited range. By contrast, PU, which easily integrates with LC-MS to form an automated, high-throughput system, is more effective for describing receptor–ligand binding characteristics, drug metabolism, and product identification. In conclusion, the advantage of ultrafiltration-based methods lies in their ability to rapidly provide binding information between drug targets and compounds. These methods can be used to study the synergistic or antagonistic effects of multiple compounds. The history of MS traces back to the early 20th century with the invention of the parabolic mass spectrometer by J.J. Thomson. In 1919, Aston developed the first velocity-focusing MS, marking a significant milestone in the field. Initially, MS was primarily used to determine the atomic weight of elements and isotopes. With advancements in ion optics theory, the technology continually improved, and by the late 1950s, it was widely applied in the analysis of inorganic and organic compounds. Owing to its high sensitivity, accuracy, and resolution, MS has become one of the most crucial analytical techniques in life sciences, medicine, and chemistry . The advent of MS technology, particularly soft ionization methods like ESI and matrix-assisted laser desorption/ionization (MALDI), has extended the application of MS to the early stages of drug discovery, specifically in the identification of lead compounds . Compared with earlier detection methods, MS does not require derivatization or isotope labeling, thereby expanding the range of applicable compounds, accelerating detection, and enhancing sensitivity and specificity. Thus, integrating MS with target affinity techniques—referred to as target molecule affinity-MS—has made drug screening more efficient and effective. In recent years, numerous MS techniques have been developed to address the increasing demand for analyzing and identifying specific components within complex substrates from multiple perspectives. These include techniques such as AUF-LC-MS, ESI-Q-TOF-MS, ultrahigh-performance LC (UPLC)–Orbitrap–(time-of-flight) TOF-MS, MALDI-TOF-MS, LC-MS, GC-MS, FT-ICR-MS, and DART-MS. Based on the above explanation, the ultrafiltration method effectively enriches and separates ligands that bind to target proteins while being easy to operate and cost-effective. AUF can screen ligand–protein complexes from unbound substances, and when combined with LC and MS, it enables rapid separation and identification of potential active ingredients. It can identify target substances at various concentrations, and it is suitable for analyzing small quantities of complex mixtures such as combinatorial compound libraries and extracts or fractions of medicinal plants. When AUF is combined with LC-MS n , the high sensitivity of MS compensates for the limitations of LC in detecting minute components with low sample content . As a high-throughput method, AUF-LC-MS performs well in screening active substances without stringent sample size requirements and offers additional advantages such as simplicity of operation and strong specificity. However, this method has certain limitations: false positives resulting from nonspecific adsorption in the ultrafiltration process typically need to be addressed through parallel control experiments using an inactivated target protein group or a serum protein replacement group. Additionally, ultrafiltration screening is primarily based on the affinity between the target protein and the ligand. As a result, while it evaluates the ligand’s affinity for the target protein, it does not directly reflect the ligand’s biological activity . Currently, various methods exist for screening active ingredients, such as cell membrane chromatography, magnetic bead screening, UV-visible spectroscopy, nuclear magnetic resonance (NMR), fluorescence, and electrochemical methods . Compared with these methods, the combination of AUF and MS for screening small-molecule active substances in TCM offers several advantages, including ease of operation, high sensitivity, and specific results. Traditional chromatographic methods based on optical or radioactive substances often encounter matrix interference . This interference complicates the identification and analysis of complex components in natural products. For instance, UV-visible spectroscopy measures α-glucosidase activity by hydrolyzing p -nitrophenyl- α - D -glucopyranoside, producing p -nitrophenol, detectable at 400 nm . However, NMR is time-consuming and not well suited for rapid inhibitor screening. Additionally, fluorescence and electrochemical methods suffer from significant interference issues . Consequently, a rapid and accurate method to screen active compounds with inhibitory effects is urgently needed. The ligand matrix does not affect the screening process of affinity MS. Thus, this unique advantage renders it particularly suitable for screening active ingredients in complex systems, especially traditional medicinal plants. Owing to its high sensitivity and selectivity, AUF-LC-MS has been effectively utilized to isolate and identify target substances from complex samples, playing a pivotal role in extracting active molecules from natural products. In AUF, researchers study the interactions between small drug molecules and biological targets in solution. Binding between AUF receptors and ligands occurs in solution, which avoids alterations in their properties from labeling or chemical coupling to solid supports, thereby preserving their natural conformation and interactions. Ultrafiltration requires only small quantities of the target, and some protein targets can be reused, making it a viable option when targets are costly, scarce, or available in limited quantities . Alternatively, the retention capability of the ultrafiltration membrane allows for the direct selection of active components that bind to target substances without the need for pretreatment, such as in immobilized enzyme online MS and cell membrane chromatography-MS . AUF-MS enables rapid determination of binding constants between biological targets and small drug molecules while concurrently providing activity data for these molecules. In the combined AUF-MS approach, AUF exhibits robust specificity and screening capabilities for small ligands in complex mixtures. Meanwhile, LC-MS offers potent functionality for efficient separation and structural identification, effectively minimizing matrix interference. AUF-LC-MS is extensively utilized for screening active ingredients from complex substrates because of its high-throughput capabilities. However, this technique has limitations, including the possibility that some identified candidates may not exhibit the expected activity or may show elevated activity, leading to potential false positives . Various factors must be considered during the experimental process, including the concentration of the target and screening substances, the material of the ultrafiltration membrane, the selection of the dissociation solvent, the interception volume, the co-incubation time, the centrifugal speed, and the solution pH, to mitigate false-positive or false-negative results. The screening conditions must be optimized to ensure the high efficiency and specificity of the screening results, and operations should be rationally designed and standardized. Additionally, the design of negative control experiments is crucial for reducing false positives and improving the accuracy of the results . 4.1. Concentration of the Target and the Screened Substances The concentrations of targets and screening substances are critical factors influencing the affinity filtration process. If the ligand concentration is significantly higher than that of the target protein, it may prevent some active ingredients from binding to the target proteins because ligand binding to target proteins is inherently competitive, leading to false negative results. Conversely, if the ligand concentration is too low, it may enhance nonspecific adsorption, thus increasing the likelihood of false positives. These false positives are often due to the nonspecific binding of the compound to the target protein. Yang et al. were the first to verify AUF-LC screening results to eliminate false positives by using competitive binding experiments. In fact, competitive binding experiments not only eliminate false positives but also exclude ligands that bind to different sites than those of competitively binding compounds. Wang et al. evaluated the feasibility of using competitive binding experiments combined with AUF-LC to identify xanthine oxidase (XOD) inhibitors in Perilla frutescens (L.) Britt., aiming to reduce false positives. In the experiment, P. frutescens extracts were incubated with XOD-free, XOD-present, or XOD-blocked active sites before ultrafiltration, and the total binding degree and specific binding degree of each compound were calculated on the basis of peak area. The results indicated that AUF-LC significantly reduced the number of false positives identified. However, this method cannot eliminate all false positives and may exclude some effective inhibitors. Therefore, a thorough methodological review is essential to obtain reliable binding results. The equilibrium dissociation constant (KD) is a critical metric for evaluating the interaction between a ligand and its target protein, with each component having its own distinct KD value. The KD values of the receptor and target ligand should be closely matched; otherwise, significant discrepancies may result in false positives or false negatives. In general, the receptor concentration should be close to the KD value of the weakest ligand. If the ligand concentration is too high, only ligands with strong binding affinity could bind to the target protein at competitive binding sites. Therefore, in actual experiments, the ligand concentration should be equal to or less than that of the receptor. Wang et al. developed an AUF-UPLC method to directly determine the KD of compounds in P. frutescens extracts and their target proteins, including the KD determination for α-glucosidase ligands in the ethyl acetate fraction of P. frutescens. The recovery rate, binding degree, and signal-to-noise ratio of α-glucosidase ligands in PFEA were determined using AUF-LC, followed by KD calculation using the proposed equilibrium. Oleanolic acid and apigenin were identified as high-affinity ligands of α-glucosidase, with KDs of 44.9 and 88.5 μM, respectively. These values were consistent with the results from isothermal titration calorimetry, kinetic analysis, and molecular docking simulations. The results demonstrate that this method is simple and easy to implement, allowing direct determination of KD values for compounds in natural product extracts without the need for internal standards or calibration agents. Optimizing these methods can enhance the screening accuracy and reliability of AUF-LC-MS, providing a robust foundation for the identification of active ingredients in complex substrates. 4.2. Ultrafiltration Membrane Material In AUF-LC-MS, ultrafiltration membranes separate ligand–receptor complexes from unbound components. The selection of ultrafiltration membranes primarily involves two factors: pore size and material . An ideal ultrafiltration membrane should effectively retain the target biological macromolecules while preventing leakage or clogging. The pore size should be less than one-third of the biomacromolecule’s size to ensure effective retention . Selecting the appropriate pore size improves separation efficiency and prevents leakage of unbound components. An ideal ultrafiltration membrane material should minimize specific adsorption with potential ligands and receptors. Common ultrafiltration membrane materials include polyvinyl fluoride, polysulfone, polyether ketone, and methylcellulose . These materials exhibit low nonspecific binding and are therefore widely used in ultrafiltration membrane production. Selecting the appropriate pore sizes and materials optimizes the separation efficiency of AUF-LC-MS and enhances the accuracy and reliability of the experiment, ensuring the authenticity of ligand–receptor interactions and reducing false-positive results. 4.3. Choice of Dissociation Solvent The complex components, diverse structures, and varying polarities of TCM extracts make it challenging to successfully dissociate ligands from the affinity target while minimizing nonspecific adsorption, a key factor affecting screening results. Two main methods are currently used to denature enzymes: adding acid to the dissociation solvent to inactivate the enzyme in a low pH environment or using organic solvents for enzyme denaturation. However, using organic solvent-based dissociation solutions only can sometimes increase nonspecific adsorption. Some related studies have demonstrated that acid-containing organic solvents, as opposed to those with organic solvents only, can effectively reduce nonspecific adsorption of non-affinity interacting substances. For example, Xie et al. used a methanol–water (90:10) mixture to screen potential TCM components targeting 5-lipoxygenase and cyclooxygenase-2. Comparison of the ultrafiltrate chromatograms between the experimental and control groups revealed significant differences in the peak areas of active ingredients, with lower signals for nonspecifically adsorbed substances. Conversely, some researchers have successfully screened small-molecule inhibitors of cyclooxygenase and glutathione reductase from TCM by using dissociation solutions containing organic solvents only . The findings indicate that different dissociation solutions yield varying effects, necessitating multiple experimental attempts to optimize dissociation conditions. In summary, selecting an appropriate dissociation solvent is essential for reducing nonspecific adsorption and enhancing the accuracy of screening results. Multiple experimental attempts are recommended to identify the optimal dissociation conditions by comparing results, thereby effectively screening the active ingredients in TCM extracts. The concentrations of targets and screening substances are critical factors influencing the affinity filtration process. If the ligand concentration is significantly higher than that of the target protein, it may prevent some active ingredients from binding to the target proteins because ligand binding to target proteins is inherently competitive, leading to false negative results. Conversely, if the ligand concentration is too low, it may enhance nonspecific adsorption, thus increasing the likelihood of false positives. These false positives are often due to the nonspecific binding of the compound to the target protein. Yang et al. were the first to verify AUF-LC screening results to eliminate false positives by using competitive binding experiments. In fact, competitive binding experiments not only eliminate false positives but also exclude ligands that bind to different sites than those of competitively binding compounds. Wang et al. evaluated the feasibility of using competitive binding experiments combined with AUF-LC to identify xanthine oxidase (XOD) inhibitors in Perilla frutescens (L.) Britt., aiming to reduce false positives. In the experiment, P. frutescens extracts were incubated with XOD-free, XOD-present, or XOD-blocked active sites before ultrafiltration, and the total binding degree and specific binding degree of each compound were calculated on the basis of peak area. The results indicated that AUF-LC significantly reduced the number of false positives identified. However, this method cannot eliminate all false positives and may exclude some effective inhibitors. Therefore, a thorough methodological review is essential to obtain reliable binding results. The equilibrium dissociation constant (KD) is a critical metric for evaluating the interaction between a ligand and its target protein, with each component having its own distinct KD value. The KD values of the receptor and target ligand should be closely matched; otherwise, significant discrepancies may result in false positives or false negatives. In general, the receptor concentration should be close to the KD value of the weakest ligand. If the ligand concentration is too high, only ligands with strong binding affinity could bind to the target protein at competitive binding sites. Therefore, in actual experiments, the ligand concentration should be equal to or less than that of the receptor. Wang et al. developed an AUF-UPLC method to directly determine the KD of compounds in P. frutescens extracts and their target proteins, including the KD determination for α-glucosidase ligands in the ethyl acetate fraction of P. frutescens. The recovery rate, binding degree, and signal-to-noise ratio of α-glucosidase ligands in PFEA were determined using AUF-LC, followed by KD calculation using the proposed equilibrium. Oleanolic acid and apigenin were identified as high-affinity ligands of α-glucosidase, with KDs of 44.9 and 88.5 μM, respectively. These values were consistent with the results from isothermal titration calorimetry, kinetic analysis, and molecular docking simulations. The results demonstrate that this method is simple and easy to implement, allowing direct determination of KD values for compounds in natural product extracts without the need for internal standards or calibration agents. Optimizing these methods can enhance the screening accuracy and reliability of AUF-LC-MS, providing a robust foundation for the identification of active ingredients in complex substrates. In AUF-LC-MS, ultrafiltration membranes separate ligand–receptor complexes from unbound components. The selection of ultrafiltration membranes primarily involves two factors: pore size and material . An ideal ultrafiltration membrane should effectively retain the target biological macromolecules while preventing leakage or clogging. The pore size should be less than one-third of the biomacromolecule’s size to ensure effective retention . Selecting the appropriate pore size improves separation efficiency and prevents leakage of unbound components. An ideal ultrafiltration membrane material should minimize specific adsorption with potential ligands and receptors. Common ultrafiltration membrane materials include polyvinyl fluoride, polysulfone, polyether ketone, and methylcellulose . These materials exhibit low nonspecific binding and are therefore widely used in ultrafiltration membrane production. Selecting the appropriate pore sizes and materials optimizes the separation efficiency of AUF-LC-MS and enhances the accuracy and reliability of the experiment, ensuring the authenticity of ligand–receptor interactions and reducing false-positive results. The complex components, diverse structures, and varying polarities of TCM extracts make it challenging to successfully dissociate ligands from the affinity target while minimizing nonspecific adsorption, a key factor affecting screening results. Two main methods are currently used to denature enzymes: adding acid to the dissociation solvent to inactivate the enzyme in a low pH environment or using organic solvents for enzyme denaturation. However, using organic solvent-based dissociation solutions only can sometimes increase nonspecific adsorption. Some related studies have demonstrated that acid-containing organic solvents, as opposed to those with organic solvents only, can effectively reduce nonspecific adsorption of non-affinity interacting substances. For example, Xie et al. used a methanol–water (90:10) mixture to screen potential TCM components targeting 5-lipoxygenase and cyclooxygenase-2. Comparison of the ultrafiltrate chromatograms between the experimental and control groups revealed significant differences in the peak areas of active ingredients, with lower signals for nonspecifically adsorbed substances. Conversely, some researchers have successfully screened small-molecule inhibitors of cyclooxygenase and glutathione reductase from TCM by using dissociation solutions containing organic solvents only . The findings indicate that different dissociation solutions yield varying effects, necessitating multiple experimental attempts to optimize dissociation conditions. In summary, selecting an appropriate dissociation solvent is essential for reducing nonspecific adsorption and enhancing the accuracy of screening results. Multiple experimental attempts are recommended to identify the optimal dissociation conditions by comparing results, thereby effectively screening the active ingredients in TCM extracts. 5.1. High-Throughput Screening (HTS) of Active Ingredients of TCM Efficient and rapid screening of active ingredients from complex systems, such as TCM, remains a key challenge in modern pharmaceutical research. Traditional methods of chemical separation, structural identification, and activity screening face the following several issues: unclear objectives, cumbersome procedures, high workload, lengthy processes, and potential loss of active ingredients. Recent pharmacological research has demonstrated that the affinity between drugs and biological macromolecules—such as enzymes, receptors, DNA, and RNA—is crucial for drug action. Molecular targeting strategies for drug screening have emerged, focusing on disease-related biological macromolecules as targets. Ultrafiltration offers excellent separation and minimizes matrix interference, whereas LC-MS provides powerful analytical capabilities for the rapid identification of multiple components. Combining these technologies to discover small-molecule active ingredients in TCM holds significant potential. Recently, this combined approach has been successfully applied to the screening of lead compounds, compound libraries, and active ingredients from natural products. Numerous studies have confirmed that this method rapidly screens and identifies complex ligands in natural products . In recent years, scientists have frequently combined AUF with MS detection to screen active ingredients in combinatory chemical libraries, identifying novel inhibitors of key targets like α-glucosidase. α-Glucosidase is a key enzyme in carbohydrate hydrolysis, cleaving the α-1,4-glucoside bond at the non-reducing end of oligosaccharides, thereby releasing glucose and raising blood sugar levels. α-Glucosidase inhibitors reduce glucose production by inhibiting this enzyme’s activity, and they are widely used in the treatment of type 2 diabetes mellitus (T2DM) . Although some α-glucosidase inhibitors derived from microorganisms, such as acarbose and voglibose, are used clinically, they can cause severe gastrointestinal side effects . Natural α-glucosidase inhibitors from medicinal plants offer potential as alternative treatments for T2DM due to their low toxicity. Consequently, researchers have recently screened potential α-glucosidase inhibitors from various natural plants, including Cichorium glandulosum Boiss. et Huet, a chicory species in the Asteraceae family and a traditional Uighur medicinal plant. C. glandulosum is listed as a “medicinal food homology” item in the 2015 Catalogue of Homologous Medicine and Food by the National Health and Family Planning Commission of China. Studies have shown that chicory exhibits significant hypoglycemic activity and inhibits α-glucosidase . Chen et al. used AUF-LC-MS to screen and identify four potential α-glucosidase inhibitors from C. glandulosum seed extract to further investigate its hypoglycemic components. The preliminary identification included esculetin, chlorogenic acid, isochlorogenic acid B, and osochlorogenic acid A. Subsequently, Abudurexiti A et al. used AUF to screen C. glandulosum extracts, identifying the following six potential α-glucosidase inhibitors: quercetin, lactucin, 3- O -methylquercetin, hyperoside, lactucopicrin, and isochlorogenic acid B. Potential α-glucosidase inhibitors have been screened from various natural plants, including the leaves of Rubus suavissimus and Inonotus obliquus and the roots of Siraitia grosvenorii . The screening results of α-glucosidase-targeted active ingredients are detailed in . Medicinal plants have been widely used to treat various diseases for thousands of years owing to their value as natural resources. Extracting biologically active compounds from medicinal plants has become a major focus of research worldwide. Chemical components in medicinal plants often have low abundance, complex structures, and multiple biological targets. The active ingredients and mechanisms of action are often challenging to define precisely. AUF-LC-MS is well suited for screening active ingredients in complex natural products. This technology combines the separation and analytical strengths of AUF and LC-MS, facilitating HTS and rapid identification of bioactive components in complex natural products. Andrographis paniculata (Burm. f.) Wall. ex Nees is derived from the dried aboveground parts of the plant. It exhibits a broad range of pharmacological activities in vivo and in vitro studies, with anti-inflammatory effects being the most prominent. Cyclooxygenase-2 (COX-2) is a key enzyme in prostaglandin (PG) synthesis, and its inhibitors are effective anti-inflammatory agents. Jiao developed an AUF-based analytical method combined with UPLC and quadrupole TOF-MS (BAUF-UPLC-Q-TOF-MS) for rapid screening and identification of COX-2 ligands. Five COX-2 inhibitors were identified from A. paniculata extracts. Apart from its anti-inflammatory properties, A. paniculata exhibits immunomodulatory and antiviral effects. Feng screened 11 potential ligands from A. paniculata targeting COX-2, IL-6, and ACE2. In addition to the previously mentioned disease-related targets, AUF-MS can be used to screen 24 target active ingredients, including lipase, thrombin, and tyrosinase (TYR). Lipase catalyzes the hydrolysis of fats (lipids). Lipase inhibitors regulate lipids by inhibiting the catalytic activity of human pancreatic lipase, a key enzyme in triacylglycerol hydrolysis, aiding in the control or treatment of obesity-related conditions. TYR is a rate-limiting enzyme in melanin production. Albinism is a genetic disorder caused by mutations in the TYR gene, leading to impaired TYR production. Thrombin (FIIα) is a key enzyme in thrombosis and a downstream component of the coagulation pathway. It converts fibrinogen into fibrin and coagulation factor XIII into factor XIIIα. This process combines with calcium ions to form the fibrin network, a critical step in thrombosis. Consequently, FIIα has gained widespread attention as a target for antithrombotic therapies. summarizes the applications of AUF-MS in screening natural product extracts from January 2014 to May 2024. 5.2. Screening of Active Ingredients in TCM Compound Preparations TCM compound preparations are formulated on the basis of TCM theory. Their chemical components are highly complex, making it challenging to rapidly screen and identify active ingredients using conventional analytical methods. Historically, clarifying the bioactive components and mechanisms of action in single medicinal plants has been difficult, let alone in natural drug formulas, due to their low content, complex chemical structures, and multicomponent, multitarget effects. AUF-LC-MS remains one of the most powerful tools for screening active compounds from complex natural products . In recent years, Ronghua Dai’s research group has employed AUF-LC-MS to study the interactions between extracts of Zishen Pills, a TCM compound preparation, and biological target proteins. COX-2 is a key enzyme that catalyzes the conversion of arachidonic acid (AA) into PGs. It is specifically induced during inflammation, degeneration, and tumorigenesis. The research group employed AUF-LC-MS to investigate the interaction between Zishen Pill extract and COX-2, selecting celecoxib and glipizide as positive and negative controls, respectively. The study identified 20 compounds that specifically bind to COX-2, 8 of which are potential COX-2 inhibitors. Their structures were elucidated using Fourier transform ion cyclotron resonance MS. Further validation was conducted using in vitro COX-2 inhibition assays and molecular docking studies. Additionally, the research group further investigated the interaction between Zishen Pills and 5-lipoxygenase (5-LOX) inhibitors . It was found that 5-LOX plays a crucial role in inflammatory processes, and it is a key enzyme in the metabolism of AA to leukotriene A4 (LTA4). The research team optimized the concentration of 5-LOX enzyme, incubation conditions (temperature and time), pH, and ionic strength based on prior experiments to achieve more accurate screening results. The screening results indicated that six compounds may possess potential 5-LOX inhibitory activity, with anemarrhenasaponin I, timosaponin AI, nyasol, and demethyleneberberine demonstrating significant enzyme inhibition. Further, structure–activity relationship studies revealed that the hydroxyl group is essential for ligand binding to the 5-LOX protein, followed by the aromatic ring, which engages in π–π interactions with amino acid residues in the 5-LOX protein. This study provides a scientific foundation for the development of 5-LOX inhibitors. Efficient and rapid screening of active ingredients from complex systems, such as TCM, remains a key challenge in modern pharmaceutical research. Traditional methods of chemical separation, structural identification, and activity screening face the following several issues: unclear objectives, cumbersome procedures, high workload, lengthy processes, and potential loss of active ingredients. Recent pharmacological research has demonstrated that the affinity between drugs and biological macromolecules—such as enzymes, receptors, DNA, and RNA—is crucial for drug action. Molecular targeting strategies for drug screening have emerged, focusing on disease-related biological macromolecules as targets. Ultrafiltration offers excellent separation and minimizes matrix interference, whereas LC-MS provides powerful analytical capabilities for the rapid identification of multiple components. Combining these technologies to discover small-molecule active ingredients in TCM holds significant potential. Recently, this combined approach has been successfully applied to the screening of lead compounds, compound libraries, and active ingredients from natural products. Numerous studies have confirmed that this method rapidly screens and identifies complex ligands in natural products . In recent years, scientists have frequently combined AUF with MS detection to screen active ingredients in combinatory chemical libraries, identifying novel inhibitors of key targets like α-glucosidase. α-Glucosidase is a key enzyme in carbohydrate hydrolysis, cleaving the α-1,4-glucoside bond at the non-reducing end of oligosaccharides, thereby releasing glucose and raising blood sugar levels. α-Glucosidase inhibitors reduce glucose production by inhibiting this enzyme’s activity, and they are widely used in the treatment of type 2 diabetes mellitus (T2DM) . Although some α-glucosidase inhibitors derived from microorganisms, such as acarbose and voglibose, are used clinically, they can cause severe gastrointestinal side effects . Natural α-glucosidase inhibitors from medicinal plants offer potential as alternative treatments for T2DM due to their low toxicity. Consequently, researchers have recently screened potential α-glucosidase inhibitors from various natural plants, including Cichorium glandulosum Boiss. et Huet, a chicory species in the Asteraceae family and a traditional Uighur medicinal plant. C. glandulosum is listed as a “medicinal food homology” item in the 2015 Catalogue of Homologous Medicine and Food by the National Health and Family Planning Commission of China. Studies have shown that chicory exhibits significant hypoglycemic activity and inhibits α-glucosidase . Chen et al. used AUF-LC-MS to screen and identify four potential α-glucosidase inhibitors from C. glandulosum seed extract to further investigate its hypoglycemic components. The preliminary identification included esculetin, chlorogenic acid, isochlorogenic acid B, and osochlorogenic acid A. Subsequently, Abudurexiti A et al. used AUF to screen C. glandulosum extracts, identifying the following six potential α-glucosidase inhibitors: quercetin, lactucin, 3- O -methylquercetin, hyperoside, lactucopicrin, and isochlorogenic acid B. Potential α-glucosidase inhibitors have been screened from various natural plants, including the leaves of Rubus suavissimus and Inonotus obliquus and the roots of Siraitia grosvenorii . The screening results of α-glucosidase-targeted active ingredients are detailed in . Medicinal plants have been widely used to treat various diseases for thousands of years owing to their value as natural resources. Extracting biologically active compounds from medicinal plants has become a major focus of research worldwide. Chemical components in medicinal plants often have low abundance, complex structures, and multiple biological targets. The active ingredients and mechanisms of action are often challenging to define precisely. AUF-LC-MS is well suited for screening active ingredients in complex natural products. This technology combines the separation and analytical strengths of AUF and LC-MS, facilitating HTS and rapid identification of bioactive components in complex natural products. Andrographis paniculata (Burm. f.) Wall. ex Nees is derived from the dried aboveground parts of the plant. It exhibits a broad range of pharmacological activities in vivo and in vitro studies, with anti-inflammatory effects being the most prominent. Cyclooxygenase-2 (COX-2) is a key enzyme in prostaglandin (PG) synthesis, and its inhibitors are effective anti-inflammatory agents. Jiao developed an AUF-based analytical method combined with UPLC and quadrupole TOF-MS (BAUF-UPLC-Q-TOF-MS) for rapid screening and identification of COX-2 ligands. Five COX-2 inhibitors were identified from A. paniculata extracts. Apart from its anti-inflammatory properties, A. paniculata exhibits immunomodulatory and antiviral effects. Feng screened 11 potential ligands from A. paniculata targeting COX-2, IL-6, and ACE2. In addition to the previously mentioned disease-related targets, AUF-MS can be used to screen 24 target active ingredients, including lipase, thrombin, and tyrosinase (TYR). Lipase catalyzes the hydrolysis of fats (lipids). Lipase inhibitors regulate lipids by inhibiting the catalytic activity of human pancreatic lipase, a key enzyme in triacylglycerol hydrolysis, aiding in the control or treatment of obesity-related conditions. TYR is a rate-limiting enzyme in melanin production. Albinism is a genetic disorder caused by mutations in the TYR gene, leading to impaired TYR production. Thrombin (FIIα) is a key enzyme in thrombosis and a downstream component of the coagulation pathway. It converts fibrinogen into fibrin and coagulation factor XIII into factor XIIIα. This process combines with calcium ions to form the fibrin network, a critical step in thrombosis. Consequently, FIIα has gained widespread attention as a target for antithrombotic therapies. summarizes the applications of AUF-MS in screening natural product extracts from January 2014 to May 2024. TCM compound preparations are formulated on the basis of TCM theory. Their chemical components are highly complex, making it challenging to rapidly screen and identify active ingredients using conventional analytical methods. Historically, clarifying the bioactive components and mechanisms of action in single medicinal plants has been difficult, let alone in natural drug formulas, due to their low content, complex chemical structures, and multicomponent, multitarget effects. AUF-LC-MS remains one of the most powerful tools for screening active compounds from complex natural products . In recent years, Ronghua Dai’s research group has employed AUF-LC-MS to study the interactions between extracts of Zishen Pills, a TCM compound preparation, and biological target proteins. COX-2 is a key enzyme that catalyzes the conversion of arachidonic acid (AA) into PGs. It is specifically induced during inflammation, degeneration, and tumorigenesis. The research group employed AUF-LC-MS to investigate the interaction between Zishen Pill extract and COX-2, selecting celecoxib and glipizide as positive and negative controls, respectively. The study identified 20 compounds that specifically bind to COX-2, 8 of which are potential COX-2 inhibitors. Their structures were elucidated using Fourier transform ion cyclotron resonance MS. Further validation was conducted using in vitro COX-2 inhibition assays and molecular docking studies. Additionally, the research group further investigated the interaction between Zishen Pills and 5-lipoxygenase (5-LOX) inhibitors . It was found that 5-LOX plays a crucial role in inflammatory processes, and it is a key enzyme in the metabolism of AA to leukotriene A4 (LTA4). The research team optimized the concentration of 5-LOX enzyme, incubation conditions (temperature and time), pH, and ionic strength based on prior experiments to achieve more accurate screening results. The screening results indicated that six compounds may possess potential 5-LOX inhibitory activity, with anemarrhenasaponin I, timosaponin AI, nyasol, and demethyleneberberine demonstrating significant enzyme inhibition. Further, structure–activity relationship studies revealed that the hydroxyl group is essential for ligand binding to the 5-LOX protein, followed by the aromatic ring, which engages in π–π interactions with amino acid residues in the 5-LOX protein. This study provides a scientific foundation for the development of 5-LOX inhibitors. The fingerprint analysis of active ingredients in TCM is crucial for quality control and evaluation. Although traditional chemical fingerprints can reflect the overall characteristics of TCM, they are limited because the selected chemical components may not correspond directly to those that produce clinical effects. Therefore, integrating high-throughput screening technologies, such as AUF-LC-MS, to identify active ingredients in TCM and further obtain their biological fingerprints can address the limitations of chemical fingerprints and offer a novel approach for evaluating the efficacy of TCM. Recently, Mingquan Guo’s research group has made significant progress in studying Rhamnus davurica Pall. by using AUF-LC-MS. They established an AUF-LC-MS-based method to successfully screen and identify ligands in R. davurica that are potentially active against therapeutic targets like top I and COX-2 . The study identified 12 potential top I ligands and 11 potential COX-2 ligands, further demonstrating that these components exhibit anti-inflammatory and anti-proliferative activities in vitro. This study not only proposes a novel method to reveal the diverse active ingredients of TCM and their potential targets but also underscores the importance of biological fingerprint analysis in TCM research. By integrating bioaffinity technology with MS, the characteristics of active ingredients in TCM can be understood more comprehensively, providing a more scientific basis for its quality control. This approach not only addresses the limitations of traditional chemical fingerprinting but also enhances the accuracy of TCM efficacy evaluation. Future research should focus on exploring the biological fingerprints of various TCMs to advance the quality standardization and modernization of TCM, ultimately supporting its broader application in clinical practice. In the analysis of small-molecule drug metabolites, modern analytical methods are diverse and highly efficient, with LC-MS being particularly prominent. This technology not only efficiently separates and detects drugs and their metabolites but also provides detailed structural information and supports metabolic pathway research, significantly advancing the fields of pharmacokinetics and pharmacodynamics. AUF-LC-MS, a pretreatment technique, has been widely applied in drug metabolism research. This method combines ultrafiltration technology with online LC-MS analysis to rapidly and efficiently assess the metabolic rate and extent of drugs at affinity targets like liver microsomes. Van Breemen and colleagues successfully used AUF-LC-MS to evaluate the metabolic characteristics of tricyclic psychotropic drugs like promethazine and to reveal the structural features of their main metabolites. Huang et al. demonstrated the potential of AUF-LC-MS in studying the pharmacological activity of natural products. They employed this technique to screen for potential lipoxygenase inhibitors in Saposhnikovia divaricata (Trucz.) Schischk. They also identified multiple metabolic pathways by using semi-preparative HPLC separation and in vitro cytochrome P450 metabolism studies, offering new approaches for evaluating the medicinal value of natural products. Methodologically, the advantage of AUF-LC-MS lies in its simplicity and high-throughput capabilities, making it particularly suitable for metabolite analysis and structural identification of complex samples. This technology not only allows researchers to quickly obtain pharmacokinetic data but also provides valuable structure–activity relationship information during drug design and optimization. In recent years, TCM has demonstrated unique advantages in treating complex diseases owing to its multicomponent and multitarget characteristics. Traditional methods often struggle to analyze the chemical components and pharmacological mechanisms of TCM. Bioaffinity MS offers a novel approach to address this issue. Notably, AUF-LC-MS has been widely applied in screening active ingredients in TCM owing to its high efficiency and simplicity. AUF-LC-MS shows significant potential in TCM research. This approach involves combining medicinal plant extracts with specific protein targets, using ultrafiltration to separate the conjugates, and then identifying the bound active ingredients through LC-MS. Recent studies have shown that the AUF-LC-MS yielded remarkable results in screening targets such as α-glucosidase, cyclooxygenase-2, and thrombin. This study found that compounds isolated from traditional Chinese medicine by using this method exhibited excellent enzyme inhibitory activity, with high selectivity and specificity. Although the AUF-LC-MS method holds promising prospects for screening active ingredients in TCM, it still possesses limitations and faces various challenges. First, given the complex nature of TCM compounds, molecular interactions may compromise analytical accuracy. Future efforts might incorporate computational biology techniques to predict and confirm inter-component interactions, thereby enhancing the accuracy of screening and analysis for potentially active components. Secondly, employing multi-stage ultrafiltration membranes or a series of ultrafiltration tubes could facilitate the development of multichannel or high-throughput AUF systems, significantly enhancing the efficiency and precision of multi-target screening. Additionally, despite the therapeutic benefits of TCM volatiles like monoterpenes and sesquiterpenes, their volatile and low-density nature leads to immobilization and trapping challenges during the AUF process. Improvements might be achieved by employing tightly sealed reaction vessels, developing specialized ultrafiltration membranes, or operating in low-temperature conditions to minimize the loss of volatile components. Given the unique characteristics of volatile components, integrating auxiliary technologies such as gas chromatography for pretreatment or post-treatment might improve screening accuracy and efficiency. Moreover, false positives and nonspecific binding restrict the wider application of this technique. Utilizing enzyme denaturation controls, together with enzyme activity assays and molecular docking, can significantly mitigate nonspecific binding and boost screening accuracy. Considering that MS and LC analysis tools have become more miniaturized and automated, the application of AUF-LC-MS is expected to become more widespread and in-depth. In the future, the use of this technology in screening TCM active ingredients should extend beyond common targets to include more significant protein targets. This approach could facilitate the discovery of new drugs and enhance the understanding of the pathogenesis of complex diseases. Therefore, AUF-MS is a powerful tool for identifying and studying the mechanisms of active ingredients in TCM. With ongoing innovations and improvements, this method is likely to play a more significant role in natural product research and new drug development. Future research should focus on overcoming current technical bottlenecks and identifying more disease-related protein targets, which would advance modern research on TCM. |
Virtual supervision in ophthalmology: a scoping review | 524ffb48-7bcd-4a87-9006-f0a29e6a6511 | 10074343 | Ophthalmology[mh] | The use of telemedicine (TM) in the USA increased dramatically in response to the COVID-19 public health emergency . One application of TM is virtual supervision (VS), where the supervisor and supervisee who are not in the same physical location interact for an episode of patient care via synchronous audio and/or video modalities .Virtual supervision can facilitate cost-effective mentoring by specialists from remote locations and is also useful as a teaching tool in graduate medical education (GME) . The information available in the literature on VS in ophthalmology is not well described. This scoping review aims to better understand the evidence and potential role for VS in ophthalmic practice and education. In consultation with a reference librarian, a literature search strategy was developed in accordance with Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) . We searched six databases (PubMed, Embase, Web of Science, CINAHL, PsycINFO, and ERIC) for articles published between January 1950 and August 2022. Our search included database-specific thesaurus terms such as medical subject headings (MeSH) and Emtree, as well as keywords related to VS in ophthalmology (Table ). The inclusion criteria were full-text articles published in an English-language peer-reviewed journal. We included studies that involved VS between physicians or between physicians and trainees in ophthalmology. We excluded studies in which the supervisor and supervisee were in the same physical location and studies in which clinical supervision was carried out retrospectively. We also excluded studies that included supervisors who were not ophthalmology physicians. Eligible studies were de-duplicated in EndNote (Clarivate Analytics, Philadelphia, PA) and imported into the systematic review software Covidence (Melbourne, Victoria, Australia) for screening, full-text review, and data extraction. The screening and selection process is displayed in a PRISMA flowchart (Fig. ). Two investigators (CK and CS) independently conducted title/abstract screening, full-text review, and data extraction in Covidence. Disagreements were resolved by the senior investigator (PBG). A data template was developed in Covidence to extract relevant information, including year of publication, location, study design, characteristics of study participants, sample size, and any type of outcome. Two investigators (CK and CS) independently appraised the methodological quality of the studies using the Mixed Methods Appraisal Tool (MMAT), and disagreements were resolved by the senior investigator (PBG). The Mixed Methods Appraisal Tool (MMAT) appraises studies based on five questions assessing the sampling strategy, outcome measurement, confounders, and statistical analysis of the study . All studies were scored on a scale of 1 to 5 based on the MMAT. Study characteristics The initial database search yielded a total of 2700 articles. Following duplicate removal, title and abstract screening, and full-text review, seven articles were included in our qualitative synthesis (Fig. ): four were case reports or case series (57%) , two were cohort studies (29%) , and one was a cross-sectional study (14%) (Table ) . The publication year of the articles ranged from 1998 to 2022. Most studies were conducted in the USA. Other studies were conducted in Finland , Israel , the UK , and the Philippines . Three studies were prospective in design (43%) , while other studies were retrospective (14%) or both (14%) . The objective of the studies ranged from determining the feasibility of a teleconference device (43%) , describing the use of VS in an emergency department (ED) or rural hospital (43%) , or determining whether fellows could safely operate independently with remote supervision (14%) . A total of 619 patients and 367 supervisees were included (Table ). Two studies included supervisees who were physicians, including an ophthalmic surgeon from a foreign country (14%) and a general practitioner in a rural hospital (14%) . Five studies included supervisees who were medical trainees, such as ophthalmology residents (43%) , vitreoretinal fellows (14%) , and emergency medicine (EM) residents (14%) . Study setting varied between the ED , operating room (OR) , rural hospital , and eye clinic . Virtual supervision was used for a clinical examination in three studies (43%) , surgical procedure in three studies (43%) , and in-office procedure in one study (14%) . For studies involving a surgical or in-office procedure, VS was used to teach surgeons endoscopic laser-assisted dacryocystorhinostomy (ELA-DCR) , perform orbital decompression on a patient with extensive facial trauma , remove a corneal foreign body and rust ring in a rural hospital , and supervise retina fellows repairing macula-on rhegmatogenous retinal detachments (RRD) during bank holidays and weekends . Devices used for VS included an optical coherence tomography (OCT)/fundus camera , slit lamp camera , network camera , and ceiling-mounted camera (Table ) . Synchronous communication between the supervisor and supervisee was facilitated through various modalities such as email , telephone , video conferencing software , and local area network (LAN) . Four studies reported successful transmission of real-time video of surgical or in-office procedures . In three studies on VS for clinical examinations, supervisees held synchronous communication with their supervisors to reach an agreement in diagnosis . Studies noted that VS was helpful for diagnosing ocular surface, anterior segment, and macular diseases but difficult to use when diagnosing vitreous and peripheral retinal conditions . One study noted that the slit lamp camera was sufficient to provide high enough real-time image quality for procedures , while another study using a network camera noted decreased transmission rate due to electronic traffic . Several studies also noted machine malfunction, image artifacts, insufficient image resolution, and lighting problems as factors that decreased the effectiveness of VS . One study reported patient outcomes related to VS and found no significant difference in 6-month retinal re-detachment rate when comparing remotely supervised to unsupervised fellows performing RRD repairs on bank holidays and weekends . Only one study reported educational outcomes. In this study, four out of eight residents stated that VS helped them diagnose patients more accurately, as they could synchronously share the images with the supervisor and talk through their thought processes . Critical appraisal of included studies The MMAT was used to critically appraise the quality of the studies. Limitations were noted in outcome measurement , statistical analysis , sampling strategy , and inclusion of confounding factors . Overall, two studies scored a 5/5 , one study scored a 4/5 , and four studies scored a 3/5 . The initial database search yielded a total of 2700 articles. Following duplicate removal, title and abstract screening, and full-text review, seven articles were included in our qualitative synthesis (Fig. ): four were case reports or case series (57%) , two were cohort studies (29%) , and one was a cross-sectional study (14%) (Table ) . The publication year of the articles ranged from 1998 to 2022. Most studies were conducted in the USA. Other studies were conducted in Finland , Israel , the UK , and the Philippines . Three studies were prospective in design (43%) , while other studies were retrospective (14%) or both (14%) . The objective of the studies ranged from determining the feasibility of a teleconference device (43%) , describing the use of VS in an emergency department (ED) or rural hospital (43%) , or determining whether fellows could safely operate independently with remote supervision (14%) . A total of 619 patients and 367 supervisees were included (Table ). Two studies included supervisees who were physicians, including an ophthalmic surgeon from a foreign country (14%) and a general practitioner in a rural hospital (14%) . Five studies included supervisees who were medical trainees, such as ophthalmology residents (43%) , vitreoretinal fellows (14%) , and emergency medicine (EM) residents (14%) . Study setting varied between the ED , operating room (OR) , rural hospital , and eye clinic . Virtual supervision was used for a clinical examination in three studies (43%) , surgical procedure in three studies (43%) , and in-office procedure in one study (14%) . For studies involving a surgical or in-office procedure, VS was used to teach surgeons endoscopic laser-assisted dacryocystorhinostomy (ELA-DCR) , perform orbital decompression on a patient with extensive facial trauma , remove a corneal foreign body and rust ring in a rural hospital , and supervise retina fellows repairing macula-on rhegmatogenous retinal detachments (RRD) during bank holidays and weekends . Devices used for VS included an optical coherence tomography (OCT)/fundus camera , slit lamp camera , network camera , and ceiling-mounted camera (Table ) . Synchronous communication between the supervisor and supervisee was facilitated through various modalities such as email , telephone , video conferencing software , and local area network (LAN) . Four studies reported successful transmission of real-time video of surgical or in-office procedures . In three studies on VS for clinical examinations, supervisees held synchronous communication with their supervisors to reach an agreement in diagnosis . Studies noted that VS was helpful for diagnosing ocular surface, anterior segment, and macular diseases but difficult to use when diagnosing vitreous and peripheral retinal conditions . One study noted that the slit lamp camera was sufficient to provide high enough real-time image quality for procedures , while another study using a network camera noted decreased transmission rate due to electronic traffic . Several studies also noted machine malfunction, image artifacts, insufficient image resolution, and lighting problems as factors that decreased the effectiveness of VS . One study reported patient outcomes related to VS and found no significant difference in 6-month retinal re-detachment rate when comparing remotely supervised to unsupervised fellows performing RRD repairs on bank holidays and weekends . Only one study reported educational outcomes. In this study, four out of eight residents stated that VS helped them diagnose patients more accurately, as they could synchronously share the images with the supervisor and talk through their thought processes . The MMAT was used to critically appraise the quality of the studies. Limitations were noted in outcome measurement , statistical analysis , sampling strategy , and inclusion of confounding factors . Overall, two studies scored a 5/5 , one study scored a 4/5 , and four studies scored a 3/5 . There is a limited evidence base on VS in ophthalmology. The few available studies suggest that VS is technologically feasible, may provide a positive experience for both supervisees and patients, and can permit synchronous communication and transmission of clinical data, which can be used to formulate diagnosis and management plans or learn new surgical skills. Virtual supervision can improve diagnostic accuracy by allowing fast, synchronous communication between supervisor and trainee. In one study, ophthalmology residents (training level unspecified) used VS to consult their supervisors regarding complicated cases during night shifts in the ED . The residents communicated with their supervisors via telephone and emailed images and videos captured by a miniature slit lamp camera. The authors found high agreement in diagnosis made during the night shift and the on-site examination by the supervisor the next day. In another study, first-year ophthalmology residents sent OCT/fundus photographs to the supervisor to synchronously discuss a differential diagnosis . The use of the device did not correlate with a longer duration of visit in the ED. Virtual supervision can also impact ophthalmic surgical GME. In one study, retina fellows who were deemed ready to operate independently by their supervisors performed surgery while supervisors provided synchronous feedback from a remote location during holidays and weekends . This readiness was assessed on a case-by-case basis by the supervisors after 2–3 months of supervised training based on a review of recorded surgeries and didactics. The study included only primary uncomplicated macula-on RRDs; patients with more complex retinal detachments were excluded. The study reported no statistical difference in the 6-month retinal re-detachment rate between off-hour cases utilizing VS and off-hour cases without any supervision. The authors did not mention if there were any adverse events, difficulties with VS, or a safety plan if a significant complication were to occur during the surgery. In addition, ophthalmologists can use VS to assist physicians in areas with limited access to ophthalmic care or continuing medical education (CME). In one study, a general practitioner in a remote area provided time-sensitive ophthalmic procedures with the assistance of synchronous communication with ophthalmologists . In another study, ophthalmologists in the USA were able to teach a new procedure to ophthalmologists in the Philippines who had the necessary equipment but not the surgical expertise . This review suggests that various methods can be used to ensure high image and video quality during VS, although challenges remain. In one study, patients were instructed to keep voluntary movements to a minimum and to fixate on a target, and a designated remote operator kept the slit lamp image focused . Despite these methods, there were technological glitches, image artifacts, and internet connection issues. Most studies reported that quality of images and videos were adequate for clinical diagnosis and procedure guidance, although in some cases, supervisees needed to spend a significant amount of time to find the optimal setup . Furthermore, the blue filter on the slit lamp with fluorescein was particularly difficult to visualize during VS due to degradation of the video quality . We identified several gaps in the literature on the use of VS in ophthalmology. First, most studies on ophthalmic GME lacked standardized curricula to implement the interventions and methodology to assess the outcomes. Surveys were often used to assess feasibility and usability of VS; only one study collected data from all supervisors, supervisees, and patients . Furthermore, no studies included control groups, which makes it difficult to compare the effectiveness of VS versus direct supervision in GME. Second, most studies focused on feasibility of technology for VS rather than on clinical or educational outcomes. One study that assessed educational outcomes relied on self-reported data without a comparison group . Third, there was significant heterogeneity in the types of supervisees such as residents, fellows, general practitioners, and physicians. Supervisees have varying needs based on their background and training experience, which makes it difficult to generalize the results . This review has several limitations. First, we did not include studies published in gray literature or in languages other than English. Second, the reproducibility of MMAT ratings is limited by the authors’ judgments about the quality of the study design. In conclusion, this scoping review highlights the potential utility of VS in ophthalmology and identifies areas for future research on this topic. Recent technological advancements allow supervisors to guide other physicians and trainees through VS, with potentially improved efficiency and collaboration between different institutions and countries. Future studies should employ larger sample sizes and more rigorous methodology to better define the safety and efficacy of VS in ophthalmic practice and education . |
Artificial intelligence in rheumatology research: what is it good for? | 7d3c40d1-83d7-4857-9752-1bebfc080dae | 11748787 | Internal Medicine[mh] | Discriminative artificial intelligence (AI) is advancing rheumatology with machine learning models that enhance disease diagnosis and prediction by analysing structured data, imaging data and text. Generative AI, using large language models, may significantly support research by assisting the process and refining study development via general and specialised chatbots, although its application in rheumatology is still in early development. To fully harness AI’s potential in rheumatology research, it is crucial to balance innovation with responsibility, ensuring robust methodologies and the preservation of research integrity. Artificial intelligence (AI) has emerged as a transformative technology in medicine, providing rheumatology with innovative tools for research. AI, known as the capability of computational systems to perform tasks that typically require human intelligence, include learning patterns from prior data, understanding natural language, perception, reasoning, problem-solving. The impact of this technology in health sciences research is increasingly evident, with multiple applications gradually being integrated into the field of rheumatology. Indeed, different algorithms have led to the development of models for the diagnosis, evaluation, prognosis and prediction of disease. As AI has evolved, it has become increasingly important to differentiate between discriminative AI, widely used for studies on disease classification and prediction, and the more recently emergent generative AI, which holds promise for novel applications in research like hypothesis generation, clinical trial design, drug development, literature synthesis and writing support. Discriminative and generative AI differ in how they process data and apply their learning algorithms. While discriminative models focus on finding decision limits to predict labels, generative models analyse the underlying data distribution aiming to generate new data. summarises the main models used by these technologies and applications for research that we will explore in this review. Discriminative AI includes a wide range of capabilities, such as distinguishing data to make classifications or predictions, as well as performing tasks like outlier detection and clustering. Radiology exemplifies the significant impact of discriminative AI. As a matter of fact, 723 of the 950 (76%) AI/ML (machine learning)-enabled medical devices approved by the Food and Drug Administration (FDA) as of August 2024 are related to this specialty. Other notable examples can be found in ophthalmology, where algorithms have shown the ability not only to diagnose ocular pathologies with greater accuracy than expert ophthalmologists but also to predict cardiovascular risk factors undetectable by fundus examination. Discriminative AI models have proven effective beyond image analysis, extending to fields like linguistics and other data types. For instance, a model used voice recordings and demographic data to predict dementia onset in patients with mild cognitive impairment, achieving around 80% accuracy. Although no FDA-approved AI/ML applications currently exist in rheumatology, numerous promising studies within the field of discriminative AI will be discussed in further detail. Generative AI has recently transformed the AI landscape, particularly following the release of the chatbot ChatGPT in 2022, which has made AI more accessible to the general public. Generative AI can create new content based on existing data from various sources, including text generation, image or video creation. The tools based on this technology have shown promising applications in medicine, including demonstrating clinical knowledge by successfully achieving high accuracy in standardised examinations. Beyond knowledge-based tasks, ChatGPT’s responses to patient questions were shown to be often preferred to those by doctors for their quality and empathy. These results are remarkable given that the ChatGPT model is general-purpose and not specifically designed for medicine. The accuracy and reasoning skills of large language models (LLMs) have also been demonstrated in clinical examinations. Research has led to the development of specialised medical models like Med-Pathways Language Model (PaLM2), which achieved an accuracy similar to clinician answers (both exceeding 90%) in answering medical questions after being trained on six medical question-answering datasets. Rheumatology is a rapidly evolving specialty, thanks to the advent of advanced therapies and new technologies. Focusing on the management of chronic diseases with potential systemic involvement, it is a field rich in data and complex decision-making. Therefore, the use of AI tools holds the promise to transform clinical practice, leading to more informed decision making. Beyond the diagnostic level, there are promising predictive capabilities. In this regard, AI can assess disease activity, predict flares, determine optimal treatment dosages and anticipate patient responses based on clinical and serological biomarkers. Moreover, generative AI can function as a clinical decision support system, assist with administrative tasks, and enhance the quality of patient information and education. The aim of this review is to highlight the current and future applications of AI in rheumatology, examine the mechanisms of AI, analyse state-of-the-art investigations and explore its integration into daily research practice. To achieve this, we conducted a narrative review including an electronic search in Medline and Embase for English-language sources from inception to September 2024. We employed a range of free-text terms including, but not limited to: “Artificial intelligence AND rheumatology”, “Machine learning AND rheumatology”, “Deep learning AND rheumatology”, “(Machine learning OR Deep learning) AND (rheumatoid arthritis OR spondyloarthritis OR psoriatic arthritis OR osteoarthritis OR lupus OR Sjogren)”, “Large language models”, “Natural language processing”, “(Predictive modeling OR Electronic medical records OR Risk stratification) AND rheumatology”, “(Large language models OR Natural language processing) AND rheumatology”, “ChatGPT AND rheumatology”. Furthermore, we conducted a manual search by examining the references cited in the included studies and technical computer science books. Priority was given to seminal references or those published within the last 2 years. The integration of AI into rheumatology research is becoming increasingly relevant, as the availability of complex datasets and advanced computation redefine how we approach and conduct scientific investigations. Given its capacity for use in research, AI-related concepts can help rheumatologists to effectively use these technologies in their work. summarises the core principles in the most widely used AI algorithms, as well as their application in rheumatology. An AI algorithm is a computational model designed to perform tasks by learning from data and identifying patterns, rather than relying solely on a predefined set of rules or instructions. These algorithms may improve their performance over time through experience, which is gained through an iterative process. These algorithms are typically used for classification or predictive purposes, such as diagnostic or prescriptive applications in medicine. This can be achieved by analysing large datasets, identifying relevant features and applying learnt patterns to new, unseen data. ML is a branch of AI that operates by feeding an algorithm with input data that reflects past observations, enabling it to construct a model to assess new, previously unseen observations. ML algorithms can be classified into four main types according to their training: supervised, unsupervised, self-supervised and reinforcement learning. Supervised algorithms are trained on a dataset where the output results are known and are used to label the outcomes. These have been the most used for clinical research. Unsupervised algorithms work with unlabelled data to identify patterns or clusters within datasets, making them useful for exploratory data analysis. There are two main types: clustering, which groups similar data points (eg, K-means, DBSCAN, hierarchical clustering), and dimensionality reduction, which simplifies data by reducing the number of features while preserving essential information (eg, PCA, t-SNE). These methods help uncover hidden structures without the need for labelled examples. Self-supervised learning creates internal labels within an unlabelled dataset, allowing models to learn without external annotation and guidance. Reinforcement learning adapts dynamically using reward-based feedback to maximise the performance of the algorithm. Deep learning (DL) is a subtype of ML that involves neural networks. A neural network is a particular ML algorithm based on successive layers of data transformation, inspired by the neural connections in the human brain. Neural networks are particularly effective with large volumes of data and demand significant processing power, which can be provided by processing units working in parallel. DL uses a high number of neuron layers, allowing for multiple levels of abstraction and has achieved noteworthy results in various applications, including text and image recognition. Transfer learning enables the adaptation of a DL model to specific imaging tasks (eg, rheumatological imaging classification) by leveraging pre-existing knowledge from extensive, non-specialised image datasets, enhancing model performance and reducing the need for large, specialised training datasets. Applications of DL include image recognition and natural language processing (NLP), which use images and text as input data, respectively. Indeed, one type of deep neural network algorithm gave birth to transformer technology. Transformers have revolutionised NLP with the so-called self-attention mechanism, which allows for capturing relations between words, allowing for efficient and accurate text generation. The seminal paper on this technology has garnered 140 000 citations by November 2024, reflecting the significant impact of language models on society. LLMs are advanced neural networks based on the transformer architecture. They are pretrained on vast amounts of unlabelled text data, typically sourced from the web, using self-supervised learning. This self-supervised learning involves predicting the next word in a sentence given the previous words (context), for which the model uses the surrounding context as the signal to learn and improve. These models are fine-tuned for specific tasks like question-answering and named entity recognition, showing their versatility and effectiveness in language understanding and generation. When models are able to process and integrate multiple types of data such as text, images and audio, this is known as multimodality. Validation is a process that ensures a model’s generalisability and reliability by assessing its performance on unseen data before it is deployed in real-world applications. Evaluating discriminative AI models involves metrics familiar to rheumatologists, such as sensitivity (also known as recall in the field of ML) and specificity, which assess the ability to correctly identify true positives and true negatives. Precision, similar to positive predictive value (PPV), measures the proportion of true positives among all positive predictions, while the F1 score combines precision and recall into a single measure. Accuracy, which represents the overall correctness of the model’s predictions, reflects the proportion of true results (both true positives and true negatives) among the total cases. The area under the receiver operating characteristic curve (AUC-ROC) offers a visual summary of the model’s performance across thresholds, with the area reflecting the ability of the model to distinguish between classes. In evaluating generative AI models, additional metrics provide an understanding of model performance besides accuracy. Perplexity measures the model’s ability to predict the next word in a sequence, with lower scores indicating more precise predictions. Bilingual Evaluation Understudy assesses the similarity between the model-generated text and a human reference by comparing overlapping word sequences. Recall-Oriented Understudy for Gisting Evaluation measures the degree of n-gram overlap between generated and reference texts, emphasising recall. BERTScore further enhances evaluation by using Bidirectional Encoder Representations from Transformers (BERT) embeddings to compare the semantic similarity between generated and reference texts, capturing context-sensitive alignment. These complementary metrics allow for a comprehensive assessment of generative AI in clinical applications. Discriminative AI algorithms are focused on two main objectives. Classification analysis aims to evaluate previously described phenomena, attempting to describe features and ideally associations between risk factors (independent or predictor variables) and outcomes (dependent variables or events). On the other hand, predictive analysis, aims to forecast future events. Traditionally, this has been achieved using regression methodologies, including linear, logistic or Cox regression. Recently, AI has been employed to classify diseases and predict their progression using ML algorithms. These approaches may use various data types, including structured data, images and free-text information. Interpreting performances across studies is complex due to variations in datasets, patient cohorts and study outcomes, making direct comparison challenging. While metrics such as AUC and F1 metrics offer insights into model performance, their practical value depends on whether these AI advances lead to real-world clinical benefits. Validations against conventional models and interventions remain essential to establish AI’s utility and ensure its impact on patient care. The application of AI in analysing structured data is advancing diagnostic accuracy, risk prediction and patient management in rheumatic and RMDs. In rheumatoid arthritis (RA), for example, a neural network model trained on demographic and laboratory data (including age, sex, rheumatoid factor, anti-citrullinated cyclic peptide and anti-carbamylated protein) achieved an F1 score of 0.92 in diagnosing RA, demonstrating accuracy comparable to, or exceeding, conventional diagnostic approaches. In predicting difficult-to-treat (D2T) RA, an extreme gradient boosting (XGBoost) model combined structured and unstructured data from 1873 patients, achieving an AUC-ROC of 0.88 for D2T identification and 0.73 for future D2T development prediction. In combination with structured data, integrating unstructured data—such as clinical notes and imaging—further enhances AI models’ ability to predict complex outcomes, as demonstrated in the previous study identifying RA subsets like D2T RA. Collectively, these applications underscore AI’s potential to exceed or complement standard statistical approaches by improving accuracy, sensitivity and specificity across RMD diagnostic and prognostic tasks. For prognostic applications, structured data analyses have provided insights across various RMDs. In spondyloarthritis (SpA), K-means clustering applied to a longitudinal dataset identified two distinct disease activity trajectories—one with persistently high activity and another evolving to low activity—highlighting potential therapeutic approaches based on trajectory patterns. Similarly, in systemic lupus erythematosus (SLE), a random forest (RF) model predicted hospitalisations with an AUC-ROC of 0.75, using clinical markers such as dsDNA positivity, C3 levels, blood cell counts, inflammatory markers and albumin. Additionally, in pregnancy outcomes for women with SLE, a pre-pregnancy RF model achieved an AUC-ROC of 0.92, demonstrating high sensitivity (0.89) and specificity (0.94) in identifying adverse outcomes, a notable improvement compared with traditional models. The use of structured data and their analysis through AI has become an important axis in drug development and molecule generation, mainly based on ML and DL algorithms. Some of the use cases aim to identify drug targets and binding sites as well as to predict chemical properties (affinity, ability, lipophilicity, solubility, toxicity) of a compound. ML and DL algorithms may be used for efficacy evaluations of drugs through big data modelling and analysis. A relevant advancement in this regard has been conducted by AlphaFold, developed by DeepMind, in predicting structures of proteins. This enabled researchers to understand molecular targets more precisely, therefore supporting in identifying binding sites, refining drug designs and predicting protein interactions. Additionally, AI may assist clinical trial design and implementation by supporting selection of promising lead molecules based on patient-specific profiles, identifying suitable patient profiles and improving recruitment for clinical trials. Predicting the suitability of treatments is crucial for improving research and clinical practice. One study used an ML model to predict methotrexate (MTX) response in RA patients using clinical data. A Least Absolute Shrinkage and Selection Operator algorithm, a method for fitting linear models, was employed in this project, achieving better performance than RF, with an AUC-ROC=0.79 (vs 0.68 in RF); this effective categorisation of patients into good and poor responders was achieved with baseline Disease Activity Score 28 (DAS-28), anti-citrullinated protein antibody and Health Assessment Questionnaire as top predictors. Combining clinical data with genomic biomarkers (single-nucleotide polymorphisms) and baseline DAS-28 has also shown promise in predicting MTX response in early RA; metrics of different supervised ML methods showed an AUC-ROC=0.84 in the training cohort, and a validation cohort accuracy of 0.76. Similarly, different ML models (linear regression, random forest, XGBoost and CatBoost) were evaluated for their ability to predict the probability of therapeutic response for bDMARDs in RA in the ESPOIR cohort, predicting response to tumorous necrosis factor inhibitors with an AUC-ROC of 0.72 (0.68 to 0.73), and yielding key predictors such as DAS28, lymphocytes, aspartate aminotransferase, neutrophils, age, weight and smoking status. Models aiming to predict the readmission risk of patients with RMDs after discharge, or the evolution of a given disease, have also been developed. By analysing data from electronic health records (EHRs), RF-based models aiming to predict patient’s return to the clinic achieved an AUC-ROC of 0.65, a sensitivity of 0.38 and a specificity of 0.79; follow-up duration, the prescription of DMARDs, corticosteroids, diagnosis of chronic polyarthritis, quality of life and patient occupation were identified as key variables. In the context of life-threatening illnesses such as systemic sclerosis, predictive modelling has been employed to estimate mortality rates drawing on clinical, demographic and spirometric data. The Naïve Bayes Classifier, a supervised ML algorithm, achieved an AUC-ROC=0.76 to predict 5-year mortality rates after internal cross-validation, which demonstrated superior predictive capability as compared with other algorithms, including RF (AUC-ROC=0.73), logistic regression (AUC-ROC=0.75) and Cox regression (AUC-ROC=0.724). Concerning imaging , studies have used various techniques from simple to complex. Using X-rays, ML models achieved up to 90.7% accuracy in distinguishing RA and OA from normal hand radiographs, though accuracy decreased (80.6%) when classifying all three classes together. Moving on to osteoarthritis (OA), a DL model was trained on knee radiographs to identify patients with and without pain progression, as measured by the Western Ontario and McMaster Universities Arthritis Index (WOMAC) pain score. The DL model achieved an AUC-ROC=0.80 in predicting pain progression, significantly higher (p<0.001) than a traditional model trained on demographic, clinical and radiographic risk factors. In axial imaging, a neural network based on 1553 pelvis X-rays evaluating the presence or absence of definite radiographic sacroiliitis as agreed in a central reading session, identified definite sacroiliitis with an AUC-ROC=0.94, a sensitivity of 0.92 and a specificity of 0.81 for the test dataset. CT has also benefited from AI, where neural networks trained on CT-derived 3D joint shapes distinguished hand joint patterns in RA with AUC-ROC=75%, psoriatic arthritis (PsA) with AUC-ROC=68% and healthy controls with AUC-ROC=82%. These models additionally identified disease-specific regions prone to erosions and bony spurs, contributing to classifying undifferentiated arthritis. Convolutional neural networks (CNNs) trained on sacroiliac joint images detected structural lesions such as erosion and ankylosis, achieving sensitivities of 0.95 and 0.82 and specificities of 0.85 and 0.97. Additionally, in Sjögren’s syndrome, a DL model using 500 CT images detected salivary gland damage in parotid glands with 96% accuracy, comparable to diagnosis of experienced radiologists. Regarding MRI, CNNs have also demonstrated the ability to differentiate between patients with RA and PsA based on patterns from hand MRIs, achieving AUC-ROC=0.75 for seropositive RA versus PsA, 0.74 for seronegative RA versus PsA and 0.67 for seropositive versus seronegative RA. Interestingly, adding demographic or clinical data to the networks did not provide improve classification. Non-articular MRI applications, such as brain MRI, have been used to evaluate fatigue in RA, showing that brain structural metrics were superior to clinical measures, with the highest prediction accuracy reaching 0.67. In SpA, MRI models have been developed to detect sacroiliac joint active damage. In fact, a deep neural network developed to detect MRI changes in sacroiliac joints indicative of axial SpA (axSpA) achieved a sensitivity of 0.88 and specificity of 0.71 for detecting inflammatory changes, and a sensitivity of 0.85 and specificity of 0.78 for structural changes in external validation. A multi-purpose MRI-based model using compound image transformations analysed knee cartilage in T2-weighted images to predict progression to symptomatic OA with an accuracy of 0.75, as defined by the WOMAC score 3 years post-baseline. Other imaging modalities, such as ultrasound, have demonstrated potential in predicting RA relapses and assessing joint conditions. A study comparing three ML classifiers found XGBoost to be the best-performing model (AUC-ROC=0.75), identifying 10 key features, including superb microvascular imaging scores of wrist and metatarsophalangeal joints. On vasculitis ultrasound, a study assessed the use of a CNN for detecting the halo sign in colour Doppler images for diagnosing giant cell arteritis, achieving an AUC-ROC=0.84 on the test set, with a 0.95 specificity and 0.60 sensitivity. For Sjögren’s syndrome, DL models used transfer learning to improve the automated segmentation of salivary gland ultrasonography, achieving a higher Intersection-over-Union (0.85) compared with both inter-observer agreement (0.76) and intra-observer agreement (0.84), indicating superior accuracy and consistency. Thermography, combined with AI, can detect RA activity by analysing temperature changes in hand joints. An ML-based method, ThermoJIS, for detecting joint inflammation in RA using hand thermography, correlated moderately with ultrasound scores and demonstrated with good diagnostic performance (AUC-ROC=0.78). Building on this, the study developed and validated two composite disease activity indices, ThermoDAI and ThermoDAI-CRP, which showed stronger correlations with ultrasound-determined synovitis (GS=0.52–0.58; PD=0.56–0.61) compared with patient global assessment (PGA) and PGA+CRP, and strong correlations with clinical indices (ρ>0.81). In the context of text analysis , discriminative AI using NLP has aided the analysis of vast amounts of EHRs, including tasks such as disease identification and clinical characteristics assessments. Several studies highlight NLP’s utility in rheumatology for disease detection. For instance, a validated ML pipeline identified RA patients with high performance, with support vector machines (AUC-ROC=0.98, F1 score 0.83) and gradient boosting (AUC-ROC=0.94, F1 score 0.82) outperforming simpler word-matching methods. Other study demonstrated the capability to identify axSpA through an unsupervised algorithm, incorporating both the NLP concept and ICD codes, with a sensitivity of 0.78, a specificity of 0.94 and an AUC-ROC of 0.93. This has also been explored in PsA, in which a sensitivity of 0.79 and a PPV of 0.93 were achieved when NLP was combined with billing codes. Further, a tool combining text mining with NLP-based exclusion accurately identified ANCA-associated vasculitis cases, achieving a PPV of 0.86 and outperforming traditional ICD-10 coding. This growing body of evidence supports the adoption of NLP technologies in accurately identifying RMDs. Additional studies have focused on extracting clinical information beyond diagnoses from EHRs. For example, a recent study that included a dataset with around 64 million EHRs focused on the demographic and clinical characteristics of RA patients with interstitial lung disease (RA-ILD), yielding relevant information on prevalence, comorbidities and drug use in real life, with a high precision (F1 score over 0.7) for most of the assessed variables. Another algorithm extracted forced vital capacity from EHRs, strongly correlating (r=0.94) with pulmonary function test values. In RA, a study identified MTX-induced liver toxicity using NLP with a string-matching algorithm, achieving a PPV of 0.76. In another study, the analysis of structured and free-text EHR data from three hospitals showed limited disease activity evaluations in axSpA and PsA patients. For systemic autoimmune rheumatic diseases, an ML model predicted autoantibody testing needs and specialist referrals in systemic autoimmune diseases with AUC-ROC values from 0.91 to 0.94, enabling early detection up to 5 years before diagnosis. Another example illustrating the potential of AI in using large-scale real-world data is EPIC Cosmos, a vast inter-hospital database aggregating de-identified EHR from millions of patients across multiple health systems. EPIC Cosmos has enabled studies in different fields including rheumatology, such as recent work on SLE where researchers used Cosmos to enhance disease phenotyping and diagnosis. This study applied ICD codes to identify SLE patients and validated data quality against EULAR/ACR classification criteria, highlighting the need for integrating clinical notes to improve data completeness beyond structured EHR fields. While the study primarily relied on structured ICD codes for SLE phenotyping, the authors acknowledge plans to develop an NLP pipeline to analyse clinical notes, aiming to improve data completeness. Generative AI, the latest advancement in AI, is an emerging technology capable of creating new content in audio, image, video and text formats. It is based on foundation models—large-scale AI systems that acquire emergent capabilities across domains such as language, vision, robotics, reasoning and interaction. Their versatility allows them to adapt to diverse tasks, from NLP to computer vision and robotic control, by leveraging unlabelled data and self-supervised learning techniques. Among these, text-oriented applications have shown the most potentialities for research. At the core of generative AI are LLMs, which use transformer architecture to generate human-like responses based on input data. LLMs process and analyse input to generate outputs that mimic human reasoning based on statistical correlations, a capability that distinguishes them from discriminative AI, whose models produce a label or category based on the input, requiring explicit interpretation of the results. An example of interacting with these models is through widely recognised chatbots such as ChatGPT by OpenAI or Gemini by Google. Clinical workflow and decision-making Generative AI has yielded some results in research on clinical practice use, though its applications are still in the early stages. Current LLMs fine-tuned on medical data such as Med-PalM or Meditron show promise, nearing expert human performance in answering medical questions, which could serve as a decision support; nonetheless, they may fall short when addressing individual patient circumstances. Moreover, LLMs can significantly reduce administrative burdens by summarising and rephrasing information, aiding in the composition of clinical notes and discharge reports with real-time suggestions. Future developments will likely see major software companies integrating LLMs into administrative workflows, serving as clinical decision support systems and automating tasks such as documenting information from consultations, video calls and emails. Concerning disease diagnosis, LLMs have demonstrated significant results. One study conducted in early 2023 with ChatGPT-3.5 highlighted its strong performance across various clinical tasks, achieving an overall accuracy of 76.9% in making final diagnoses. The multimodal ChatGPT-4 has shown diagnostic capabilities in musculoskeletal radiology, performing at a level comparable to radiology residents when inputting the medical history and imaging findings (accuracy rates of 43% vs 41%) but not matching board-certified radiologists (53%). Interestingly, its text-based diagnostic performance surpassed that of its vision-based counterpart ( Vision ChatGPT4 version) when processing radiology findings rather than images. LLM performance has also been assessed in comparison to physicians for differentiating inflammatory rheumatic diseases from non-inflammatory conditions, highlighting its capacity to generate diagnostic insights through pattern recognition in language. ChatGPT-4 correctly identified the most likely diagnosis in 35% of cases, closely matching rheumatologists’ 39% (p=0.30). In cases of inflammatory rheumatic disease, ChatGPT-4 performed better, with 71% accuracy versus 62% for rheumatologists. However, it was less accurate in non-inflammatory rheumatic cases. Other LLM-based applications, such as DxGPT, have shown relatively high accuracy in diagnosing rare diseases. This decision support tool revealed that models like Claude 3 Opus achieved 55% strict accuracy and 70% top-5 accuracy using real-world datasets of rare diseases. While these findings highlight the capacity of AI to assist rheumatologists in diagnosing non-prevalent conditions, further validation in clinical settings is essential. Indeed, as a decision support tool, ChatGPT has proven useful and reliable for answering questions about some RMDs. In a study evaluating LLMs on MTX information for RA, GPT-4 achieved 100% accuracy and completeness, with all 23 MTX-related responses correct and complete as evaluated by two reviewers. In contrast, BARD (now Gemini) scored 73.9% correct answers. In the ChatSLE study, ChatGPT-4 was evaluated against leading rheumatology experts, providing answers to 100 patient-related questions from Lupus100.org. ChatGPT-4’s responses were rated as high quality, with a mean quality score of 4.55 (95% CI 4.48 to 4.62) compared with 4.31 (95% CI 4.23 to 4.39) for expert responses (p<0.0001). Both sources showed similar empathy scores, but ChatGPT-4 was preferred in 57% of cases (p=0.01). Additionally, ChatGPT-4 provided relatively accurate patient information, with a mean score of 8.4±0.7 on a 0–10 scale. Further studies have evaluated ChatGPT’s reliability and utility in providing information on common RMDs. For instance, an assessment of ChatGPT’s responses regarding conditions such as RA, AS and OA on a 7-point Likert scale, found that ChatGPT achieved the highest reliability score for OA (mean±SD 5.62±1.17), indicating that while the model is a promising tool, clinicians should remain vigilant of its probability to provide misleading information. Some studies have compared the performance of models in rheumatology. The recent Rheum2Guide study compared treatment plans generated by GPT-4 and GPT-3.5 with those created by a clinical rheumatology board using 20 fictional patient vignettes. GPT-4’s plans were selected more frequently than GPT-3.5’s for first-line treatments, indicating GPT-4’s closer alignment with clinical expectations. Although GPT-4 and GPT-3.5 generated safe and high-quality treatment plans, the rheumatology board’s plans were preferred in 68.8% of cases due to higher ratings in guideline adherence, medical appropriateness, completeness and overall quality. Another study evaluated the diagnostic capabilities of ChatGPT-4 and other LLMs like Claude 1.3, Claude 2 and Bard using standardised prompts in The Lancet’s Picture Quiz Gallery focused on rheumatic diseases—including the text and not images as part of the input. ChatGPT-4 and Claude 2 both achieved 81% accuracy, outperforming Claude 1.3 (72%) and Bard (66%). However, all models, except Claude 2, struggled with cases involving uncommon infectious diseases, where ChatGPT-4’s accuracy dropped to 57%. The accuracy and reasoning skills of LLMs have also been demonstrated in challenging clinical examinations, in which could be used to aid medical education. For instance, ChatGPT-4 has repeatedly demonstrated proficiency in standardised tests like the US Medical Licensing Examination, where it provided coherent and intuitive responses, surpassing the performance of previous earlier AI systems. Additionally, it successfully answered 93.7% of all rheumatology-related questions from the Spanish Medical Training Examination (MIR) within the years 2009–2023, with a median clinical reasoning score of 4.67 on a 5-point Likert scale, outperforming earlier LLM versions. Although currently the use of AI in day-to-day clinical practice may represent a complement rather than a stand-alone solution, it has been shown that the clinician with AI support versus traditional methods does not improve the situation, but AI alone has shown better results than previous groups, so the potential of this technology will be derived based on the clinician’s learning to use it. Drug development, clinical trials and digital twins Drug development has leveraged generative AI for molecule generation and molecular property prediction. For instance, BERT—a transformer-based model—has been adapted to learn molecular representations, supporting drug discovery tasks. Similarly, other language models have been fine-tuned for molecule generation and annotation, significantly enhancing efficiency and accuracy in drug design. In clinical trials, generative models can create synthetic data that closely mirrors real-world data, as illustrated with digital twins (DTs). Pretrained on patient vitals, clinical trajectories, lab results and diagnoses, DTs simulate patient evolution over time based on treatment decisions. Indeed, DT may facilitate the creation of synthetic control arms, which can replicate patient groups for comparative analyses without recruiting additional participants. External controlled arms based clinical trials have been supported by both the FDA and EMA; in rheumatology, these approaches have been applied to research in RA. Optimising the research process AI, particularly through LLMs, has the opportunity to transform research by offering advanced tools that can support every stage of the process. Central to adopting these capabilities is the concept of prompting —the process of giving instructions to AI systems. Prompts can range from simple, direct queries to complex, structured inputs designed to elicit detailed responses. In research, prompting can be executed as ‘zero-shot’ learning, where the AI is given a task without any prior example or training; more refined prompts can guide AI to produce more focused and relevant information based on examples, such as few-shot prompting (giving examples) or chain-of-thought (providing step-by-step) prompting. The role of AI in all aspects of research goes from idea generation and literature review to data analysis and manuscript preparation . In the early stages of research, AI can significantly help via brainstorming sessions, in which a wide range of ideas and hypotheses can be explored. LLMs such as ChatGPT-4, Gemini, Perplexity and Claude are capable of generating diverse perspectives on a given research question, helping to promote the creativity that helps refine research objectives. These tools allow researchers to quickly iterate their ideas, explore conceivable investigative angles and develop a clear roadmap for studies. A recent study found that an AI model generated research ideas rated as more original and exciting than those of human scientists, though with slightly lower feasibility. Using the Claude 3.5 model, researchers produced 4000 ideas across several topics, and reviewers assessed these ideas without knowing their source. Despite AI’s high novelty scores, only about 200 ideas were genuinely unique, with creativity diminishing over time. The overwhelming abundance of options produced by AI challenges conventional creative processes, pushing researchers to shift from seeking single insights to generating numerous ideas for refinement. Rather than asking for one idea, AI enables researchers to request many altogether, allowing to sift through diverse suggestions and strategies. This surplus demands the skill of curating and discerning the best quality, which highlights a core value of AI in augmenting intellectual creativity and decision-making in research. As the research project progresses, AI tools can assist in refining the design and structure of the study. Beyond generating ideas, these systems can suggest detailed article structures, and help in drafting sections of a manuscript. Conducting a literature review can also benefit from AI support. Tools such as Elicit or Research Rabbit provide curated bibliographies by searching with NLP. In addition, they can offer insights into the state-of-the-art of research concepts and visualise the relationships between key studies, potentially uncovering connections between research articles that may not be immediately apparent. They can also generate summaries from articles, extract precise data and even highlight emerging trends in the field, enabling researchers to stay ahead in their field. AI tools can also take on a more active role during the data analysis phase. LLMs can assist in data analysis by guiding researchers through their analysis, performing statistical tests and generating insights from datasets. For instance, ChatGPT-4, can provide AI-generated code for statistical tools like SPSS, R or Python that facilitates the execution of complex analyses. Moreover, it can directly perform advanced computational calculations and data queries by inputting the prompt and the dataset to the system. As an additional support, it can also provide preliminary interpretations of data and give insights on the results. A recent study of 187 489 software developers using GitHub Copilot demonstrated how AI tools can reshape work by shifting focus from non-core management tasks to primary tasks, such as coding. This shift allowed developers to work more autonomously, explore new methods, and potentially reduce hierarchical dependencies. Finally, the writing phase—often the most time-consuming—can be facilitated using AI. Besides ChatGPT or Gemini, other tools such as Jenni AI can suggest article structures, and help draft coherent and well-organised manuscripts. These platforms can assist with translation, paraphrasing and ensuring that the text adheres to publication standards. As for the presentation of the results, image models can support on creating graphs, images or presentations. For example, platforms like Microsoft Copilot can create a presentation of the research. Challenges ahead The integration of AI into the field of rheumatology research is a double-edged sword, offering significant chances alongside profound ethical and practical challenges. AI models have demonstrated high accuracy in diagnosing and predicting outcomes of rheumatic diseases, sometimes even surpassing traditional methods. Predictive analytics can identify patients at higher risk of disease progression, facilitating proactive management. Nonetheless, accuracy and reliability of these models in clinical practice is yet to be explored. While there are some studies including external validation of the algorithms, clinical trials assessing the efficacy of these models in randomised controlled trials are lacking in rheumatology. Another primary concern is the rapid pace at which AI is being adopted, particularly as health systems deploy AI-driven support tools with minimal clinician training. Without structured guidance, clinicians may struggle to use these tools effectively, which could limit their impact on diagnostic accuracy. A recent randomised trial demonstrated that access to an LLM alone did not improve physicians’ diagnostic reasoning in challenging cases, even though the LLM performed well when operating independently. Unexpectedly, the LLM alone significantly outperformed physicians in diagnostic reasoning for complex cases. This finding suggests that simply having access to AI tools does not inherently enhance clinical reasoning skills and that effective use of these tools requires comprehensive training. Besides, flawed training data may possibly lead to algorithmic bias; AI models trained on non-representative data might produce skewed outcomes, disadvantaging certain patient groups. As an example in SLE, AI models trained predominantly on data from non-Hispanic white populations may produce less accurate predictions for under-represented groups, such as black, Hispanic or Asian patients, due to differing symptom patterns and disease progression, potentially leading to skewed outcomes in diagnosis and treatment. Concerning generative AI, some researchers have raised concerns about the readiness of LLMs for medical application. For example, there have been instances where the unethical use of LLMs, such as generating fraudulent research or using undisclosed AI assistance in manuscript writing, has led to the retraction of scientific papers. Additionally, LLMs are prone to ‘hallucinations’, where they generate plausible-sounding but incorrect information, which is a matter of debate for clinical practice, where accuracy and evidence-based knowledge are paramount. In addition, the black-box nature of many AI algorithms also raises transparency issues, making it difficult for practitioners to understand and trust AI-generated insights. To address these concerns, new reporting guidelines have emerged for both discriminative and generative models, such as TRIPOD-AI for validating AI interventions, CONSORT-AI for clinical trials, DECIDE-AI for decision support systems and CLAIM-AI for imaging technologies. Additionally, guidelines like CANGARU have been developed specifically for generative AI models, reflecting the growing need for transparency and accountability in AI-driven research. In addition, regulatory frameworks, such as the European Union’s AI Act, set to be enforced in 2026, are now being formulated. As we stand at the crossroads of innovation, regulation and ethics, the responsible evolution of AI in rheumatology requires a collective commitment from researchers to thoughtfully use these technologies, ensuring they enhance both research and patient care. With an ageing population and a projected increase in RMDs, AI can assist in managing the growing demand on healthcare systems. In this regard, AI may enhance collaboration between general practitioners and rheumatologists. Decision-support systems can aid in the early detection of RMDs at the primary care level, improving the accuracy and timeliness of referrals. Enhanced communication platforms can lead to a more integrated approach to patient care. Moving forward, it is crucial to balance the promising capabilities of AI with a mindful consideration of its limitations. Training healthcare professionals in AI technologies will facilitate their effective integration into clinical practice. Ongoing research is necessary to enhance the robustness of AI models and adapt them to the evolving needs of rheumatology. Generative AI has yielded some results in research on clinical practice use, though its applications are still in the early stages. Current LLMs fine-tuned on medical data such as Med-PalM or Meditron show promise, nearing expert human performance in answering medical questions, which could serve as a decision support; nonetheless, they may fall short when addressing individual patient circumstances. Moreover, LLMs can significantly reduce administrative burdens by summarising and rephrasing information, aiding in the composition of clinical notes and discharge reports with real-time suggestions. Future developments will likely see major software companies integrating LLMs into administrative workflows, serving as clinical decision support systems and automating tasks such as documenting information from consultations, video calls and emails. Concerning disease diagnosis, LLMs have demonstrated significant results. One study conducted in early 2023 with ChatGPT-3.5 highlighted its strong performance across various clinical tasks, achieving an overall accuracy of 76.9% in making final diagnoses. The multimodal ChatGPT-4 has shown diagnostic capabilities in musculoskeletal radiology, performing at a level comparable to radiology residents when inputting the medical history and imaging findings (accuracy rates of 43% vs 41%) but not matching board-certified radiologists (53%). Interestingly, its text-based diagnostic performance surpassed that of its vision-based counterpart ( Vision ChatGPT4 version) when processing radiology findings rather than images. LLM performance has also been assessed in comparison to physicians for differentiating inflammatory rheumatic diseases from non-inflammatory conditions, highlighting its capacity to generate diagnostic insights through pattern recognition in language. ChatGPT-4 correctly identified the most likely diagnosis in 35% of cases, closely matching rheumatologists’ 39% (p=0.30). In cases of inflammatory rheumatic disease, ChatGPT-4 performed better, with 71% accuracy versus 62% for rheumatologists. However, it was less accurate in non-inflammatory rheumatic cases. Other LLM-based applications, such as DxGPT, have shown relatively high accuracy in diagnosing rare diseases. This decision support tool revealed that models like Claude 3 Opus achieved 55% strict accuracy and 70% top-5 accuracy using real-world datasets of rare diseases. While these findings highlight the capacity of AI to assist rheumatologists in diagnosing non-prevalent conditions, further validation in clinical settings is essential. Indeed, as a decision support tool, ChatGPT has proven useful and reliable for answering questions about some RMDs. In a study evaluating LLMs on MTX information for RA, GPT-4 achieved 100% accuracy and completeness, with all 23 MTX-related responses correct and complete as evaluated by two reviewers. In contrast, BARD (now Gemini) scored 73.9% correct answers. In the ChatSLE study, ChatGPT-4 was evaluated against leading rheumatology experts, providing answers to 100 patient-related questions from Lupus100.org. ChatGPT-4’s responses were rated as high quality, with a mean quality score of 4.55 (95% CI 4.48 to 4.62) compared with 4.31 (95% CI 4.23 to 4.39) for expert responses (p<0.0001). Both sources showed similar empathy scores, but ChatGPT-4 was preferred in 57% of cases (p=0.01). Additionally, ChatGPT-4 provided relatively accurate patient information, with a mean score of 8.4±0.7 on a 0–10 scale. Further studies have evaluated ChatGPT’s reliability and utility in providing information on common RMDs. For instance, an assessment of ChatGPT’s responses regarding conditions such as RA, AS and OA on a 7-point Likert scale, found that ChatGPT achieved the highest reliability score for OA (mean±SD 5.62±1.17), indicating that while the model is a promising tool, clinicians should remain vigilant of its probability to provide misleading information. Some studies have compared the performance of models in rheumatology. The recent Rheum2Guide study compared treatment plans generated by GPT-4 and GPT-3.5 with those created by a clinical rheumatology board using 20 fictional patient vignettes. GPT-4’s plans were selected more frequently than GPT-3.5’s for first-line treatments, indicating GPT-4’s closer alignment with clinical expectations. Although GPT-4 and GPT-3.5 generated safe and high-quality treatment plans, the rheumatology board’s plans were preferred in 68.8% of cases due to higher ratings in guideline adherence, medical appropriateness, completeness and overall quality. Another study evaluated the diagnostic capabilities of ChatGPT-4 and other LLMs like Claude 1.3, Claude 2 and Bard using standardised prompts in The Lancet’s Picture Quiz Gallery focused on rheumatic diseases—including the text and not images as part of the input. ChatGPT-4 and Claude 2 both achieved 81% accuracy, outperforming Claude 1.3 (72%) and Bard (66%). However, all models, except Claude 2, struggled with cases involving uncommon infectious diseases, where ChatGPT-4’s accuracy dropped to 57%. The accuracy and reasoning skills of LLMs have also been demonstrated in challenging clinical examinations, in which could be used to aid medical education. For instance, ChatGPT-4 has repeatedly demonstrated proficiency in standardised tests like the US Medical Licensing Examination, where it provided coherent and intuitive responses, surpassing the performance of previous earlier AI systems. Additionally, it successfully answered 93.7% of all rheumatology-related questions from the Spanish Medical Training Examination (MIR) within the years 2009–2023, with a median clinical reasoning score of 4.67 on a 5-point Likert scale, outperforming earlier LLM versions. Although currently the use of AI in day-to-day clinical practice may represent a complement rather than a stand-alone solution, it has been shown that the clinician with AI support versus traditional methods does not improve the situation, but AI alone has shown better results than previous groups, so the potential of this technology will be derived based on the clinician’s learning to use it. Drug development has leveraged generative AI for molecule generation and molecular property prediction. For instance, BERT—a transformer-based model—has been adapted to learn molecular representations, supporting drug discovery tasks. Similarly, other language models have been fine-tuned for molecule generation and annotation, significantly enhancing efficiency and accuracy in drug design. In clinical trials, generative models can create synthetic data that closely mirrors real-world data, as illustrated with digital twins (DTs). Pretrained on patient vitals, clinical trajectories, lab results and diagnoses, DTs simulate patient evolution over time based on treatment decisions. Indeed, DT may facilitate the creation of synthetic control arms, which can replicate patient groups for comparative analyses without recruiting additional participants. External controlled arms based clinical trials have been supported by both the FDA and EMA; in rheumatology, these approaches have been applied to research in RA. AI, particularly through LLMs, has the opportunity to transform research by offering advanced tools that can support every stage of the process. Central to adopting these capabilities is the concept of prompting —the process of giving instructions to AI systems. Prompts can range from simple, direct queries to complex, structured inputs designed to elicit detailed responses. In research, prompting can be executed as ‘zero-shot’ learning, where the AI is given a task without any prior example or training; more refined prompts can guide AI to produce more focused and relevant information based on examples, such as few-shot prompting (giving examples) or chain-of-thought (providing step-by-step) prompting. The role of AI in all aspects of research goes from idea generation and literature review to data analysis and manuscript preparation . In the early stages of research, AI can significantly help via brainstorming sessions, in which a wide range of ideas and hypotheses can be explored. LLMs such as ChatGPT-4, Gemini, Perplexity and Claude are capable of generating diverse perspectives on a given research question, helping to promote the creativity that helps refine research objectives. These tools allow researchers to quickly iterate their ideas, explore conceivable investigative angles and develop a clear roadmap for studies. A recent study found that an AI model generated research ideas rated as more original and exciting than those of human scientists, though with slightly lower feasibility. Using the Claude 3.5 model, researchers produced 4000 ideas across several topics, and reviewers assessed these ideas without knowing their source. Despite AI’s high novelty scores, only about 200 ideas were genuinely unique, with creativity diminishing over time. The overwhelming abundance of options produced by AI challenges conventional creative processes, pushing researchers to shift from seeking single insights to generating numerous ideas for refinement. Rather than asking for one idea, AI enables researchers to request many altogether, allowing to sift through diverse suggestions and strategies. This surplus demands the skill of curating and discerning the best quality, which highlights a core value of AI in augmenting intellectual creativity and decision-making in research. As the research project progresses, AI tools can assist in refining the design and structure of the study. Beyond generating ideas, these systems can suggest detailed article structures, and help in drafting sections of a manuscript. Conducting a literature review can also benefit from AI support. Tools such as Elicit or Research Rabbit provide curated bibliographies by searching with NLP. In addition, they can offer insights into the state-of-the-art of research concepts and visualise the relationships between key studies, potentially uncovering connections between research articles that may not be immediately apparent. They can also generate summaries from articles, extract precise data and even highlight emerging trends in the field, enabling researchers to stay ahead in their field. AI tools can also take on a more active role during the data analysis phase. LLMs can assist in data analysis by guiding researchers through their analysis, performing statistical tests and generating insights from datasets. For instance, ChatGPT-4, can provide AI-generated code for statistical tools like SPSS, R or Python that facilitates the execution of complex analyses. Moreover, it can directly perform advanced computational calculations and data queries by inputting the prompt and the dataset to the system. As an additional support, it can also provide preliminary interpretations of data and give insights on the results. A recent study of 187 489 software developers using GitHub Copilot demonstrated how AI tools can reshape work by shifting focus from non-core management tasks to primary tasks, such as coding. This shift allowed developers to work more autonomously, explore new methods, and potentially reduce hierarchical dependencies. Finally, the writing phase—often the most time-consuming—can be facilitated using AI. Besides ChatGPT or Gemini, other tools such as Jenni AI can suggest article structures, and help draft coherent and well-organised manuscripts. These platforms can assist with translation, paraphrasing and ensuring that the text adheres to publication standards. As for the presentation of the results, image models can support on creating graphs, images or presentations. For example, platforms like Microsoft Copilot can create a presentation of the research. The integration of AI into the field of rheumatology research is a double-edged sword, offering significant chances alongside profound ethical and practical challenges. AI models have demonstrated high accuracy in diagnosing and predicting outcomes of rheumatic diseases, sometimes even surpassing traditional methods. Predictive analytics can identify patients at higher risk of disease progression, facilitating proactive management. Nonetheless, accuracy and reliability of these models in clinical practice is yet to be explored. While there are some studies including external validation of the algorithms, clinical trials assessing the efficacy of these models in randomised controlled trials are lacking in rheumatology. Another primary concern is the rapid pace at which AI is being adopted, particularly as health systems deploy AI-driven support tools with minimal clinician training. Without structured guidance, clinicians may struggle to use these tools effectively, which could limit their impact on diagnostic accuracy. A recent randomised trial demonstrated that access to an LLM alone did not improve physicians’ diagnostic reasoning in challenging cases, even though the LLM performed well when operating independently. Unexpectedly, the LLM alone significantly outperformed physicians in diagnostic reasoning for complex cases. This finding suggests that simply having access to AI tools does not inherently enhance clinical reasoning skills and that effective use of these tools requires comprehensive training. Besides, flawed training data may possibly lead to algorithmic bias; AI models trained on non-representative data might produce skewed outcomes, disadvantaging certain patient groups. As an example in SLE, AI models trained predominantly on data from non-Hispanic white populations may produce less accurate predictions for under-represented groups, such as black, Hispanic or Asian patients, due to differing symptom patterns and disease progression, potentially leading to skewed outcomes in diagnosis and treatment. Concerning generative AI, some researchers have raised concerns about the readiness of LLMs for medical application. For example, there have been instances where the unethical use of LLMs, such as generating fraudulent research or using undisclosed AI assistance in manuscript writing, has led to the retraction of scientific papers. Additionally, LLMs are prone to ‘hallucinations’, where they generate plausible-sounding but incorrect information, which is a matter of debate for clinical practice, where accuracy and evidence-based knowledge are paramount. In addition, the black-box nature of many AI algorithms also raises transparency issues, making it difficult for practitioners to understand and trust AI-generated insights. To address these concerns, new reporting guidelines have emerged for both discriminative and generative models, such as TRIPOD-AI for validating AI interventions, CONSORT-AI for clinical trials, DECIDE-AI for decision support systems and CLAIM-AI for imaging technologies. Additionally, guidelines like CANGARU have been developed specifically for generative AI models, reflecting the growing need for transparency and accountability in AI-driven research. In addition, regulatory frameworks, such as the European Union’s AI Act, set to be enforced in 2026, are now being formulated. As we stand at the crossroads of innovation, regulation and ethics, the responsible evolution of AI in rheumatology requires a collective commitment from researchers to thoughtfully use these technologies, ensuring they enhance both research and patient care. With an ageing population and a projected increase in RMDs, AI can assist in managing the growing demand on healthcare systems. In this regard, AI may enhance collaboration between general practitioners and rheumatologists. Decision-support systems can aid in the early detection of RMDs at the primary care level, improving the accuracy and timeliness of referrals. Enhanced communication platforms can lead to a more integrated approach to patient care. Moving forward, it is crucial to balance the promising capabilities of AI with a mindful consideration of its limitations. Training healthcare professionals in AI technologies will facilitate their effective integration into clinical practice. Ongoing research is necessary to enhance the robustness of AI models and adapt them to the evolving needs of rheumatology. The combined strengths of discriminative and generative AI are revolutionising rheumatology research. Discriminative AI’s precise classification and prediction capabilities, paired with generative AI’s ability to synthesise and create content, may provide rheumatology researchers with powerful tools to enhance their work. These advancements can accelerate the research process and therefore contribute to the development of efficient processes in rheumatology. However, as we integrate these technologies into our research, we must proceed with caution, balancing innovation with responsibility to maximise their prospective impact on the field. The future trajectory of AI in rheumatology is within our hands, with its ultimate impact determined by our collective efforts and thoughtful application. |
Detection of the 30-bp deletion and protein expression of Epstein-Barr virus latent membrane protein 1 in extranodal NK/T cell lymphoma and its clinicopathological significance | 5ea502be-d023-44a0-ae91-3d58b1ee5fa2 | 11823043 | Anatomy[mh] | Extranodal natural killer/T-cell lymphoma (ENKTCL) is a rare subtype of EBV-associated non-Hodgkin lymphoma with a highly aggressive clinical course . It has a predilection for extranodal involvement, especially in the upper aerodigestive tract (UADT), but also affects the skin, gastrointestinal tract, soft tissue, and testis . Histologically, ENKTCL is characterized by angioinvasion and angiodestruction, accompanied by prominent coagulative necrosis and a cytotoxic phenotype . ENKTCL predominantly occurs in East Asia and Latin America rather than in Europe and North America . Although its pathogenesis is still unclear, the effect of the strong geographic distribution of ENKTCL on specific populations suggests a genetic predisposition for ENKTCL. Epstein-Barr virus (EBV), a member of the Herpesviridae family, infects approximately 95% of the world’s population and persists as an asymptomatic, life-long infection . The importance of EBV in ENKTCL lymphomagenesis was first recognized in 1990 and has been confirmed in various studies since then . Further studies have focused on the viral proteins and their genetic variants to investigate the role of EBV in this lymphoma’s pathogenesis . Latent membrane protein 1 (LMP1) is one of the important oncogenes of EBV carcinogenicity and has the ability to induce malignant transformation in epithelial and B cells . The LMP1 gene has been shown to have polymorphisms, among which the 30-base-pair (bp) deletion (del-LMP1) is the most common variant in the C-terminus and was first detected in nasopharyngeal carcinoma (NPC) patients from southern China . It occurs at the 3’ end of the C-terminal tail and is associated with the CTAR2 functional domain. However, there were conflicting views about the significant role of the 30-bp deletion in conferring the more tumorigenic potential of the LMP1 gene . Knecht et al. reported an association of del-LMP1 with cases of European Hodgkin’s disease (HD) and suggested that del-LMP1 was associated with aggressive histology or behavior; however, a study from Mexico held a negative attitude and disputed a definite pathogenetic role for del-LMP1 in ENKTCL . Most reports of del-LMP1 have suggested that this variant is widespread worldwide but varies in different geographic regions . A recent review by Montes-Mojarro et al. analyzed the distribution of LMP1 variants in 6 studies involving 140 ENKTCL patients showed a clear predominance of the wild-type LMP1 (wt-LMP1) (52.1%) with 37.1% of patients harboring del-LMP1 . While, other authors have found a high frequency of del-LMP1 detected in Asian ENKTCL patients . However, to the best of our knowledge, the clinicopathological significance of del-LMP1 in ENKTCL has not been fully elucidated to date. Previous studies have found that B-cell-derived LMP1 isolate protein is highly immunogenic . Kingma et al. suggested that mutations in the LMP1 gene reduced the immunogenicity of LMP1 and thus escaped immune surveillance in immunocompetent hosts. Therefore, simultaneous assessment of LMP1 expression at the protein level may contribute to our better understanding of the underlying significance of LMP1 in ENKTCL tumorigenesis and development. Herein, we conducted a retrospective cohort study to investigate whether 30-bp deletion of the LMP1 gene and LMP1 protein expression play any role in the clinicopathological manifestations and prognosis of ENKTCL in Wenzhou. Sampling and data collection We retrospectively reviewed records and samples of ENKTCL from 125 patients between January 2016 and June 2022 at the Department of Pathology, the First Affiliated Hospital of Wenzhou Medical University. The ENKTCL diagnosis was according to the World Health Organization’s (WHO) Classification of Tumours of Haematopoietic and Lymphoid Tissues . Patients with other malignancies, patients exposed to neoadjuvant chemotherapy and radiation therapy before surgery, and patients with incomplete data were excluded. We selected 42 cases with adequate tissue material for further investigation using nested polymerase chain reaction (PCR) and immunohistochemistry (IHC). Patients’ clinical information was collected from electronic medical records. In addition, 10 nasopharyngeal chronic inflammation tissues were selected as the control group. The presence of EBV was confirmed by in situ hybridization (ISH) for EBV-encoded small RNA (EBER) (Zhongshan Golden Bridge Company, Beijing, China). DNA extraction Fifty-two eligible samples (42 ENKTCL and 10 control tissues) were obtained from formalin-fixed paraffin-embedded (FFPE) tissue blocks by cutting 5-µm-thick sections. Each specimen used a fresh microtome blade to reduce the risk of cross-contamination. According to the manufacturer’s protocol, deoxyribonucleic acid (DNA) was extracted and purified from samples using the Biospin FFPE Tissue Genomic DNA Extraction Kit. Nested PCR for detection of the LMP1 gene Nested PCR was performed to amplify the C-terminus of the LMP1 gene. The first round of amplification was done using the outer primers of LMP1-F1 (5’-ATTGGCACAAGATGGAAAGC-3’) and LMP1-R1 (5’-TCCTTTGGCTCCTCCTGTTT-3’) in a total volume of 20 µl containing Taq PCR Master Mix 12.5 µl, 1 µl of DNA extracted from tissues, 1 µl of each primer, and 4.5 µl deionized sterile water (ddH2O). The cycling protocol was started with an initial denaturation for 5 min at 98℃, then 30 cycles of denaturation for 30 s at 95 ℃, annealing for 30 s at 54 ℃, and extension for 75 s at 72 ℃. A final elongation step for 10 min at 72 ℃ was then conducted. For the second round of PCR, we employed the inner primers of LMP1-F2 (5’-GAGGGAGAGTCAGTCAGGC-3’) and LMP1-R2 (5’-AGACGGAAGAGGTTGAAAAC-3’); then 1 µl of the PCR product was used as the template. The second round utilized similar experimental conditions except for the extension time, which was 30 s. In each experiment, positive and negative controls were conducted, respectively. The primers were designed according to the standard strain EBV 95.8 and synthesized by Sangon Biotech Co., Ltd. (Shanghai, China). Agarose gel electrophoresis The nested PCR products were separated by electrophoresis on a 1.2% agarose gel in 1× Tris-Acetate-EDTA (TAE) buffer for 45 min. The gel was stained with 4S green nucleic acid gel stain and photographed under ultraviolet light. The DNA molecular weight standard (100-5000) was used as the marker. DNA sequencing The specific band of interest was excised from the gel and purified using the Agarose Gel DNA Extraction Kit (Sangon Biotech Co., Ltd., Shanghai, China) following the manufacturer’s protocols. The PCR products were sequenced by Sangon Biotech Co., Ltd., a service provider, Sanger sequencing method was used to determine the presence of deletion mutations, and the sequencing primer was LMP1-R2. The results were analyzed using MegAlign software. IHC for the detection of LMP1 protein IHC was performed on sections of 3.5 μm thickness. After being deparaffinized in xylene and rehydrated using a graded alcohol series, the antigen retrieval was followed in an EDTA buffer at 95 ℃ for 20 min. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 10 min at room temperature. Slides were then incubated with mouse anti-LMP1 monoclonal antibody (clone CS1-4, MXB, Fujian, China) overnight at 4 ℃ and horseradish peroxidase (HRP)-labeled secondary antibody for 20 min at room temperature. Diaminobenzidine was used as the chromogen for the immunostaining. Finally, the sections were counterstained with hematoxylin. EBV-positive Hodgkin’s lymphoma sections were performed as positive controls according to the manufacturer’s instructions. Negative controls were obtained by substituting the primary antibody with phosphate-buffered saline. LMP1-positive was determined when the lymphoma cells were positive according to the methods described by Kume et al. . Statistical analysis Statistical analyses were performed with SPSS 26.0 software. The Chi-squared test and Fisher’s exact test were used to compare variables among groups (calculated categorical variables). Overall survival (OS) was calculated from the date of initial diagnosis until death from any cause or the date of last follow-up. Survival curves were calculated with the Kaplan-Meier method, and statistical significance was assessed using the log-rank test. All statistical tests were two-sided; a p -value ≤ 0.05 was considered statistically significant. We retrospectively reviewed records and samples of ENKTCL from 125 patients between January 2016 and June 2022 at the Department of Pathology, the First Affiliated Hospital of Wenzhou Medical University. The ENKTCL diagnosis was according to the World Health Organization’s (WHO) Classification of Tumours of Haematopoietic and Lymphoid Tissues . Patients with other malignancies, patients exposed to neoadjuvant chemotherapy and radiation therapy before surgery, and patients with incomplete data were excluded. We selected 42 cases with adequate tissue material for further investigation using nested polymerase chain reaction (PCR) and immunohistochemistry (IHC). Patients’ clinical information was collected from electronic medical records. In addition, 10 nasopharyngeal chronic inflammation tissues were selected as the control group. The presence of EBV was confirmed by in situ hybridization (ISH) for EBV-encoded small RNA (EBER) (Zhongshan Golden Bridge Company, Beijing, China). Fifty-two eligible samples (42 ENKTCL and 10 control tissues) were obtained from formalin-fixed paraffin-embedded (FFPE) tissue blocks by cutting 5-µm-thick sections. Each specimen used a fresh microtome blade to reduce the risk of cross-contamination. According to the manufacturer’s protocol, deoxyribonucleic acid (DNA) was extracted and purified from samples using the Biospin FFPE Tissue Genomic DNA Extraction Kit. Nested PCR was performed to amplify the C-terminus of the LMP1 gene. The first round of amplification was done using the outer primers of LMP1-F1 (5’-ATTGGCACAAGATGGAAAGC-3’) and LMP1-R1 (5’-TCCTTTGGCTCCTCCTGTTT-3’) in a total volume of 20 µl containing Taq PCR Master Mix 12.5 µl, 1 µl of DNA extracted from tissues, 1 µl of each primer, and 4.5 µl deionized sterile water (ddH2O). The cycling protocol was started with an initial denaturation for 5 min at 98℃, then 30 cycles of denaturation for 30 s at 95 ℃, annealing for 30 s at 54 ℃, and extension for 75 s at 72 ℃. A final elongation step for 10 min at 72 ℃ was then conducted. For the second round of PCR, we employed the inner primers of LMP1-F2 (5’-GAGGGAGAGTCAGTCAGGC-3’) and LMP1-R2 (5’-AGACGGAAGAGGTTGAAAAC-3’); then 1 µl of the PCR product was used as the template. The second round utilized similar experimental conditions except for the extension time, which was 30 s. In each experiment, positive and negative controls were conducted, respectively. The primers were designed according to the standard strain EBV 95.8 and synthesized by Sangon Biotech Co., Ltd. (Shanghai, China). The nested PCR products were separated by electrophoresis on a 1.2% agarose gel in 1× Tris-Acetate-EDTA (TAE) buffer for 45 min. The gel was stained with 4S green nucleic acid gel stain and photographed under ultraviolet light. The DNA molecular weight standard (100-5000) was used as the marker. The specific band of interest was excised from the gel and purified using the Agarose Gel DNA Extraction Kit (Sangon Biotech Co., Ltd., Shanghai, China) following the manufacturer’s protocols. The PCR products were sequenced by Sangon Biotech Co., Ltd., a service provider, Sanger sequencing method was used to determine the presence of deletion mutations, and the sequencing primer was LMP1-R2. The results were analyzed using MegAlign software. IHC was performed on sections of 3.5 μm thickness. After being deparaffinized in xylene and rehydrated using a graded alcohol series, the antigen retrieval was followed in an EDTA buffer at 95 ℃ for 20 min. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 10 min at room temperature. Slides were then incubated with mouse anti-LMP1 monoclonal antibody (clone CS1-4, MXB, Fujian, China) overnight at 4 ℃ and horseradish peroxidase (HRP)-labeled secondary antibody for 20 min at room temperature. Diaminobenzidine was used as the chromogen for the immunostaining. Finally, the sections were counterstained with hematoxylin. EBV-positive Hodgkin’s lymphoma sections were performed as positive controls according to the manufacturer’s instructions. Negative controls were obtained by substituting the primary antibody with phosphate-buffered saline. LMP1-positive was determined when the lymphoma cells were positive according to the methods described by Kume et al. . Statistical analyses were performed with SPSS 26.0 software. The Chi-squared test and Fisher’s exact test were used to compare variables among groups (calculated categorical variables). Overall survival (OS) was calculated from the date of initial diagnosis until death from any cause or the date of last follow-up. Survival curves were calculated with the Kaplan-Meier method, and statistical significance was assessed using the log-rank test. All statistical tests were two-sided; a p -value ≤ 0.05 was considered statistically significant. Patients characteristics The characteristics of 42 ENKTCL patients were summarized in Table . There were 31 men and 11 women with a male-to-female ratio of 2.8 to 1. The age range was from 21 to 87 years, with a median age of 55 years and 16 patients over 60 years old. The most common site of occurrence was the UADT ( n = 28); epistaxis, rhinorrhea, and nasal obstruction were initial manifestations. Thirteen patients had B symptoms, and eighteen had elevated serum lactate dehydrogenase (LDH) levels. Thirty-two patients were classified as stages I/II according to the Ann Arbor stage. Morphologically, most cases showed pleomorphic infiltrates, angiodestruction occurred in 22 samples, and variable degrees of necrosis were found in 23 tissues (Fig. A). CD56 (clone UMAB83, Zhongshan Golden Bridge Company, Beijing, China) and Granzyme B (clone EP230, Zhongshan Golden Bridge Company, Beijing, China) were positive in all cases. The median value of Ki-67 (clone UMAB107, Zhongshan Golden Bridge Company, Beijing, China) was 80% (range: 35%-98%). All cases showed EBER-positive signals in tumor cell nuclei (Fig. B). Twenty patients had chemotherapy alone, while twelve received combined chemoradiotherapy. Three patients received autologous stem cell transplants and three more received allogeneic stem cell transplantations. By the end of the follow-up, ten patients had died of disease. The median survival was 19.1 months (range: 0.1-48 months). The 1-year and 3-year estimated survival rates were 79.2% and 69.3%, respectively. Detection of the LMP1 gene and del-LMP1 The LMP1 gene was successfully amplified from 37/42 ENKTCL cases and 6/10 non-malignant controls, as shown in Table . Two types of PCR products, 251 bp and 221 bp, were obtained following amplification of the LMP1 gene. Representative gel electrophoresis results were shown in Fig. . Sanger sequencing confirmed that the 221 bp product contained a special 30-bp deletion (from nucleotide 168255 to 168285). The results indicated that wt-LMP1 was present in 2 ENKTCL patients, while del-LMP1 was present in 35 ENKTCL patients and 6 non-malignant controls, as depicted in the relevant sequence diagram in Fig. . No statistical difference was found for del-LMP1 between ENKTCL and non-malignant nasopharyngeal tissues ( p = 0.190, Table ). We analyzed correlations between del-LMP1 and clinicopathologic characteristics, including gender, age, Ann Arbor stage, B symptoms, LDH, vascular invasion, and necrosis, and all results lacked statistical significance (Table ). Detection of LMP1 protein expression IHC staining was used for detecting LMP1 protein expression (Fig. ). As shown in Table , LMP1 protein expression was detected in 21 out of 42 ENKTCL specimens and 4 out of 10 control samples, and the difference between the two groups was not statistically significant ( p = 0.729). We evaluated correlations between LMP1 protein expression and the clinical features of ENKTCL patients. Although most parameters were not statistically significant, LMP1 protein expression in coagulation necrosis tissues increased significantly ( p = 0.030), and there was a trend toward increased expression of LMP1 protein in patients with elevated LDH ( p = 0.075) and younger patients (≤ 60 years old, p = 0.057) (Table ). In the del-LMP1 group, 19 cases were LMP1 protein-positive. There was no correlation between protein expression and 30-bp deletion ( p = 0.489). Table demonstrated that young patients (≤ 60 years old) had significantly higher LMP1 protein expression in the del-LMP1 group than older patients (> 60 years old) ( p = 0.004). Association between del-LMP1, LMP1 protein expression, and survival outcomes Among 35 ENKTCL cases with del-LMP1, the median survival was 18.2 months (range: 0.1-48 months), and the estimated 1- and 3-year OS rates were 77.5% and 62.0%, respectively (Fig. A). Two cases with wt-LMP1 survived, with a follow-up period of 20 months and 42 months, respectively. There was no significant difference in the overall survival between the two groups, but it could be seen that the wt-LMP1 cases had a trend toward a better prognosis ( p = 0.331, Fig. A). The estimated 1-year survival rate was lower in the LMP1 protein-positive group (73.5%) than in the negative group (84.4%), but there was no statistically significant difference in overall survival between LMP1 protein-positive and negative cases ( p = 0.592, Fig. B). In the del-LMP1 group, there was no statistically significant difference in overall survival between LMP1protein-positive and negative cases ( p = 0.580, Fig. C). The characteristics of 42 ENKTCL patients were summarized in Table . There were 31 men and 11 women with a male-to-female ratio of 2.8 to 1. The age range was from 21 to 87 years, with a median age of 55 years and 16 patients over 60 years old. The most common site of occurrence was the UADT ( n = 28); epistaxis, rhinorrhea, and nasal obstruction were initial manifestations. Thirteen patients had B symptoms, and eighteen had elevated serum lactate dehydrogenase (LDH) levels. Thirty-two patients were classified as stages I/II according to the Ann Arbor stage. Morphologically, most cases showed pleomorphic infiltrates, angiodestruction occurred in 22 samples, and variable degrees of necrosis were found in 23 tissues (Fig. A). CD56 (clone UMAB83, Zhongshan Golden Bridge Company, Beijing, China) and Granzyme B (clone EP230, Zhongshan Golden Bridge Company, Beijing, China) were positive in all cases. The median value of Ki-67 (clone UMAB107, Zhongshan Golden Bridge Company, Beijing, China) was 80% (range: 35%-98%). All cases showed EBER-positive signals in tumor cell nuclei (Fig. B). Twenty patients had chemotherapy alone, while twelve received combined chemoradiotherapy. Three patients received autologous stem cell transplants and three more received allogeneic stem cell transplantations. By the end of the follow-up, ten patients had died of disease. The median survival was 19.1 months (range: 0.1-48 months). The 1-year and 3-year estimated survival rates were 79.2% and 69.3%, respectively. The LMP1 gene was successfully amplified from 37/42 ENKTCL cases and 6/10 non-malignant controls, as shown in Table . Two types of PCR products, 251 bp and 221 bp, were obtained following amplification of the LMP1 gene. Representative gel electrophoresis results were shown in Fig. . Sanger sequencing confirmed that the 221 bp product contained a special 30-bp deletion (from nucleotide 168255 to 168285). The results indicated that wt-LMP1 was present in 2 ENKTCL patients, while del-LMP1 was present in 35 ENKTCL patients and 6 non-malignant controls, as depicted in the relevant sequence diagram in Fig. . No statistical difference was found for del-LMP1 between ENKTCL and non-malignant nasopharyngeal tissues ( p = 0.190, Table ). We analyzed correlations between del-LMP1 and clinicopathologic characteristics, including gender, age, Ann Arbor stage, B symptoms, LDH, vascular invasion, and necrosis, and all results lacked statistical significance (Table ). IHC staining was used for detecting LMP1 protein expression (Fig. ). As shown in Table , LMP1 protein expression was detected in 21 out of 42 ENKTCL specimens and 4 out of 10 control samples, and the difference between the two groups was not statistically significant ( p = 0.729). We evaluated correlations between LMP1 protein expression and the clinical features of ENKTCL patients. Although most parameters were not statistically significant, LMP1 protein expression in coagulation necrosis tissues increased significantly ( p = 0.030), and there was a trend toward increased expression of LMP1 protein in patients with elevated LDH ( p = 0.075) and younger patients (≤ 60 years old, p = 0.057) (Table ). In the del-LMP1 group, 19 cases were LMP1 protein-positive. There was no correlation between protein expression and 30-bp deletion ( p = 0.489). Table demonstrated that young patients (≤ 60 years old) had significantly higher LMP1 protein expression in the del-LMP1 group than older patients (> 60 years old) ( p = 0.004). Among 35 ENKTCL cases with del-LMP1, the median survival was 18.2 months (range: 0.1-48 months), and the estimated 1- and 3-year OS rates were 77.5% and 62.0%, respectively (Fig. A). Two cases with wt-LMP1 survived, with a follow-up period of 20 months and 42 months, respectively. There was no significant difference in the overall survival between the two groups, but it could be seen that the wt-LMP1 cases had a trend toward a better prognosis ( p = 0.331, Fig. A). The estimated 1-year survival rate was lower in the LMP1 protein-positive group (73.5%) than in the negative group (84.4%), but there was no statistically significant difference in overall survival between LMP1 protein-positive and negative cases ( p = 0.592, Fig. B). In the del-LMP1 group, there was no statistically significant difference in overall survival between LMP1protein-positive and negative cases ( p = 0.580, Fig. C). ENKTCL is an increasingly recognized EBV-associated disease entity with aggressive clinical behavior and unique clinicopathological features. Due to the important role of LMP1 in immunogenicity and tumorigenicity, its genetic diversity has been extensively studied in the literature . Some studies have shown that LMP1 has a geographical bias , while others have shown that a specific variant of EBV is prevalent, causing disease and malignancy . In this study, LMP1 genes were detected in 37/42 (88.10%) ENKTCL cases, compared to 6/10 (60%) in the control group. This result not only reflected the high rate of EBV infection in ENKTCL but also suggested that LMP1 may be involved in the pathogenesis and development of ENKTCL, which provided a theoretical basis for further study into its mechanisms. However, it is important note that a certain proportion of healthy carriers also harbor the LMP1 gene, indicating that EBV infection may occur before the formation of tumors. We speculated that mature resting B cells infected with EBV remain quiescent in healthy carriers and may be activated under certain conditions, leading to EBV-associated tumorigenesis . In our study, the LMP1 gene was still undetectable in 5 ENKTCL cases and four control tissues. The possible reasons are: (l) The formation and progression of ENKTCL are associated with multiple factors, and the EBV LMP1 gene may not be an essential cause; (2) the LMP1 fragment may be lost during the differentiation and proliferation of lymphoma cells; (3) some cases seem to exhibit latency I ; and (4) EBV-infected mature resting B cells do not need to express latent viral proteins to avoid immune surveillance . The present study revealed that 35/42 (83.3%) of ENKTCL patients and 6/10 (60%) of chronic nasopharyngitis patients had del-LMP1. Some studies suggested that the prevalence of del-LMP1 in ENKTCL might be attributed to the prevalence of this variant in the general population , while others reported a significantly higher association of ENKTCL compared to healthy individuals . Chiang et al. found a significantly higher prevalence of del-LMP1 in ENKTCL (91.3%) compared to normal nasal tissue (62.5%). The observation was further supported by Tai et al. , who reported that del-LMP1 was detected in all UADT lymphomas (100%) and proposed a predilection for del-LMP1 in UADT lymphomas. Despite the limited number of biopsies, majority ENKTCL patients in our study carried del-LMP1. Regarding various clinicopathologic characteristics, including gender, age, Ann Arbor stage, B symptoms, LDH, vascular invasion, and necrosis, del-LMP1 was not correlated with ENKTCL. Although del-LMP1 presumably contributes to the ability to transform during ENKTCL tumorigenesis, it may have no actual effect on the clinical characteristics or the clinicopathological manifestations. In the survival analysis, we found that the two ENKTCL patients with wt-LMP1 survived longer than the median survival time of patients with del-LMP1, suggesting a better prognostic trend. This may be due to del-LMP1’s ability to activate NF-κB abnormally, making cells resistant to apoptosis and promoting lymphoma development . Alternatively, the loss of part of the genetic structure may weaken the immunogenicity of the LMP1 protein, giving tumor cells a survival advantage under immune surveillance, prolonging and enhancing the tumorigenic effect, and maintaining or promoting the evolution of lymphoma . However, given the limited number of wild-type cases, del-LMP1 should be viewed objectively, and it cannot be regarded simply as a pathogenic factor of ENKTCL, but we cannot completely deny that del-LMP1 may play a role in the occurrence and development of this lymphoma. Although ENKTCL is considered to exhibit type II latency, LMP1 protein expression was detected in only 21 (50%) cases in our study, of which 19 positive specimens carried del-LMP1. While previous studies have demonstrated that LMP1 protein expression was highly correlated with B symptoms and anatomic location , our study did not find significant differences. However, our results indicated a trend toward increased LMP1 protein expression in patients with elevated LDH and younger patients (≤ 60 years old). In particular, LMP1 protein expression was significantly higher in young patients (≤ 60 years old) than in elderly patients (> 60 years old) within the del-LMP1 group ( p = 0.004). This may be the reason why the majority of ENKTCL patients are younger than 60 years old (median age in this study: 55 years), and this finding may indirectly support the hypothesis that LMP1 protein products carrying 30-bp deletion exhibit reduced immunogenicity, thereby conferring a survival advantage and enhancing pathogenicity . Further exploration with larger follow-up samples is necessary to establish the correlation between LMP1 protein carrying the 30-bp deletion and patient age. Unlike previous studies, we observed that LMP1-positive tumor cells were more frequently detected in tissues with coagulative necrosis, potentially due to the cytopathic effect of LMP1-positive tumor cells. Tissue necrosis is a key feature of ENKTCL, but not always observed. This may be due to the fact that the biopsy specimens were usually small or due to variations in sampling locations. Additionally, early-stage disease may not yet involve vascular invasion and destruction, leading to tissue ischemia and extensive necrosis . Knecht et al. speculated in their study of HD cases that the strong expression of LMP1 protein in necrotic areas can cause a reduced amount of protein turnover, which may lead to the cytopathic effect of LMP1. The implication of the association between LMP1-positive expression and coagulative necrosis in ENKTCL requires further investigation. Studies by Jiang et al. and Cao et al. have shown that LMP1 expression was associated with an unfavorable prognosis in ENKTCL. In our study, there was no statistically significant difference in overall survival between the LMP1 protein-positive and negative groups. However, we observed a lower 1-year survival rate in the LMP1 protein-positive group than in the negative group. This may be due to the reduced immunogenicity of the highly mutated LMP1 isolate protein, which escapes immune surveillance and promotes lymphoma progression. In recent decade, the detection methods of del-LMP1 have significantly advanced. Next generation sequencing (NGS) technology has identified additional variants, including STAT3, JAK3, STAT5B, MSN, BCOR, DDX3X, TP53, and MGA . Despite its advantages, NGS remains relatively expensive and can raise difficult challenges for analysis and interpretation, limiting its application for diagnostic applications. Sanger sequencing, characterized by high sensitivity, convenient operation, and relatively low cost, may be particularly useful in resource-limited countries. In our experiment, we observed that the detection rate of the LMP1 gene was significantly higher ( p = 0.048) than that of the protein. This may be due to varying expression levels of LMP1 protein during different stages of tumor cells differentiation and maturation, or other factors affecting transcription and translation levels, leading to decreased LMP1 protein expression or reduced immunogenicity of mutant LMP1 isolates. It may also be related to technical reasons, such as the difficulty of antigen repair after formalin treatment. The specific reasons need to be further studied and discussed. Our study demonstrated that del-LMP1 is the most prevalent variant in our population, but it was not associated with overall survival or any investigated clinicopathological characteristics of ENKTCL. Therefore, from the viewpoint of molecular epidemiology, del-LMP1 appears to be the predominant variant than a phenotype-associated polymorphism in ENKTCL. Additionally, we observed that LMP1 protein expression was more common in younger patients with del-LMP1 and in tissues with necrosis. These findings contribute to the understanding of the information on LMP1 characteristics in ENKTCL patients in Wenzhou. Expanding the sample size and continuing follow-up will be essential to fully elucidate the significance of LMP1 in ENKTCL, including its predictive and prognostic value. |
Integrated Genetic Diversity and Multi-Omics Analysis of Colour Formation in Safflower | 01c6e8ea-9507-426b-b880-28901ad15c19 | 11765828 | Biochemistry[mh] | Safflower ( Carthamus tinctorius L.) is a flowering annual plant of the Asteraceae family that originated in the Fertile Crescent between southern Israel and western Iraq, where it was first cultivated about 4000 years ago . Vavilov suggested that the three centres of origin of safflower are India, Afghanistan, and Ethiopia . According to the Food and Agriculture Organization of the United Nations (FAO), the global cultivation area of safflower is 1,204,335 hectares, with a production of 1,002,023.32 tons. The major production areas for safflower seeds are Asia and the Americas, accounting for 90% of cumulative production . In China, safflower was introduced to the Central Plains more than 2000 years ago during Zhang Qian’s mission to the West during the Han Dynasty. Currently, Xinjiang, Yunnan, Sichuan, and Gansu provinces are the major production areas, with a cultivation area of 23,069 ha and production of 33,879.09 tons . Safflower is a multipurpose crop that was originally grown as a natural dye and food flavouring . Since the first half of the 20th century, it has been grown as an oilseed crop, and its seeds are rich in unsaturated fatty acids (oleic, linoleic, and linolenic acids), which is directly linked to cardio-protective, hypolipidemic, anti-atherosclerotic, and anti-inflammatory effects . The planting of safflower is extended to other regions worldwide given the gradual attention paid to its efficacy. However, its cultivation is affected by factors such as the yield, low oil content, susceptibility to a variety of biotic stresses, and the presence of thorns . Therefore, the conservation and sustainable use of genetic resources are key to the continuous improvement of safflower. Genetic diversity is the foundation of biodiversity and the driving force behind the stability and continued evolution of species . With the development of science and technology, molecular marker technology has become one of the main ways to study genetic diversity, population structure, and relatedness. The limited genomic data of safflower compared with its major crops limit the use of applicable molecular markers. Randomly amplified polymorphic DNA (RAPD) , amplified fragment length polymorphisms (AFLPs) , simple sequence repeats (SSRs) , and sequence-related amplified polymorphism (SRAP) are used to identify and evaluate safflower varietal resources. However, these molecular markers can only obtain limited genetic information in populations. With the release of a high-quality safflower genome , the insertion/deletion (Indel) marker has been developed with 11 oil traits to construct the safflower core germplasm . This Indel marker has easy to design primers, basic PCR systems, and agarose gel electrophoresis for visualisation . Therefore, the construction of the representative core germplasm by combining Indel markers with agronomic trait data provides a means to increase the degree of variation and genetic diversity of the core collection. As a visible feature of plants, colour is important for their growth and development. The flower colour depends on the plant pigments such as flavonoids, carotenoids, and betalains . Anthocyanins are water-soluble pigments belonging to the flavonoid family and contribute mainly to red and blue colours. Carotenoids are the natural pigments that impart yellow, red, or orange colours to flowers, and most plants have similar contents, including β, ε-carotenoids and β, β-carotenoids in their green tissues. For example, deep red or purple chrysanthemums have high levels of anthocyanins, while deep yellow or green chrysanthemums have high amounts of carotenoids . In general, flavonoid biosynthetic pathways begin with phenylacetone biosynthesis, and chalcone synthase produces the chalcone scaffold from which all flavonoids are derived. The quinoxaline chalcone C-glycoside is one of the chalcone scaffolds present only in safflower species. Almost all red and yellow pigments in safflower petals belong to the C-glycoside quinoxalinone family of flavonoids. Hydroxysafflor yellow A (HSYA) is the main component of the flavonoids in safflower and consists of a C-glucosyl quinoxalinone. C-glucosyl quinuclidin chalcone is formed by the glycosylation of naringenin chalcone by Cytokinin-O-Glucosyltransferase (CGT), followed by oxidation by cytochrome P450 (P450). In addition, a high-performance liquid chromatography (HPLC) analysis of yellow pigments in the aqueous extracts of safflower found that the predominant constituent was HSYA, followed by anhydrosafflor yellow B (AHSYB) . Although the biosynthetic pathways of flavonoids and carotenoids in other plants are well-established, the balance of the expression dynamics of the relevant genes during safflower colour transition is still poorly understood. Changes in flower colour are comprehensively regulated by physiological changes, metabolite accumulation, and fluctuations in the transcript levels of relevant genes . Combined metabolite–transcriptome analyses can elucidate the interactions between gene expression and metabolite accumulation, providing insights into the regulatory mechanisms . So far, the integration of genomics, transcriptomics, and metabolomics has emerged a powerful tool to understand the complex network controlling changes in plant flower colour . It has been widely applied to many plants such as tomato , watermelon , wolfberry , Iris sanguinea , Lagerstroemia indica , and other plants. Therefore, combining the advantages of multi-omics analysis can provide a more comprehensive understanding of plant variation in flower colour, which ultimately contributes to the improvement of crop traits. Based on research findings from other plants, we hypothesise that the variation in flower colour of safflower may be related to the ratio of anthocyanin and carotenoid content. To verify this hypothesis, we developed Indel markers using safflower whole-genome and RNA sequence data, analysed the genetic diversity of 614 safflower germplasm resources collected, and constructed core germplasms. Transcriptome and metabolome analyses were also used to resolve significantly differentially accumulated metabolites and significantly different genes in three different colours of safflower. This work provides a basis for the research on the dynamic change of metabolite accumulation and the expression of the key genes involved in safflower colour formation. 2.1. Genetic Diversity Analysis of Safflower Germplasm An analysis of variance, correlation analysis, and principal component analysis were performed on 614 safflower germplasm. The results, in combination with agronomic traits data, were used to evaluate the population structure of safflower populations and effectively screen the specific germplasm. Differences were observed in seven morpho-quantitative traits of the safflower germplasm with a variation range of 29.61–77.35% . Hence, the studied safflower germplasm is rich in genetic diversity. Various degrees of correlation were observed among the agronomic traits, with most traits exhibiting significant or highly significant correlations with each other. Plant height showed a positive correlation with height of the top branch (r = 0.89 **) and height of primary branches (r = 0.68 **). The height of primary branches showed a negative correlation with the number of heads per plant (r = −0.49 **) and a positive correlation with the height of the top branch (r = 0.74 **) ( A). The results of the principal component analysis (PCA) of the 11 traits of the 615 accessions indicated that the first four principal components explained 70.34% of the total variation . The PCA was performed by combined PCA1 and PCA2 with geographic distribution ( B). The results demonstrated no obvious pattern in the distribution of different germplasm resources and a large distribution range, which confirmed the extensive genetic diversity in 614 safflower germplasm resources. A polymorphism analysis was performed on 50 pairs of Indel markers to verify the polymorphism of the developed Indel markers . The mean value of the Shannon’s information index ( I ) was 0.551, and the polymorphism information content ( PIC ) ranged from 0.107 to 0.375, with a mean value of 0.296. An unrooted phylogenetic tree was constructed by using the neighbour-joining method with Indel markers, and it showed that 614 safflower germplasms were classified into three clades ( A). The maximum ΔK value was found when K = 3 , suggesting that the 614 safflower accessions could be divided into three subgroups (STR I, STR II, STR III). Based on the Q-matrix and maximum membership coefficient, 87 safflower accessions had Q values less than 0.6, and the subgroup was STR MIX. According to the results of the AMOVA analysis, the inter-subpopulation variation in safflower accounted for 23.50% of the total variation, and most of the variation existed within the subpopulations . Fst ranged from 0.069 to 0.425, indicating great genetic differentiation between STR I and STR III, and between STR II and STR III . The principal coordinate analysis (PCoA) was carried out on 614 safflower varieties, and its results showed that four subgroups could be distinguished ( B). Phenotypic data and genotypic data of safflower were screened using Core Hunter version 3 software. Two subsets, namely, CtCore 1 and CtCore 2, were generated , and 214 core germplasms of safflower were obtained by combining the CtCore 1 and CtCore 2 subsets and removing the overlap. The evaluation using Indel markers demonstrated that the 214 safflower core germplasms are more representative than that of the 614 safflower germplasms . The mean difference (MD) percentages of the core collections were less than 20%, and the range coincidence rates (CRs) were greater than 80%, indicating that all core collections met the conditions. The mean difference rate, variance difference rate (VD), range coincidence rate, and coefficient of variation coincidence rate (VR) were 0.00%, 57.14%, 98.14%, and 110.92%, respectively . The results of the comparison of the distribution of the core germplasm in the original germplasm based on agronomic traits and Indel markers, respectively, revealed that the core subsets formed by CtCore 1 and CtCore 2 were more evenly distributed in both cluster methods . Therefore, the constructed core germplasm can effectively represent the population diversity of the original materials in the evaluation of agronomic traits. 2.2. Correlation Analysis Between Phenotype and Indel Makers The association analysis of safflower agronomic traits and Indel makers was performed using TASSEL version 5.0. A total of 16 loci were associated with PH, HN, PBN, PBH, TBH, SBN, and FC at p < 0.05. The cis-acting element analysis of the promoter region revealed that Loci4, 24, 29, 32, and 37 were associated with FC . Loci37 was the MYB binding site for the flavonoid metabolism regulatory genes. MYB transcription factors are widely involved in the synthesis of flavonoids, and the flavonoid metabolism pathway is the main regulatory pathway of safflower flower colour change. Thus, transcriptome and qRT-PCR analyses were performed on the genes where these five loci are located . Loci24 is relatively highly expressed at Rb, and the gene at this locus may be involved in regulating the growth of safflower buds. Loci32 is relatively highly expressed at Rs and Rb and may be involved in the formation and growth of safflower filaments and buds. Loci37 has a MYB binding site and is involved in the regulatory pathway of safflower flavonoid metabolism by binding to MYB transcription factors; it is relatively highly expressed at Wb. Loci4 is expressed at the same level in Rs, Rb, Ws, and Wb, and may be involved in the formation of flower colour through other transcriptional regulatory pathways. Loci29 is mainly involved in the formation of small flower buds in safflower and in the regulation of large flower buds in white flowers. 2.3. Transcriptome and Metabolome Analyses to Compare Different Colours of Safflower To study the physiological mechanism of different flower colour changes in safflower, we performed widely targeted metabolic profiling on samples of three different colours, namely, red (R), yellow (Y), and white (W), of safflower . The variation in the metabolite composition of the three different colours of safflower was assessed using LC-MS/MS, and a total of 627 compounds were detected, such as flavonoids, pyridines and derivatives, carboxylic acids and derivatives, terpenoids, and organic acids. All the metabolic species were annotated through KEGG (HSYA are special metabolic species of safflower that were not annotated by KEGG). The PCA of the samples revealed a distinct dispersion between groups and tight aggregation within groups, indicating that the sampling results were stable and reproducible . The significant DAMs between pairwise comparisons were screened based on the variable’s importance in projection (VIP) ≥ 1 and p -value < 0.05, which indicated that 88, 96, and 83 DAMs were detected in the three comparison groups (R vs. W, R vs. Y, and W vs. Y) . The scores plot of the OPLS-DA model discriminated the flower colours, with different colour groups all exhibiting satisfactory separation, as reflected by all samples having a significantly different metabolite composition . The loading plot in the scores plot of the OPLS-DA model with different colour groups shows that HSYA, oxazepam glucuronide, quercetin 3-(6″-malonyl-glucoside), etc., are dominant metabolites in the R group compared with the W group. In parallel, astragalin, luteolin 7-galactoside, and kaempferol 3-alpha-D-galactoside are dominant metabolites in the W group . Based on the comparison of the R group and Y groups, riboflavin, HSYA, and quercetin-3-(6″-malonyl)-glucoside are also the main metabolites in the R group, and luteolin 7-galactoside, isoquercitrin, and alpha-curcumene are the main metabolites in the Y group . Kaempferol 3-alpha-D-galactoside, astragalin, and luteolin 7-galactoside are the dominant metabolites in the W group compared with that in the Y group; biorobin, datiscin, quercetin-3,4′-O-di-beta-glucopyranoside, etc., are the dominant metabolites in the Y group . Most of the DAMs were enriched to the flavonoid biosynthesis (map00941), such as astragalin, luteolin 7-galactoside, isoquercitrin, cyanidin 3-glucoside, HSYA, etc. . The k-means of DAMs were studied to investigate trends in the relative content of metabolites in different colour groups. The metabolites were well-distinguished at k = 8. Metabolites in Cluster1 were significantly higher in the W group than in the R and Y groups, indicating that these metabolites, mainly including kaempferol, quercetin, coumarin, etc., were dominant in white safflower ( A). The content of metabolites in Cluster3 was significantly higher in group Y than in the two remaining groups, and they were the major metabolites of yellow safflower (mainly including isorhamnetin, petunia pigments, herbaceous pigments, etc.). The dominant metabolites of the R group are in Cluster6 and include HSYA, lignans, catechin, etc. The hierarchical cluster analysis of all differentially expressed metabolites in Cluster1, Cluster3, and Cluster6 showed that many metabolites were highly expressed in the R and W groups ( B). Applying filtering criteria of |log2Fold Change| ≥ 1 and FDR < 0.05, a total of 2636 DEGs were detected in the three compared combinations of R vs. W, R vs. Y, and W vs. Y; 716, 1165, and 755 DEGs were identified in the compared combinations ( A). The annotation of the differential genes using the gene ontology (GO) resource revealed that they are involved in a variety of functions and pathways. The anthocyanin-containing compound biosynthetic process (GO:0009718), glucose metabolic process (GO:0006006), UDP-glycosyltransferase activity (GO:0008194), and xyloglucan metabolic process (GO:0010411) were upregulated (GO:0010411) in the R vs. W compared combination; beta-galactosidase activity (GO:0004565), quercetin 7-O-glucosyltransferase activity (GO:0080044), the L-phenylalanine catabolic process (GO:0006559), and flavonoid biosynthetic process (GO:0009813) were downregulated ( C). In the R vs. Y compared combination, the anthocyanin-containing compound biosynthetic process (GO:0009718), jasmonic acid hydrolase (GO:0120091), and carotenoid biosynthetic process (GO:0016117) were upregulated, and Delta12-fatty-acid desaturase activity (GO:0102985), beta-galactosidase activity (GO:0004565), leucocyanidin oxygenase activity (GO:0050589), and quercetin 7-O-glucosyltransferase activity (GO:0080044) were downregulated ( D). In the R colour group combinations (R vs. W and R vs. Y), the anthocyanin-containing compound biosynthetic process was enriched, and quercetin 7-O-glucosyltransferase activity and beta-galactosidase activity were enriched in both the Y and W groups. Meanwhile, jasmonic acid hydrolase (GO:0120091), cellulose synthase (UDP-forming) activity (GO:0016760), the flavonoid biosynthetic process (GO:0009813) and quercetin 7-O-glucosyltransferase activity (GO:0080044) were upregulated in the W vs. Y compared combination. The downregulated GO items included the auxin metabolic process (GO:0009850), response to hydrogen peroxide (GO:0042542), monooxygenase activity (GO:0004497), cinnamoyl-CoA reductase activity (GO:0016621), etc. ( E). Therefore, the expression levels of key enzyme genes in the flavonoid biosynthesis pathway were investigated. The expression of most upstream key enzyme genes, such as CHS, phenylalanine aminotransferase (PAL), and 4-coumarate-CoA ligase (4CL), was higher in groups W and Y than in the R group; meanwhile, the expression of downstream key enzyme genes, such as F3H, ANS, and BZ1, was higher than that of the R group ( B). The metabolites of the flavonoid biosynthesis pathway in different coloured safflowers varied considerably due to the absence of HSYA in the W group and the different levels of HSYA in the R and Y groups. The metabolic species and the gene expression levels were integrated into a schematic diagram . In this pathway, L-phenylalanine is converted into p-coumaroyl-CoA by phenylalanine ammonia lyase (PAL), cinnamate-4-hydroxylase (C4H), and 4-coumarate CoA ligase (4CL); this process is common to many secondary metabolism pathways. Then, chalcone synthase (CHS) produced chalcones such as naringenin chalcone and C-glucosyl quinochalcones. The pathway of HSYA converted from C-glucosyl quinochalcones is still not well-understood. Key enzyme genes from L-phenylalanine to dihydrokaempferol have higher expression levels in the W and Y groups than in the R group; from dihydrokaempferol, most enzymes in the anthocyanin metabolic pathway such as ANS, FLS, and BZ1, were highly expressed in the R group. The comparison of the R and Y groups showed a distraction beginning downstream from quercetin. Isoquercitrin, quercetin, and glucuronide were higher in the Y group, and rutin was high in the R group. Those flavonoids from naringenin in the R group and W group show different flow directions. In the W group, apigenin and kaempferol flowed more, whereas in the R group, the anthocyanin metabolic pathway flowed more. Apart from that, the W and Y group safflower flavonoids were different from dihydrokaempferol. The content of safflower yellow (SY) in safflower is about 20–30%, and HYSA is one of the constituent compounds, which account for 80–90% of the total SY. To investigate the other pigments in safflower, we extracted anthocyanin and carotenoid from three different colours of safflowers, R, Y, and W. The contents were initially determined by the spectrophotometric method. The anthocyanin extracts showed different colours among the three colours groups; the R group had deep-red colours, the Y group had yellow colours, and the W group had less-yellow colours ( A). The absorbance of these extracts was recorded at 520 nm and quantified in each colour group. In comparison with the other groups, the anthocyanin content was higher in the R group, and had an average value of 10 mg/g. The anthocyanin content was very low in groups Y and W ( B). The extraction of carotenoids from the three colour groups showed that the Y group is yellow in colour and the W group is nearly transparent. The absorbance of carotenoids was recorded at 440 nm. Compared with that in Y group, the total carotenoid content was higher in the R group and lower in the W group ( C,D). An analysis of variance, correlation analysis, and principal component analysis were performed on 614 safflower germplasm. The results, in combination with agronomic traits data, were used to evaluate the population structure of safflower populations and effectively screen the specific germplasm. Differences were observed in seven morpho-quantitative traits of the safflower germplasm with a variation range of 29.61–77.35% . Hence, the studied safflower germplasm is rich in genetic diversity. Various degrees of correlation were observed among the agronomic traits, with most traits exhibiting significant or highly significant correlations with each other. Plant height showed a positive correlation with height of the top branch (r = 0.89 **) and height of primary branches (r = 0.68 **). The height of primary branches showed a negative correlation with the number of heads per plant (r = −0.49 **) and a positive correlation with the height of the top branch (r = 0.74 **) ( A). The results of the principal component analysis (PCA) of the 11 traits of the 615 accessions indicated that the first four principal components explained 70.34% of the total variation . The PCA was performed by combined PCA1 and PCA2 with geographic distribution ( B). The results demonstrated no obvious pattern in the distribution of different germplasm resources and a large distribution range, which confirmed the extensive genetic diversity in 614 safflower germplasm resources. A polymorphism analysis was performed on 50 pairs of Indel markers to verify the polymorphism of the developed Indel markers . The mean value of the Shannon’s information index ( I ) was 0.551, and the polymorphism information content ( PIC ) ranged from 0.107 to 0.375, with a mean value of 0.296. An unrooted phylogenetic tree was constructed by using the neighbour-joining method with Indel markers, and it showed that 614 safflower germplasms were classified into three clades ( A). The maximum ΔK value was found when K = 3 , suggesting that the 614 safflower accessions could be divided into three subgroups (STR I, STR II, STR III). Based on the Q-matrix and maximum membership coefficient, 87 safflower accessions had Q values less than 0.6, and the subgroup was STR MIX. According to the results of the AMOVA analysis, the inter-subpopulation variation in safflower accounted for 23.50% of the total variation, and most of the variation existed within the subpopulations . Fst ranged from 0.069 to 0.425, indicating great genetic differentiation between STR I and STR III, and between STR II and STR III . The principal coordinate analysis (PCoA) was carried out on 614 safflower varieties, and its results showed that four subgroups could be distinguished ( B). Phenotypic data and genotypic data of safflower were screened using Core Hunter version 3 software. Two subsets, namely, CtCore 1 and CtCore 2, were generated , and 214 core germplasms of safflower were obtained by combining the CtCore 1 and CtCore 2 subsets and removing the overlap. The evaluation using Indel markers demonstrated that the 214 safflower core germplasms are more representative than that of the 614 safflower germplasms . The mean difference (MD) percentages of the core collections were less than 20%, and the range coincidence rates (CRs) were greater than 80%, indicating that all core collections met the conditions. The mean difference rate, variance difference rate (VD), range coincidence rate, and coefficient of variation coincidence rate (VR) were 0.00%, 57.14%, 98.14%, and 110.92%, respectively . The results of the comparison of the distribution of the core germplasm in the original germplasm based on agronomic traits and Indel markers, respectively, revealed that the core subsets formed by CtCore 1 and CtCore 2 were more evenly distributed in both cluster methods . Therefore, the constructed core germplasm can effectively represent the population diversity of the original materials in the evaluation of agronomic traits. The association analysis of safflower agronomic traits and Indel makers was performed using TASSEL version 5.0. A total of 16 loci were associated with PH, HN, PBN, PBH, TBH, SBN, and FC at p < 0.05. The cis-acting element analysis of the promoter region revealed that Loci4, 24, 29, 32, and 37 were associated with FC . Loci37 was the MYB binding site for the flavonoid metabolism regulatory genes. MYB transcription factors are widely involved in the synthesis of flavonoids, and the flavonoid metabolism pathway is the main regulatory pathway of safflower flower colour change. Thus, transcriptome and qRT-PCR analyses were performed on the genes where these five loci are located . Loci24 is relatively highly expressed at Rb, and the gene at this locus may be involved in regulating the growth of safflower buds. Loci32 is relatively highly expressed at Rs and Rb and may be involved in the formation and growth of safflower filaments and buds. Loci37 has a MYB binding site and is involved in the regulatory pathway of safflower flavonoid metabolism by binding to MYB transcription factors; it is relatively highly expressed at Wb. Loci4 is expressed at the same level in Rs, Rb, Ws, and Wb, and may be involved in the formation of flower colour through other transcriptional regulatory pathways. Loci29 is mainly involved in the formation of small flower buds in safflower and in the regulation of large flower buds in white flowers. To study the physiological mechanism of different flower colour changes in safflower, we performed widely targeted metabolic profiling on samples of three different colours, namely, red (R), yellow (Y), and white (W), of safflower . The variation in the metabolite composition of the three different colours of safflower was assessed using LC-MS/MS, and a total of 627 compounds were detected, such as flavonoids, pyridines and derivatives, carboxylic acids and derivatives, terpenoids, and organic acids. All the metabolic species were annotated through KEGG (HSYA are special metabolic species of safflower that were not annotated by KEGG). The PCA of the samples revealed a distinct dispersion between groups and tight aggregation within groups, indicating that the sampling results were stable and reproducible . The significant DAMs between pairwise comparisons were screened based on the variable’s importance in projection (VIP) ≥ 1 and p -value < 0.05, which indicated that 88, 96, and 83 DAMs were detected in the three comparison groups (R vs. W, R vs. Y, and W vs. Y) . The scores plot of the OPLS-DA model discriminated the flower colours, with different colour groups all exhibiting satisfactory separation, as reflected by all samples having a significantly different metabolite composition . The loading plot in the scores plot of the OPLS-DA model with different colour groups shows that HSYA, oxazepam glucuronide, quercetin 3-(6″-malonyl-glucoside), etc., are dominant metabolites in the R group compared with the W group. In parallel, astragalin, luteolin 7-galactoside, and kaempferol 3-alpha-D-galactoside are dominant metabolites in the W group . Based on the comparison of the R group and Y groups, riboflavin, HSYA, and quercetin-3-(6″-malonyl)-glucoside are also the main metabolites in the R group, and luteolin 7-galactoside, isoquercitrin, and alpha-curcumene are the main metabolites in the Y group . Kaempferol 3-alpha-D-galactoside, astragalin, and luteolin 7-galactoside are the dominant metabolites in the W group compared with that in the Y group; biorobin, datiscin, quercetin-3,4′-O-di-beta-glucopyranoside, etc., are the dominant metabolites in the Y group . Most of the DAMs were enriched to the flavonoid biosynthesis (map00941), such as astragalin, luteolin 7-galactoside, isoquercitrin, cyanidin 3-glucoside, HSYA, etc. . The k-means of DAMs were studied to investigate trends in the relative content of metabolites in different colour groups. The metabolites were well-distinguished at k = 8. Metabolites in Cluster1 were significantly higher in the W group than in the R and Y groups, indicating that these metabolites, mainly including kaempferol, quercetin, coumarin, etc., were dominant in white safflower ( A). The content of metabolites in Cluster3 was significantly higher in group Y than in the two remaining groups, and they were the major metabolites of yellow safflower (mainly including isorhamnetin, petunia pigments, herbaceous pigments, etc.). The dominant metabolites of the R group are in Cluster6 and include HSYA, lignans, catechin, etc. The hierarchical cluster analysis of all differentially expressed metabolites in Cluster1, Cluster3, and Cluster6 showed that many metabolites were highly expressed in the R and W groups ( B). Applying filtering criteria of |log2Fold Change| ≥ 1 and FDR < 0.05, a total of 2636 DEGs were detected in the three compared combinations of R vs. W, R vs. Y, and W vs. Y; 716, 1165, and 755 DEGs were identified in the compared combinations ( A). The annotation of the differential genes using the gene ontology (GO) resource revealed that they are involved in a variety of functions and pathways. The anthocyanin-containing compound biosynthetic process (GO:0009718), glucose metabolic process (GO:0006006), UDP-glycosyltransferase activity (GO:0008194), and xyloglucan metabolic process (GO:0010411) were upregulated (GO:0010411) in the R vs. W compared combination; beta-galactosidase activity (GO:0004565), quercetin 7-O-glucosyltransferase activity (GO:0080044), the L-phenylalanine catabolic process (GO:0006559), and flavonoid biosynthetic process (GO:0009813) were downregulated ( C). In the R vs. Y compared combination, the anthocyanin-containing compound biosynthetic process (GO:0009718), jasmonic acid hydrolase (GO:0120091), and carotenoid biosynthetic process (GO:0016117) were upregulated, and Delta12-fatty-acid desaturase activity (GO:0102985), beta-galactosidase activity (GO:0004565), leucocyanidin oxygenase activity (GO:0050589), and quercetin 7-O-glucosyltransferase activity (GO:0080044) were downregulated ( D). In the R colour group combinations (R vs. W and R vs. Y), the anthocyanin-containing compound biosynthetic process was enriched, and quercetin 7-O-glucosyltransferase activity and beta-galactosidase activity were enriched in both the Y and W groups. Meanwhile, jasmonic acid hydrolase (GO:0120091), cellulose synthase (UDP-forming) activity (GO:0016760), the flavonoid biosynthetic process (GO:0009813) and quercetin 7-O-glucosyltransferase activity (GO:0080044) were upregulated in the W vs. Y compared combination. The downregulated GO items included the auxin metabolic process (GO:0009850), response to hydrogen peroxide (GO:0042542), monooxygenase activity (GO:0004497), cinnamoyl-CoA reductase activity (GO:0016621), etc. ( E). Therefore, the expression levels of key enzyme genes in the flavonoid biosynthesis pathway were investigated. The expression of most upstream key enzyme genes, such as CHS, phenylalanine aminotransferase (PAL), and 4-coumarate-CoA ligase (4CL), was higher in groups W and Y than in the R group; meanwhile, the expression of downstream key enzyme genes, such as F3H, ANS, and BZ1, was higher than that of the R group ( B). The metabolites of the flavonoid biosynthesis pathway in different coloured safflowers varied considerably due to the absence of HSYA in the W group and the different levels of HSYA in the R and Y groups. The metabolic species and the gene expression levels were integrated into a schematic diagram . In this pathway, L-phenylalanine is converted into p-coumaroyl-CoA by phenylalanine ammonia lyase (PAL), cinnamate-4-hydroxylase (C4H), and 4-coumarate CoA ligase (4CL); this process is common to many secondary metabolism pathways. Then, chalcone synthase (CHS) produced chalcones such as naringenin chalcone and C-glucosyl quinochalcones. The pathway of HSYA converted from C-glucosyl quinochalcones is still not well-understood. Key enzyme genes from L-phenylalanine to dihydrokaempferol have higher expression levels in the W and Y groups than in the R group; from dihydrokaempferol, most enzymes in the anthocyanin metabolic pathway such as ANS, FLS, and BZ1, were highly expressed in the R group. The comparison of the R and Y groups showed a distraction beginning downstream from quercetin. Isoquercitrin, quercetin, and glucuronide were higher in the Y group, and rutin was high in the R group. Those flavonoids from naringenin in the R group and W group show different flow directions. In the W group, apigenin and kaempferol flowed more, whereas in the R group, the anthocyanin metabolic pathway flowed more. Apart from that, the W and Y group safflower flavonoids were different from dihydrokaempferol. The content of safflower yellow (SY) in safflower is about 20–30%, and HYSA is one of the constituent compounds, which account for 80–90% of the total SY. To investigate the other pigments in safflower, we extracted anthocyanin and carotenoid from three different colours of safflowers, R, Y, and W. The contents were initially determined by the spectrophotometric method. The anthocyanin extracts showed different colours among the three colours groups; the R group had deep-red colours, the Y group had yellow colours, and the W group had less-yellow colours ( A). The absorbance of these extracts was recorded at 520 nm and quantified in each colour group. In comparison with the other groups, the anthocyanin content was higher in the R group, and had an average value of 10 mg/g. The anthocyanin content was very low in groups Y and W ( B). The extraction of carotenoids from the three colour groups showed that the Y group is yellow in colour and the W group is nearly transparent. The absorbance of carotenoids was recorded at 440 nm. Compared with that in Y group, the total carotenoid content was higher in the R group and lower in the W group ( C,D). Agronomic traits are unique to crop varieties and are determined by their genetic background and environmental influences. In this study, we analysed the agronomic traits of 614 safflower varieties. The coefficients of variation of seven quantitative agronomic traits were greater than 25%, indicating significant genetic variation in the population. The correlation results indicated that the agronomic traits of safflower may have mutual constraints and influences and therefore should be considered and analysed comprehensively during germplasm creation. The PCA showed that the distribution trends of the safflower germplasm in China and abroad differed somewhat, indicating that the size of the range of the selected area and the intricacies of the environment should be fully considered. Meanwhile, the population structure analysis, phylogenetic analysis, and PCoA results showed that the collection of 614 safflower germplasms was assigned to four subgroups. Kumar et al. worked on 135 different safflower genotypes with SRAP and SSR markers, which showed four clusters within the safflower germplasm, while there was no strong correlation between geographic origin and estimated genetic diversity . Based on AMOVA, the genetic diversity within populations was more pronounced than between populations, and most of the genetic variation was present within populations. This finding is consistent with other results .These results indicate that the 614 safflower germplasm is highly differentiated and rich in genetic diversity, which provides valuable material for molecular breeding. The core germplasm can represent the genetic diversity level of the original population to the greatest extent by selecting a small, representative genetic resource from the whole, and can maximise the genetic diversity of the germplasm in a limited breeding cycle . This is because agronomic trait data can be used to visualise plant differences and are easy to manipulate. However, agronomic traits are greatly influenced by the environment, especially for the germplasm collected from different regions . Therefore, core germplasm populations are usually constructed based on molecular data, supplemented by agronomic trait data. The exploitation of the genetic diversity of genetic resources in safflower is important to construct the core germplasm and for improving breeding studies. In this paper, 214 core germplasm sources of safflower were constructed by combined agronomic traits and Indel markers using Core Hunter 3. The results showed high levels of dependence for most parameters, indicating that the selected core germplasm maintained the genetic diversity of the native germplasm. The established safflower core collection can be used in future genome-wide association studies. It can also play an important role in diversifying the genetic base of the working collection as well as in breeding programs. Therefore, the construction method has been validated in species such as Robinia pseudoacacia L. and Juglans regia L. . In recent years, the application of association analysis in plant molecular marker-assisted breeding has effectively accelerated the selection of superior varieties. Especially in crops, 91 significant trait-related SNPs were detected by marker-trait association analysis using GLM, MLM, and mrMLM in cotton; among them, 33 were related to PSB, 21 were related to SB, and 37 were related to AB . A total of 96 loci were identified for eight agronomic traits using GLM and MLM in safflower . Transcriptomic and qRT-PCR analyses of the genes for five loci (Loci4, Loci24, Loci29, Loci32 and Loci37) in the association analyses showed that they were associated with the formation of flower colour, indicating that the phenotypic identification of the Indel marker screen was consistent. Based on these results, the reliability of the marker-phenotype association was demonstrated, and the applicability of the core germplasm constructed on the basis of Indel markers was illustrated. However, it is not sufficient to identify putative candidate genes, which may be involved in complex biological pathways, because there is still a lot of information missing. Fully understanding these metabolic pathways could help to elucidate the precise role of these genes affecting particular traits and could be a good starting point to obtain high-quality safflower. Thus, additional studies, such as QTL mapping and functional validation by means of molecular cloning, will be required in the future. Flower colour is the result of the synergistic accumulation of several metabolites, especially flavonoids and carotenoids. The expression levels of structural genes were further examined in this study. Based on the transcriptome results, the anthocyanin biosynthetic process and most metabolic processes associated with glycosylation were enriched in the R group, indicating that anthocyanins are generally found in the form of glycosides. The flavonoid biosynthetic process, L-phenylalanine catabolic process, and catechol oxidase activity were enriched in the W group, which indicates that more flavonoid compounds are accumulated in white flowers. On the significant DEGs, most of the upstream expressed genes such as 4CL , CHI , and CHS have higher expression levels in the W and Y groups compared with the R group, and the majority of enzymes in the anthocyanin metabolic pathway, such as ANS , FLS , and BZ1 , were highly expressed in the R group. The expression of these genes is increased and the metabolites are increased through catalysis by various enzymes, which ultimately leads to the formation of more colourless anthocyanins, which were catalysed into coloured anthocyanins by the dehydrogenation, isomerization, and dehydration of the ANS genes under acidic conditions . This probably a consequence of different types of flavonoid compounds like flavones, flavanols, and anthocyanins. Differences in accumulated levels were observed among the three colour groups of safflowers; anthocyanins were accumulated in the R group, whereas anthocyanins were lacking in white flower, and more colourless flavonoids like apigenin, kaempferol, and naringenin were accumulated in the W group. These findings are supported by previous studies that have emphasized the importance of transcriptome analysis in understanding flavonoid biosynthesis. For example, Hong et al. used the mechanism of the effect of colour change on CtbHLH41 , CtMYB63 , and CtWD40-6 of Sichuan safflower to demonstrate that CtMYB63/CtWD40-6 enhances the transcriptional activity of CtbHLH41 on CtDFR (dihydroflavonol 4-reductase) to accumulate anthocyanins . Hong et al. found that the differential expression of F3′5′H and F3′H is a factor in the colour diversity of crape myrtle . Based on the metabolome results, naringenin chalcone is catalysed by a series of enzymes to produce naringenin, and the W group safflower metabolites downstream of naringenin begin to differ from those of the R group and Y group safflowers. The metabolites of the W and Y groups of safflowers flowed more to dihydrokaempferol and downstream of kaempferol and quercetin. Metabolites in the R group safflower flowed more to eriodictyol and anthocyanidins. It is the different metabolic flows in naringenin and eriodictyol that lead to the different contents of delphinidin and downstream anthocyanin-related products, which may be the main reason for the formation of different flower colours. These findings are supported by previous studies, in which Wang et al., using transcriptomic and metabolomic analyses, found that differential expression of the CHS gene was one of the main reasons for the differences in flavonoid species and content between different coloured safflowers . Further, the initial quantification of anthocyanin and carotenoid contents in different coloured safflower flowers (red, yellow, and white) with extracted anthocyanins and carotenoids showed that the differences in safflower flower colours were not only related to flavonoid chemistry, but also co-determined the flower colours together with several other pigments, which led to a rich variety of colours in the flowers. These findings provide a wealth of data for future use in genome-wide association analysis. 4.1. Plant Materials and Samples Collection A total of 614 safflower materials were provided by the germplasm repository of the Wuhan Institute of Oilseed Crops, Chinese Academy of Agricultural Sciences. The safflower material was planted in an experimental field at 58 m above sea level in Huangpi District, Wuhan City, Hubei Province, China (30°45′59″ N, 114°27′46″ E). Each safflower variety was sown in a row 4 m long, 50 cm apart, with an average plant spacing of 40 cm. The total number of seedlings left in each row after interplanting was 10. Agronomic traits were investigated following the guidelines provided by the Oils Crops Research Institute of the Chinese Academy of Agricultural Sciences for safflower . These guidelines encompass the following: plant height (PH), primary branch height (PBH), terminal branch height (TBH), number of primary branches (PBNs), number of secondary branches (SBNs), seed germination rate (SGR), load resistance (LR), flower colour (FC), number of heads per plant (HN), leaf margin with thorns or without thorns (LSP), and leaf margin (LM). The data were recorded for five healthy plants of each accession. 4.2. DNA Extraction and Indel Marker Development The best growing plant from the five plants with good consistency was selected as the sample plant, and the leaves were collected and stored at −20 °C. Genomic DNA was extracted using a modified version of the cetyltrimethylammonium bromide (CTAB) method . The quality and quantity of the extracted genomic DNA were determined by agarose gel electrophoresis and spectrophotometry. The DNA concentration was diluted to 50 ng/μL, and the diluted DNA was placed and stored at −20 °C for subsequent studies. Indel markers were developed based on the safflower whole genome and transcriptome data . The specific methods were as follows: (1) Illumina sequencing of multiple safflower accessions, (2) removal of low-quality sequences by using Trimmomatic version 0.35 software (3) alignment of processed transcriptome sequences to safflower genome sequences by utilizing Burrows–Wheeler Aligner (BWA) version 0.7.17 software, (4) use of GATK software to find insertion/deletion sites (Indel) and retain only Indel sites ≥ 20 bp in length for subsequent detection by agarose gel electrophoresis, and (5) use of Primer version 5.0 software to design primers based on flanking sequences of Indel sites with the following default parameters: primer length 18–25 bp, annealing temperature 57–63 °C, GC content 50–60%, and product length 100–300 bp . 4.3. Analysis of Genetic Diversity in Safflower Germplasm Five core parameters, including number of different alleles ( N A ), number of effective alleles ( N E ), Shannon’s coefficient, polymorphism information content ( PIC ), and expected heterozygosity ( H e ) were evaluated using GenAlEx version 6.51 and POPGENE version 1.3.2 software . The Nei 1983 Distance method, neighbour-joining (NJ) tree method, and bootstrap number were selected by POWERMAKER version 3.25 and set to 1000 times to generate a marker-based tree file . Based on the UPGMA algorithm on Past 3 software, the similarity coefficient was selected as Bray–Curtis, and the bootstrap number was set to 1000 times to generate a phenotype-based tree file . An evolutionary tree was prepared by iTOL webtool version 5. Core Hunter version 3.0 was employed to screen safflower core germplasms by using marker-based data and phenotypic data as input data formats . Gower’s distance was utilised to calculate distances to phenotypic traits. Modified Roger’s distance and Cavalli–Sforza and Edwards distance were supported for genetic marker data. Average entry-to-nearest-entry distance was selected as the evaluation measure for core germplasm screening. Kinship coefficients between individuals were calculated using TASSEL version 3.0 software, based on Indel marker type data . The General linear model (GLM) and Mixed linear model (MLM) included in the software were used to combine the phenotypic data to perform the association analysis. In the GLM analysis, the Q value was applied as a covariate for the analysis, and the marker-phenotype association loci were identified by regression analysis using phenotype data. In the MLM analysis, the kinship K value was added as a covariate for association analysis to identify marker-phenotype association loci. 4.4. Metabolomic and Transcriptomic Analyses Three different colours of safflower material, red (R), white (W), and yellow (Y), were used for metabolome and transcriptome sampling . When the three different coloured safflowers were in full bloom, the sampled flowers were quickly placed into a cryogenic liquid nitrogen tank for temporary storage and then transported back to the laboratory, where they were stored at −80 °C for freezing. Safflower inflorescences of the same colour were mixed to form one sample, and three replicates were taken per 100 mg. Half of each replicate was used for RNA sequencing, and the other half was used for metabolome sequencing. The extracted samples of safflower were separated on the UHPLC system (ExionLC™ AD) equipped with Waters ACQUITY ULC HSS T3 (1.8 μm, 2.1 mm × 100 mm). The analysis conditions were as follows: column temperature, 35 °C; injection volume, 2 μL; and flow rate, 0.3 mL/min. The mobile phases were water (0.1% formic acid) (phase A) and acetonitrile (0.1% formic acid) (phase B). The gradient program of phase A/phase B was 98:2 ( v / v ) at 0 min, 98:2 ( v / v ) at 2 min, 2:98 ( v / v ) at 11 min, 2:98 ( v / v ) at 13 min, and 98:2 ( v / v ) at 15 min. Samples were inserted into quality control (QC) samples in queue mode to monitor and evaluate the stability of the system and the reliability of the experimental data. Mass spectrometry was carried out using the Thermo high resolution mass spectrometry (Thermo QE Focus, Shanghai, China) and data were collected using IDA (information-dependent acquisition) mode. The ion source conditions were as follows: Spray Voltage: +3500/−3500 V, Capillary Temperature: 350 °C, Sheath Gas: 30, Aux Gas: 10, CE: 10, 30, and 50. The raw data were adjusted for peak alignment, peak area extraction, retention time correction, and feature extraction using the XCMS version 3.6.3 software. MRM data were processed using Skyline version 21.1.0.146 software. The principal components analysis (PCA), orthogonal partial least squares discriminant analysis (OPLS-DA) models, and K-means clustering were used to distinguish the significant differentially accumulated metabolites (DAMs) in different colour groups. Three groups of safflower tubular flower samples with R, W, and Y colours and biological replicates for each group were collected and analysed using transcriptome sequencing. Total RNA was isolated from frozen flowers by using a Quick RNA isolation Kit (Huayueyang, Beijing, China) according to the instructions for preparing for sequencing. The library preparations were sequenced on an ANORODA Next Seq 550AR platform and 150 bp paired-end reads were generated. The raw paired-end reads were cleaned through removing adaptor sequences, poly-N, and low-quality sequences. Clean reads after quality control were compared to the reference genome by HISAT2 to obtain position information on the reference genome or gene as well as specific sequence characteristic information of the sequenced samples . The genes were quantified with FPKMs using StringTie, and a differential expression analysis of three groups was performed using the DESeq2 R package . The threshold p -value in multiple tests to judge the significance of the gene expression difference was based on the false discovery rate (FDR) method. When FDR ≤ 0.05 and FPKM values showed at least a 2-fold difference among samples, the gene was considered a significant DEG. 4.5. Determination of Anthocyanin and Carotenoid Content In order to quantify the anthocyanin and carotenoid content, the three groups of safflowers tubular flower samples with R, O, Y, and W colours were rapidly ground to powder under liquid nitrogen. Afterward, the dried samples were stored at a low temperature for subsequent experiments. The optimal conditions used to extract anthocyanins were as follows: 60% ethanol (pH = 3), soil–liquid ratio of 1:15 (g/mL), ultrasonic time of 120 min, and ultrasonic temperature of 40 °C. A potassium chloride (KCl) buffer solution with pH = 1.0 and a sodium acetate buffer solution with pH = 4.5 were used for the dilution of each extracted anthocyanin samples from the three groups of safflowers separately. The absorbance of anthocyanins was measured at 530 nm using a spectrophotometer and three replicate samples were performed for each group . The anthocyanin yield rate in tubular flowers from the three group samples was calculated by the following Formulas (1) and (2): A = (A520 nm − A700 nm) pH1.0 − (A520 nm − A700 nm) pH4.5 (1) Anthocyanin (mg/g) = A × MW × DF × 10 3 /(ε × ℓ) (2) A: Absorbance; ε = 4.62 × 10 6 (Anthocyanin Extinction Coefficient); DF: Dilution factor. Carotenoids were extracted from the three group samples using a petroleum ether–acetone solution = 5:1 ( v / v ) with a 40 °C-water bath for 2 h. The carotenoid absorbance was measured at 440 nm and the carotenoid content was calculated by the following Formula (3) : (3) Carotenoids content ( μ g / g ) = A × V ( m L ) × 10 4 A 1 cm 1 % × P ( g ) A : Absorbance; V : Total extract volume; P : sample weight; and A 1 cm 1 % = 2592 (β-carotene Extinction Coefficient in petroleum ether). 4.6. Statistical Analysis 614 safflower germplasms were used as samples, ten plants of each variety were planted, and the average agronomic traits of five plants of each variety were taken as one sample. The mean, maximum, and minimum of agronomic traits, as well as their coefficients of variation and correlation analysis (computed as Pearson’s correlation coefficient r), were analysed using SPSS 25. The formula for calculating the coefficient of variation is as follows: CV = (SD/MN) × 100, where SD is the standard deviation and MN is the mean. The range is calculated as Max value–Min value. Principal component analysis (PCA) was performed using the FactoMineR and factoextra R packages . F-statistics, including Fst, hierarchical AMOVA, and the pairwise Fst were conducted using GenAlEx version 6.502. GraphPad Prism version 10 was utilised for additional data visualisations and statistical analyses. A total of 614 safflower materials were provided by the germplasm repository of the Wuhan Institute of Oilseed Crops, Chinese Academy of Agricultural Sciences. The safflower material was planted in an experimental field at 58 m above sea level in Huangpi District, Wuhan City, Hubei Province, China (30°45′59″ N, 114°27′46″ E). Each safflower variety was sown in a row 4 m long, 50 cm apart, with an average plant spacing of 40 cm. The total number of seedlings left in each row after interplanting was 10. Agronomic traits were investigated following the guidelines provided by the Oils Crops Research Institute of the Chinese Academy of Agricultural Sciences for safflower . These guidelines encompass the following: plant height (PH), primary branch height (PBH), terminal branch height (TBH), number of primary branches (PBNs), number of secondary branches (SBNs), seed germination rate (SGR), load resistance (LR), flower colour (FC), number of heads per plant (HN), leaf margin with thorns or without thorns (LSP), and leaf margin (LM). The data were recorded for five healthy plants of each accession. The best growing plant from the five plants with good consistency was selected as the sample plant, and the leaves were collected and stored at −20 °C. Genomic DNA was extracted using a modified version of the cetyltrimethylammonium bromide (CTAB) method . The quality and quantity of the extracted genomic DNA were determined by agarose gel electrophoresis and spectrophotometry. The DNA concentration was diluted to 50 ng/μL, and the diluted DNA was placed and stored at −20 °C for subsequent studies. Indel markers were developed based on the safflower whole genome and transcriptome data . The specific methods were as follows: (1) Illumina sequencing of multiple safflower accessions, (2) removal of low-quality sequences by using Trimmomatic version 0.35 software (3) alignment of processed transcriptome sequences to safflower genome sequences by utilizing Burrows–Wheeler Aligner (BWA) version 0.7.17 software, (4) use of GATK software to find insertion/deletion sites (Indel) and retain only Indel sites ≥ 20 bp in length for subsequent detection by agarose gel electrophoresis, and (5) use of Primer version 5.0 software to design primers based on flanking sequences of Indel sites with the following default parameters: primer length 18–25 bp, annealing temperature 57–63 °C, GC content 50–60%, and product length 100–300 bp . Five core parameters, including number of different alleles ( N A ), number of effective alleles ( N E ), Shannon’s coefficient, polymorphism information content ( PIC ), and expected heterozygosity ( H e ) were evaluated using GenAlEx version 6.51 and POPGENE version 1.3.2 software . The Nei 1983 Distance method, neighbour-joining (NJ) tree method, and bootstrap number were selected by POWERMAKER version 3.25 and set to 1000 times to generate a marker-based tree file . Based on the UPGMA algorithm on Past 3 software, the similarity coefficient was selected as Bray–Curtis, and the bootstrap number was set to 1000 times to generate a phenotype-based tree file . An evolutionary tree was prepared by iTOL webtool version 5. Core Hunter version 3.0 was employed to screen safflower core germplasms by using marker-based data and phenotypic data as input data formats . Gower’s distance was utilised to calculate distances to phenotypic traits. Modified Roger’s distance and Cavalli–Sforza and Edwards distance were supported for genetic marker data. Average entry-to-nearest-entry distance was selected as the evaluation measure for core germplasm screening. Kinship coefficients between individuals were calculated using TASSEL version 3.0 software, based on Indel marker type data . The General linear model (GLM) and Mixed linear model (MLM) included in the software were used to combine the phenotypic data to perform the association analysis. In the GLM analysis, the Q value was applied as a covariate for the analysis, and the marker-phenotype association loci were identified by regression analysis using phenotype data. In the MLM analysis, the kinship K value was added as a covariate for association analysis to identify marker-phenotype association loci. Three different colours of safflower material, red (R), white (W), and yellow (Y), were used for metabolome and transcriptome sampling . When the three different coloured safflowers were in full bloom, the sampled flowers were quickly placed into a cryogenic liquid nitrogen tank for temporary storage and then transported back to the laboratory, where they were stored at −80 °C for freezing. Safflower inflorescences of the same colour were mixed to form one sample, and three replicates were taken per 100 mg. Half of each replicate was used for RNA sequencing, and the other half was used for metabolome sequencing. The extracted samples of safflower were separated on the UHPLC system (ExionLC™ AD) equipped with Waters ACQUITY ULC HSS T3 (1.8 μm, 2.1 mm × 100 mm). The analysis conditions were as follows: column temperature, 35 °C; injection volume, 2 μL; and flow rate, 0.3 mL/min. The mobile phases were water (0.1% formic acid) (phase A) and acetonitrile (0.1% formic acid) (phase B). The gradient program of phase A/phase B was 98:2 ( v / v ) at 0 min, 98:2 ( v / v ) at 2 min, 2:98 ( v / v ) at 11 min, 2:98 ( v / v ) at 13 min, and 98:2 ( v / v ) at 15 min. Samples were inserted into quality control (QC) samples in queue mode to monitor and evaluate the stability of the system and the reliability of the experimental data. Mass spectrometry was carried out using the Thermo high resolution mass spectrometry (Thermo QE Focus, Shanghai, China) and data were collected using IDA (information-dependent acquisition) mode. The ion source conditions were as follows: Spray Voltage: +3500/−3500 V, Capillary Temperature: 350 °C, Sheath Gas: 30, Aux Gas: 10, CE: 10, 30, and 50. The raw data were adjusted for peak alignment, peak area extraction, retention time correction, and feature extraction using the XCMS version 3.6.3 software. MRM data were processed using Skyline version 21.1.0.146 software. The principal components analysis (PCA), orthogonal partial least squares discriminant analysis (OPLS-DA) models, and K-means clustering were used to distinguish the significant differentially accumulated metabolites (DAMs) in different colour groups. Three groups of safflower tubular flower samples with R, W, and Y colours and biological replicates for each group were collected and analysed using transcriptome sequencing. Total RNA was isolated from frozen flowers by using a Quick RNA isolation Kit (Huayueyang, Beijing, China) according to the instructions for preparing for sequencing. The library preparations were sequenced on an ANORODA Next Seq 550AR platform and 150 bp paired-end reads were generated. The raw paired-end reads were cleaned through removing adaptor sequences, poly-N, and low-quality sequences. Clean reads after quality control were compared to the reference genome by HISAT2 to obtain position information on the reference genome or gene as well as specific sequence characteristic information of the sequenced samples . The genes were quantified with FPKMs using StringTie, and a differential expression analysis of three groups was performed using the DESeq2 R package . The threshold p -value in multiple tests to judge the significance of the gene expression difference was based on the false discovery rate (FDR) method. When FDR ≤ 0.05 and FPKM values showed at least a 2-fold difference among samples, the gene was considered a significant DEG. In order to quantify the anthocyanin and carotenoid content, the three groups of safflowers tubular flower samples with R, O, Y, and W colours were rapidly ground to powder under liquid nitrogen. Afterward, the dried samples were stored at a low temperature for subsequent experiments. The optimal conditions used to extract anthocyanins were as follows: 60% ethanol (pH = 3), soil–liquid ratio of 1:15 (g/mL), ultrasonic time of 120 min, and ultrasonic temperature of 40 °C. A potassium chloride (KCl) buffer solution with pH = 1.0 and a sodium acetate buffer solution with pH = 4.5 were used for the dilution of each extracted anthocyanin samples from the three groups of safflowers separately. The absorbance of anthocyanins was measured at 530 nm using a spectrophotometer and three replicate samples were performed for each group . The anthocyanin yield rate in tubular flowers from the three group samples was calculated by the following Formulas (1) and (2): A = (A520 nm − A700 nm) pH1.0 − (A520 nm − A700 nm) pH4.5 (1) Anthocyanin (mg/g) = A × MW × DF × 10 3 /(ε × ℓ) (2) A: Absorbance; ε = 4.62 × 10 6 (Anthocyanin Extinction Coefficient); DF: Dilution factor. Carotenoids were extracted from the three group samples using a petroleum ether–acetone solution = 5:1 ( v / v ) with a 40 °C-water bath for 2 h. The carotenoid absorbance was measured at 440 nm and the carotenoid content was calculated by the following Formula (3) : (3) Carotenoids content ( μ g / g ) = A × V ( m L ) × 10 4 A 1 cm 1 % × P ( g ) A : Absorbance; V : Total extract volume; P : sample weight; and A 1 cm 1 % = 2592 (β-carotene Extinction Coefficient in petroleum ether). 614 safflower germplasms were used as samples, ten plants of each variety were planted, and the average agronomic traits of five plants of each variety were taken as one sample. The mean, maximum, and minimum of agronomic traits, as well as their coefficients of variation and correlation analysis (computed as Pearson’s correlation coefficient r), were analysed using SPSS 25. The formula for calculating the coefficient of variation is as follows: CV = (SD/MN) × 100, where SD is the standard deviation and MN is the mean. The range is calculated as Max value–Min value. Principal component analysis (PCA) was performed using the FactoMineR and factoextra R packages . F-statistics, including Fst, hierarchical AMOVA, and the pairwise Fst were conducted using GenAlEx version 6.502. GraphPad Prism version 10 was utilised for additional data visualisations and statistical analyses. In this study, 214 safflower core germplasms were constructed based on 11 agronomic traits and Indel markers. The association analysis of the core germplasm population identified five loci associated with flower colour. The combined metabolomics and transcriptomics analyses showed that there are differences in flavonoid metabolic pathways and differences in anthocyanin and carotenoid contents in safflowers with different flower colours, which contribute to the richness of flower colours. These findings provide a basis for understanding the molecular mechanisms underlying safflower flower colour diversity. |
Complete genomic profiles of 1496 Taiwanese reveal curated medical insights | d1b04eea-d358-433f-a19c-0ddec31e95b6 | 11675050 | Pharmacology[mh] | As the field of precision medicine continues to expand, many population-based consortiums have been collecting genomic information for biomedical studies . In Asia, a genetic reference panel based on 3,552 individuals of the Japanese (JPN) population was published in 2019 , . Another study based on 4,810 Singaporeans (SGN) was published in 2019 by the SG10K project . The results showed that the cohort analysis of a sub-sample was able to capture the population diversity. Even though the ChinaMAP project discloses the genetic profile of 10,588 Chinese , Asian genomics data are still relatively underrepresented in the public databases relative to global population percentage. The Taiwan Biobank project (TWB) was established to facilitate biomedical research on the genetic basis of the Taiwanese, a multicultural population. The majority of Taiwanese immigrated from various provinces of China over the past centuries, and there is also a group of Taiwanese aboriginals. As of 30 Sep 2021, the TWB has already recruited 153,543 subjects from the general population. Many types of genomic data are available to users, including single nucleotide variant (SNV) array data from 114,604 subjects, whole-genome sequencing (WGS) data from 2,010 subjects, and human leukocyte antigen (HLA) typing data from 1,102 subjects. Although many genotyping approaches can reveal specific variants that indicate disease susceptibility , , fragmented variant information is not sufficient to fully infer the haplotype of a gene. WGS data have the potential to reconstruct haplotype information at high resolution and cover entire genomic regions, allowing complete determination of disease-causing variants, including SNVs, small insertions/deletions (INDEL), structure variants, and the haplotypes of a gene. For example, a previous study suggested that more than 95 % of hemoglobin subunit alpha ( HBA1/2 ) pathogenic or likely pathogenic variants are from –SEA, -a 3 . 7 , -a 4 . 2 in the Chinese population . Most of these are deletions larger than 3 kb in size. The pathogenic variants of fragile X-linked mental retardation gene FMRP translational regulator 1 ( FMR1 ) are typically caused by trinucleotide (CGG) repeat expansions, variations of which require specific detection algorithms , , . A medically meaningful complete genomic profile of an individual should also include monogenic disease-causing allele screening and the haplotypes of the human leukocyte antigen (HLA) gene. Like the HLA gene, the Cytochrome P450 2D6 ( CYP2D6 ) gene is associated with the metabolism of many drugs. For functional interpretation, complete haplotype information is needed for both HLA and CPY2D6 genes. Identifying at-risk individuals with medically actionable information from hereditary diseases may benefit the individual’s whole family. Thus, expanded carrier screening for monogenic diseases has become a discipline-wide goal after a joint statement was published by the American College of Medical Genetics and Genomics (ACMG), American College of Obstetricians and Gynecologists (ACOG), National Society of Genetic Counselors (NSGC), Perinatal Quality Foundation (PQF), and Society for Maternal-Fetal Medicine (SMFM) in 2015 . All the disorders noted in abovementioned joint statement involve a cognitive or physical disability, the need for postnatal surgical or medical intervention, or a detrimental effect on the quality of life; most importantly, they are disorders where prenatal intervention could significantly improve perinatal outcomes and delivery management or where prenatal education could meet the unique needs after birth. Each year, the ACMG Secondary Findings Maintenance Working Group (SFWG) evaluates these actionable genes and associated conditions for their actionability, severity, penetrance, and impact or burden of available treatment modalities or screening recommendations. The latest version, ACMG SF v3.0 published in May 2021, contains 73 actionable genes, in contrast to 59 genes in the previous version. Many studies have disclosed the profile of actionable genes in the Asia-Pacific region , , , and WGS data allow re-analysis when the gene list has been updated. The advent of WGS helps characterize population genetic structures that can provide useful information to the public healthcare system, mitigating costs through disease risk prediction, diagnosis, and treatments. In this study, we reanalyzed the WGS data from 1496 participants present in the TWB to provide a medically useful cross-section of the population’s complete genomic profile. We also discuss disease carrier status (including all the conditions curated in the ClinVar database), ACMG actionable genes, drug responses from the PharmGKB database, and clinically relevant variants with high minor allele frequency (MAF) in the Taiwanese population. To facilitate the exploration of the variants described here, we designed a web interface for the constructed database TaiwanGenomes ( https://genomes.tw ) that allows easy browsing of the list of variants with different combinations of filters. Ethics statement All experiments involving human subjects were in accordance with the institutional review board approval from Biomedical Science Research of Academia Sinica, Taiwan (IRB-BM) and the Ethics and Governance Council (EGC) of Taiwan Biobank, Taiwan. All the informed consents were obtained and gave ethical approval for this work (AS-IRB01-18041(N)). All raw sequence data used in this study were generated as part of the Taiwan Biobank project. This study did not reveal any individual participant’s information, and none of the results can be used to identify individual participants. Data source The Taiwan Biobank (TWB) collected and recruited participants across different cities throughout Taiwan, enrolling participants from the general population. TWB has been enrolling adult volunteers between the ages of 20 and 70 who do not have the diagnosed cancer. For the subset undergoing WGS, participants were selected to match the population distribution as well as to ensure the gender balance. A total of de-identified,496 Taiwanese WGS data sets were sourced from TWB with ethical approval (AS-IRB01-18041(N)). WGS was conducted in three batches (n = 496, 497, and 503) with high-depth mapped reads. The samples were sequenced by next-generation sequencing (NGS) platforms: Illumina HiSeq 2500, 4000, and Novaseq systems. Sequencing of DNA extracted from each blood sample followed by illumina TruSeq DNA PCR-Free HT library preparation kit to generate about 90 GB of data with an average coverage of 30x ( https://taiwanview.twbiobank.org.tw/about.php ). All pairwise kinships have been checked, and there is no parent-offspring relationship in the cohort. Genotype calling, validation, and variant annotation Variant detection and joint genotype calling analyses were conducted based on the Sentieon DNAscope pipeline (Sentieon Inc., version 201808 ), a precisionFDA challenge winner implementation of GATK’s best practice . The sequence reads of each sample in FASTQ format were aligned against the human reference genome (GRCh37/ucsc.hg19.fasta) using BWA-MEM (Burrows-Wheeler Aligner with Maximal Exact Match algorithm, version 0.7.15-r1140 ). The alignment file was sorted using SAMtools , and the Sentieon Dedup algorithm was used to mark duplicated reads. The Sentieon Realigner algorithm reinforced local realignment around each indel region, and the Sentieon QualCal algorithm recalibrated base quality scores. SNVs and indels were called in genomic variant call format (GVCF) using Haplotyper. The Sentieon GVCFtyper jointly called 1496 subjects as a cohort, then Sentieon VarCal plus ApplyVarCal, a machine learning-based variant quality sequence recalibration (VQSR) algorithm for variant refinement, were used. The training sets for VQSR were those suggested by GATK’s best practice. For SNV, the datasets from 1000G phase1, omni2.5, dbSNP version 138, and HapMap version 3.3 were included as the training sets. For INDEL, the datasets from the 1000G phase1, dbSNP version 138, and Mills-and-1000G gold standard were used as the training sets. VQSR included a list of sequence-level annotations such as QD, MQ, MQRankSum, ReadPosRankSum, and FS. Multi-allelic variants were normalized into multiple bi-allelic variants at the same position through decomposition and left-aligned using bcftools (v1.9) , . We repeated the same pipeline with an additional seven Genome in A Bottle (GIAB) , samples (HG001 ∼ HG007) and jointly called with 1496 TWB subjects for benchmarking the variant calling pipeline. We stratified the variant call sets according to the VQSR tranches into strata of 100.0, 99.9, 99.8, 99.7, 99.6, 99.5, and 99.0. We further adopted the call rate, allele number (AN), and depth of coverage (DP) criteria for classifying all reference alleles into three quality classifications. Moreover, we randomly validated 149 variants in four samples (NGS2 20150510B, NGS2 20150510C, NGS2 20150510D, NGS2 20150510F) by Sanger sequencing. We were particularly interested in the variants absent, or only present, after joint calling. Due to the limited amount of available DNA, we separated samples into two groups (BC and DF) for cross-validation. We utilized the Sanger sequencing results to calculate the estimated false discovery rate (FDR) for determining the VQSR tranche. In order to provide a clear threshold with higher precision without significantly compromising recall rates, we selected VQSR tranche 99.7 as the PASS criterion based on the comparisons of HG001 (NA12878) true genetic variants and following Sanger sequencing validation results . All identified variants can be accessed from the database TaiwanGenomes ( https://genomes.tw ) together with the corresponded VQSR tranche and annotated information. Only variants with pass quality were considered in the downstream analysis. Variant annotation was done by ANNOVAR (version 20180416) with updated databases, including RefSeq Gene, UCSC Known Gene, ClinVar (v20210501), avsnp150, NHLBI-ESP 6500 exome, 1000 genome, The Genome Aggregation Database (gnomAD) genome and exome (v2.1.1), The Exome Aggregation Consortium (ExAC) 65,000 exome, Kaviar, cg69, dbnsfp33a, dbscsnv11, gwava, tfbsConsSites, wgRna, and targetScanS. We selected the annotated variants on the autosome and sex chromosome as the final variant call set. We defined nonsynonymous variants as those where a variant’s annotation hit exonic or splicing regions. The variations were defined by nonframeshift deletion, frameshift deletion, nonframeshift insertion, frameshift insertion, stopgain, stoploss, or a nonsynonymous SNV with an exonic variant function. To further elucidate the possible role of structural variant calling tools, we performed a trial using Manta (version 1.6.0) and AnnotSV (version 3.0.4) to identify and annotate the structural variants (SV) in HBA1/2 genes of a subset of the participants (n = 494). Furthermore, the SMN1/SMN2 copy number was analyzed by SMNCopyNumberCaller (version 1.1.1). Variant filtering and classification We defined the variants absent in any other public databases (e.g., without allele frequency or effect annotation) as “globally novel variants.” Variants that were not found in any previous samples while sequentially examining the 1496 samples we defined as “population novel variants”. For functional effect analysis, variants annotated as exonic in refGene were defined as “coding region variants,” whereas the variants annotated as nonsense, splicing, or frameshift in refGene were defined as “loss of function (LOF) variants.” We created in-house scripts to conduct the filtering processes. Public users can retrieve the same information by setting corresponding filters in TaiwanGenomes ( https://genomes.tw ). Clinical important pharmacogenomic alleles The HLA gene plays a crucial role in personalized medicine. It is associated with many adverse drug events, including antithyroid drug-induced agranulocytosis . The clinical variant data in the PharmGKB database (clinicalVariants.tsv, version 20210405) detail how genetic variation influences the drug response, including a list of variant-drug pairs with different levels of evidence. TWB has released individual-level HLA alleles by using NXType assay kits ( https://taiwanview.twbiobank.org.tw/about.php ). To evaluate the allele frequencies of PharmGKB-reported HLA variant-drug pairs in the Taiwanese population, we queried HLA allele typing results from the TWB websites. Here, we annotated the alleles with evidence levels 1A/1B/2A and disclosed their respective frequencies. For CYP genes, we genotyped CYP haplotypes based on 1,017 WGS data sets using ALDY and STARGAZER (a subset of our participants; manuscript in preparation). In calculating CYP variant-drug pairs, we only reported the allele frequency of the variants associated with abnormal drug metabolism. We did not include CYP wild-type allele frequencies. Carrier status – Cohort MAF and expert reviews The ClinVar database collects the interpretation of clinically related variants submitted by global researchers. Many known monogenic disease-causing variants are rare and vary across populations . However, only a few records in the database are from the Asian population. To explore medically relevant genetic variants in Taiwanese people, we retained variants annotated as pathogenic, likely pathogenic, or pathogenic/likely pathogenic (P/LP) in ClinVar (v20210501). A high-variant minor-frequency (MAF) allele ( > 0.5 %) with known pathogenicity may indicate the specificity of the population’s genetic architecture. By comparing across the East Asian population (EAS), we singled out pathogenic or likely-pathogenic variants with high frequency to address the specific genetic structure of the Taiwanese. The variant MAF was further investigated by comparing it with allelic frequencies in the gnomAD (version 2.1.1) and ExAC database. The following filtration criteria were used: Category 1) The minor allele frequency of the variant in TWB1496 was ≥ 0.01, while that in the ExAC database and in the East Asia population of the gnomAD genome and exome database were ≤ 0.01 or absent; Category 2) the allelic frequency was 0.005–0.01 in TWB1496, and ≤ 0.005 or absent in the ExAC database and in the East Asia population of the gnomeAD genome and exome database. The genetic inheritance mode of the variants was manually annotated using the Online Mendelian Inheritance in Man database (OMIM). On the other hand, experts further reviewed the variants within ACMG actionable genes. The concept of medically actionable genes originates from ACMG recommendations on reporting secondary findings from exome or genome sequencing results. To further focus on the East Asian population, we also compared the allele frequency of variants in the carrier sets to two East Asian populations (Singaporean and Japanese). The Singaporean data were acquired from the SG10K project 4 , and the Japanese data were obtained from the 3.5KJPN project . To evaluate the clinical impact of the carrier call set, we applied the expanded carrier panel , to the above data to calculate accumulation frequencies of pathogenic and likely pathogenic variants in monogenic disease-causing genes. We compared these data with the United States cohort (sequencing method) and Taiwan domestic data 6 (array genotyping method). The overall study design and analysis workflow is shown in . In summary, for SNV and INDEL, the analyses were performed for the entire genome of the cohort (N = 1496). For HLA allele, the data was from Taiwan Biobank official release ( https://taiwanview.twbiobank.org.tw/hla.php ) (N = 1,103), and allele annotation was based on PharmGKB information. For the CYP allele, the analyses were performed on a subset of the WGS cohort (N = 1,017) and only focused on CYP genes. The SV analyses were performed for a subset of the WGS cohort (N = 494), and the analyses were for the entire genome. For SMN CNV calling, the analyses were based on a subset of the WGS cohort (N = 494), and only for SMN genes. We also used the same carrier call-set data to evaluate other monogenic disease-causing genes as a concept of expanded carrier screening. We compared the P/LP variant carrier frequency with the U.S. cohort and Taiwanese cohort (using TWBv2.0 array genotyping). We also used the Westemeyer et al. (2020) gene list (274 genes) as a virtual panel to estimate how many couples may benefit from the expanded carrier screening in the Taiwanese population. All experiments involving human subjects were in accordance with the institutional review board approval from Biomedical Science Research of Academia Sinica, Taiwan (IRB-BM) and the Ethics and Governance Council (EGC) of Taiwan Biobank, Taiwan. All the informed consents were obtained and gave ethical approval for this work (AS-IRB01-18041(N)). All raw sequence data used in this study were generated as part of the Taiwan Biobank project. This study did not reveal any individual participant’s information, and none of the results can be used to identify individual participants. The Taiwan Biobank (TWB) collected and recruited participants across different cities throughout Taiwan, enrolling participants from the general population. TWB has been enrolling adult volunteers between the ages of 20 and 70 who do not have the diagnosed cancer. For the subset undergoing WGS, participants were selected to match the population distribution as well as to ensure the gender balance. A total of de-identified,496 Taiwanese WGS data sets were sourced from TWB with ethical approval (AS-IRB01-18041(N)). WGS was conducted in three batches (n = 496, 497, and 503) with high-depth mapped reads. The samples were sequenced by next-generation sequencing (NGS) platforms: Illumina HiSeq 2500, 4000, and Novaseq systems. Sequencing of DNA extracted from each blood sample followed by illumina TruSeq DNA PCR-Free HT library preparation kit to generate about 90 GB of data with an average coverage of 30x ( https://taiwanview.twbiobank.org.tw/about.php ). All pairwise kinships have been checked, and there is no parent-offspring relationship in the cohort. Variant detection and joint genotype calling analyses were conducted based on the Sentieon DNAscope pipeline (Sentieon Inc., version 201808 ), a precisionFDA challenge winner implementation of GATK’s best practice . The sequence reads of each sample in FASTQ format were aligned against the human reference genome (GRCh37/ucsc.hg19.fasta) using BWA-MEM (Burrows-Wheeler Aligner with Maximal Exact Match algorithm, version 0.7.15-r1140 ). The alignment file was sorted using SAMtools , and the Sentieon Dedup algorithm was used to mark duplicated reads. The Sentieon Realigner algorithm reinforced local realignment around each indel region, and the Sentieon QualCal algorithm recalibrated base quality scores. SNVs and indels were called in genomic variant call format (GVCF) using Haplotyper. The Sentieon GVCFtyper jointly called 1496 subjects as a cohort, then Sentieon VarCal plus ApplyVarCal, a machine learning-based variant quality sequence recalibration (VQSR) algorithm for variant refinement, were used. The training sets for VQSR were those suggested by GATK’s best practice. For SNV, the datasets from 1000G phase1, omni2.5, dbSNP version 138, and HapMap version 3.3 were included as the training sets. For INDEL, the datasets from the 1000G phase1, dbSNP version 138, and Mills-and-1000G gold standard were used as the training sets. VQSR included a list of sequence-level annotations such as QD, MQ, MQRankSum, ReadPosRankSum, and FS. Multi-allelic variants were normalized into multiple bi-allelic variants at the same position through decomposition and left-aligned using bcftools (v1.9) , . We repeated the same pipeline with an additional seven Genome in A Bottle (GIAB) , samples (HG001 ∼ HG007) and jointly called with 1496 TWB subjects for benchmarking the variant calling pipeline. We stratified the variant call sets according to the VQSR tranches into strata of 100.0, 99.9, 99.8, 99.7, 99.6, 99.5, and 99.0. We further adopted the call rate, allele number (AN), and depth of coverage (DP) criteria for classifying all reference alleles into three quality classifications. Moreover, we randomly validated 149 variants in four samples (NGS2 20150510B, NGS2 20150510C, NGS2 20150510D, NGS2 20150510F) by Sanger sequencing. We were particularly interested in the variants absent, or only present, after joint calling. Due to the limited amount of available DNA, we separated samples into two groups (BC and DF) for cross-validation. We utilized the Sanger sequencing results to calculate the estimated false discovery rate (FDR) for determining the VQSR tranche. In order to provide a clear threshold with higher precision without significantly compromising recall rates, we selected VQSR tranche 99.7 as the PASS criterion based on the comparisons of HG001 (NA12878) true genetic variants and following Sanger sequencing validation results . All identified variants can be accessed from the database TaiwanGenomes ( https://genomes.tw ) together with the corresponded VQSR tranche and annotated information. Only variants with pass quality were considered in the downstream analysis. Variant annotation was done by ANNOVAR (version 20180416) with updated databases, including RefSeq Gene, UCSC Known Gene, ClinVar (v20210501), avsnp150, NHLBI-ESP 6500 exome, 1000 genome, The Genome Aggregation Database (gnomAD) genome and exome (v2.1.1), The Exome Aggregation Consortium (ExAC) 65,000 exome, Kaviar, cg69, dbnsfp33a, dbscsnv11, gwava, tfbsConsSites, wgRna, and targetScanS. We selected the annotated variants on the autosome and sex chromosome as the final variant call set. We defined nonsynonymous variants as those where a variant’s annotation hit exonic or splicing regions. The variations were defined by nonframeshift deletion, frameshift deletion, nonframeshift insertion, frameshift insertion, stopgain, stoploss, or a nonsynonymous SNV with an exonic variant function. To further elucidate the possible role of structural variant calling tools, we performed a trial using Manta (version 1.6.0) and AnnotSV (version 3.0.4) to identify and annotate the structural variants (SV) in HBA1/2 genes of a subset of the participants (n = 494). Furthermore, the SMN1/SMN2 copy number was analyzed by SMNCopyNumberCaller (version 1.1.1). We defined the variants absent in any other public databases (e.g., without allele frequency or effect annotation) as “globally novel variants.” Variants that were not found in any previous samples while sequentially examining the 1496 samples we defined as “population novel variants”. For functional effect analysis, variants annotated as exonic in refGene were defined as “coding region variants,” whereas the variants annotated as nonsense, splicing, or frameshift in refGene were defined as “loss of function (LOF) variants.” We created in-house scripts to conduct the filtering processes. Public users can retrieve the same information by setting corresponding filters in TaiwanGenomes ( https://genomes.tw ). The HLA gene plays a crucial role in personalized medicine. It is associated with many adverse drug events, including antithyroid drug-induced agranulocytosis . The clinical variant data in the PharmGKB database (clinicalVariants.tsv, version 20210405) detail how genetic variation influences the drug response, including a list of variant-drug pairs with different levels of evidence. TWB has released individual-level HLA alleles by using NXType assay kits ( https://taiwanview.twbiobank.org.tw/about.php ). To evaluate the allele frequencies of PharmGKB-reported HLA variant-drug pairs in the Taiwanese population, we queried HLA allele typing results from the TWB websites. Here, we annotated the alleles with evidence levels 1A/1B/2A and disclosed their respective frequencies. For CYP genes, we genotyped CYP haplotypes based on 1,017 WGS data sets using ALDY and STARGAZER (a subset of our participants; manuscript in preparation). In calculating CYP variant-drug pairs, we only reported the allele frequency of the variants associated with abnormal drug metabolism. We did not include CYP wild-type allele frequencies. The ClinVar database collects the interpretation of clinically related variants submitted by global researchers. Many known monogenic disease-causing variants are rare and vary across populations . However, only a few records in the database are from the Asian population. To explore medically relevant genetic variants in Taiwanese people, we retained variants annotated as pathogenic, likely pathogenic, or pathogenic/likely pathogenic (P/LP) in ClinVar (v20210501). A high-variant minor-frequency (MAF) allele ( > 0.5 %) with known pathogenicity may indicate the specificity of the population’s genetic architecture. By comparing across the East Asian population (EAS), we singled out pathogenic or likely-pathogenic variants with high frequency to address the specific genetic structure of the Taiwanese. The variant MAF was further investigated by comparing it with allelic frequencies in the gnomAD (version 2.1.1) and ExAC database. The following filtration criteria were used: Category 1) The minor allele frequency of the variant in TWB1496 was ≥ 0.01, while that in the ExAC database and in the East Asia population of the gnomAD genome and exome database were ≤ 0.01 or absent; Category 2) the allelic frequency was 0.005–0.01 in TWB1496, and ≤ 0.005 or absent in the ExAC database and in the East Asia population of the gnomeAD genome and exome database. The genetic inheritance mode of the variants was manually annotated using the Online Mendelian Inheritance in Man database (OMIM). On the other hand, experts further reviewed the variants within ACMG actionable genes. The concept of medically actionable genes originates from ACMG recommendations on reporting secondary findings from exome or genome sequencing results. To further focus on the East Asian population, we also compared the allele frequency of variants in the carrier sets to two East Asian populations (Singaporean and Japanese). The Singaporean data were acquired from the SG10K project 4 , and the Japanese data were obtained from the 3.5KJPN project . To evaluate the clinical impact of the carrier call set, we applied the expanded carrier panel , to the above data to calculate accumulation frequencies of pathogenic and likely pathogenic variants in monogenic disease-causing genes. We compared these data with the United States cohort (sequencing method) and Taiwan domestic data 6 (array genotyping method). The overall study design and analysis workflow is shown in . In summary, for SNV and INDEL, the analyses were performed for the entire genome of the cohort (N = 1496). For HLA allele, the data was from Taiwan Biobank official release ( https://taiwanview.twbiobank.org.tw/hla.php ) (N = 1,103), and allele annotation was based on PharmGKB information. For the CYP allele, the analyses were performed on a subset of the WGS cohort (N = 1,017) and only focused on CYP genes. The SV analyses were performed for a subset of the WGS cohort (N = 494), and the analyses were for the entire genome. For SMN CNV calling, the analyses were based on a subset of the WGS cohort (N = 494), and only for SMN genes. We also used the same carrier call-set data to evaluate other monogenic disease-causing genes as a concept of expanded carrier screening. We compared the P/LP variant carrier frequency with the U.S. cohort and Taiwanese cohort (using TWBv2.0 array genotyping). We also used the Westemeyer et al. (2020) gene list (274 genes) as a virtual panel to estimate how many couples may benefit from the expanded carrier screening in the Taiwanese population. The website for browsing TaiwanGenomes is based on the open-source project VASH ( https://github.com/mbilab/vash ). VASH (a composite word of variant and flash) aims to provide a rapid and fluent user experience while browsing whole genome scale variants. VASH is implemented using Vue.js ( https://vuejs.org/ ) and Django ( https://djangoproject.com ) as frontend frameworks. MySQL is used as the database engine of Django. All requests by VASH are segmented into small chunks and cached throughout the entire processing pipeline from Vue.js to MySQL. Therefore VASH is capable of processing new chunks while users are viewing previous chunks. VASH enables infinite scrolling, which is more intuitive than pagination browsing and achieves response time in a few seconds regardless of the number of queried variants. High-quality variant call sets To evaluate the accuracy of the joint calling, we used the same pipeline to analyze seven samples (HG001 ∼ HG007) from GIAB as the internal quality control. We jointly called 1496 TWB subjects with seven reference samples and stratified the joint call set according to the VQSR tranches into strata of 100.0, 99.9, 99.8, 99.7, 99.6, 99.5, and 99.0. We then confirmed both SNV and INDEL variant calling accuracy by comparing HG001 (NA12878) genetic variants as the ground truth to the subset of HG001 results in the joint called cohort for every variant. We defined VQSR tranche 99.7 as the pass criterion to balance the overall sensitivity and specificity (a) and S2(b). All variants can be accessed from the TaiwanGenomes database ( https://genomes.tw ) including their VQSR tranche and annotated information. We only included variants with pass quality in the downstream analyses. In addition to alternative alleles, we also analyzed reference alleles and used the call rate, allele number (AN), and depth of coverage (DP) filtration for quality classification. We classified all non-alternative positions into A, B, and C categories. If the call rate of a reference allele was above 80 % (AN > 2,400 & DP > 18,000), the reference allele was defined as category A (2,686,586,573 variants). Category B (16,201,674 variants) indicated that the call rate of an allele was below 10 % (AN < 300 & DP < 4,500). Category C (30,369,949 variants) represents the call rates between categories A and B. Researchers should be careful when interpreting the variants in categories B and C as having a MAF of zero in TaiwanGenomes . It actually means missing genotypes or needs more depth of coverage to determine. The quality information for each genomic locus can be easily browsed at https://genomes.tw/#/supplement Variant detection bias or variability in population structure may have driven MAF differences between the two cohorts. We then compared the allele frequencies of 18,624 variants on chromosome 1 in which the variants were identified in both TWB1496 and gnomAD East Asian data. To derive differences in allele frequencies distribution between the two databases, we used the Python package Scipy and Statsmodels.api to fit linear regressions and produce scatter plots. We tested the correlation of MAF between the two databases among different VQSR tranches . Greater VQSR specificity and greater correlation (r-squared) to the current gnomAD East Asian dataset was found. Variants were rescued by jointly called step and confirmed by Sanger sequencing The differences between the jointly called VCF file and individual variant calls are worth noting. Theoretically, multi-sample joint calling and followed by VQSR were suggested by GATK best practices to rescue variants with moderate individual-level quality and improve the overall calling accuracy when the sample size increases ( https://gatk.broadinstitute.org/hc/en-us/articles/360035890431-The-logic-of-joint-alling-for-germline-short-variants,https://gatk.broadinstitute.org/hc/en- s/articles/360035890411 , https://gatk.broadinstitute.org/hc/en-us/articles/360035531612-Variant-Quality-Score-Recalibration-VQSR ). Hundreds of variants were checked by Sanger sequencing, but only 109 variants matching the following criteria were included in the accuracy analysis: 1) unambiguous biallelic variants; 2) clear Sanger results; 3) inconsistent results between jointly called and individually called variants. Note that the number of true negative variant calls (TN) was underestimated because the VCF file only included the positive variants. We defined VQSR tranche 99.7 as the pass criterion, which was a greater accuracy than that of the variant calls in all higher VQSR tranches, thereby reducing the false discovery rate (FDR) by 25 % (0.75 to 0.5). Based on the Sanger validation analysis, applying VQSR can remove false positive (FP) variants by 68.6 % (35/51) . Millions of novel variants found in 1496 Taiwanese Our final variant call set consisted of 59,433,212 variants from autosomes and sex chromosomes . Among these variants, there were 49,701,036 variants with MAF < 0.05 and 44,591,573 variants with MAF < 0.01. There were 480,784 variants in coding regions, and 274,265 variants were annotated as non-synonymous. For the variants satisfying the pass criterion (VQSR 99.7), the number decreased to 51,135,411. Of these, 42,557,774 variants had MAF < 0.05 and 38,599,450 had MAF < 0.01. There were 439,192 variants in coding regions and 250,940 variants annotated as nonsynonymous. To analyze the unique genetic characteristics of the Taiwanese population, we further filtered out variants that were not present in other databases (see Methods). This resulted in 16,520,159 variants classified as “globally novel.” To further investigate the population genetic structure, we sorted the samples by their total variant numbers and calculated the number of unique globally novel variants . Results from the last 50 samples suggested that a Taiwanese person has an average of 6,870.7 globally novel variants. Overall, there were 16,066,996 variants with MAF < 0.05 and 15,495,346 variants with MAF < 0.01. Most of these novel variants were in the intergenic regions. Of the novel variants, 26,664 were in coding regions, 700 were annotated as missense variants, and 4,159 were annotated as loss-of-function variants, including 151 nonsense variants, 3,174 frameshift variants, and 834 splicing variants. Loss-of-function and deleterious variants play a crucial role in Mendelian disorders. From the spectrum of variant numbers and the allele frequencies, we observed that there were far fewer loss-of-function variants than others under negative selection, which is in line with our expectations. We also observed that non-frameshift indels occurred at almost the same frequency as loss-of-function variants . 5.54% of participants had medically actionable variants We identified 1,136 variants with clinical evidence, which we termed a carrier call-set. Of these, 58 were among the 73 actionable genes recommended by ACMG 13 . These variants were further reviewed by practicing medical geneticists, resulting in 53 secondary findings. We also filtered the carrier call-set by allele frequency and found 78 variants with allele frequency ≥ 0.005. For further reduction of ambiguity, we then filtered out variants if the allele frequency was also ≥ −0.005 in any East Asian population in public databases, including ExAC EAS, gnomAD genome AF eas, and gnomAD exome AF eas. The final filtered carrier call-set consists of nine variants. On the TaiwanGenomes website, a user can set the filter combination as “CLNSIG: Pathogenic, Pathogenic/Likely pathogenic, Likely pathogenic,” resulting in these 1,136 variants. By adding the filter AF ≥ 0.005, 78 variants are retrieved, reduced to 54 by adding the filter AF ≥ 0.01. Most affected samples had only one variant from among the 53 variants considered secondary findings; average occurrence was about 5.54 % . However, the high occurrence mainly arises from three disease genes with autosomal recessive inheritance: MUTYH , ATP7B , and GAA . The ACMG guidelines suggest that athogenic or expected pathogenic variants from both alleles should be reported together. None of the samples in the TWB 1496 cohort harbored a second pathogenic variant, indicating an average occurrence of 1.7 %. Notably, one pathogenic variant with high occurrence is PTEN (NM 000314.6:c.802-2A > T, rs587782455), a splicing-acceptor variant, which did not pass the VQSR threshold and thus was excluded from further analysis. 75.3% of participants had at least one clinically important pharmacogenomic allele For PharmGKB-reported clinical variant data with a level of evidence of 1A/1B/2A, there were 13 HLA alleles and 17 clinically significant variant-drug pairs. We evaluated the frequencies of the HLA haplotypes as risk alleles with matched nomenclature fields from 1,103 TWB subjects . Most HLA haplotypes in PharmGKB are three or four fields, suggesting that high-resolution genotypes were necessary. Among the 17 variant-drug pairs, we found 16 pairs in the TWB cohort; 13 pairs had haplotype frequencies > 1 %. The most common haplotype was HLA-C*01:02:01 group alleles (16.05 %), followed by HLA-A*33:03 (12.93 %). Two alleles were associated with the adverse drug events (ADEs) of taking methazolamide and allopurinol, respectively. However, one known pharmacogenomics allele, HLA-DRB*08:03 (8 %, 178/2206), was absent from the PharmGKB clinical variant list, suggesting that a systematic review of this resource may be necessary. Among the 1,103 TWB subjects, we found 439 people who carried one risk allele (439/1103 = 39.8 %), 147 people who had at least two, 170 people who had three, 66 people who had four, and nine people who had five risk alleles. This haplotype frequency analysis suggests that approximately three out of four (831/1103 = 75.3 %) Taiwanese people have at least one risk allele, implying that a considerable proportion of people may benefit from HLA typing whenever a prescription for a corresponding drug is needed. In addition, we also evaluated the allele frequencies of PharmGKB-reported CYP variant-drug pairs in the sub-sample of 1,017 TWB volunteers (manuscript in preparation). CYP variants analysis revealed 89 pharmacogenetic variant-drug groups with a level of evidence of 1A/1B/2A . Of these groups, 54 had allele frequencies (excluding WT) ≥ 10 % in the TWB cohort. The most common variant was CYP3A5*3 (73.2 %), followed by haplotypes of the CYP2C19 group ( CYP2C19*2 , CYP2C19*3 , CYP2C19*9 , CYP2C19*10 , CYP2C19*17 , CYP2C19*24 , and CYP2C19*26 ) (36.2 %). The variants were associated with the abnormal drug metabolism of tacrolimus and omeprazole, implying that CYP typing is helpful for clinical medication safety. Since treatment alterations may occur when a genetic variant alters a treatment’s efficacy, dosage, metabolism, or pharmacokinetics or otherwise causes toxicity or an adverse drug reaction (ADR), such information is valuable for both clinicians and patients. The status of carrier variants in Taiwan, Singapore, and Japan To evaluate genomic characteristics specific to the Taiwanese, we compared the MAF of 1,136 carrier call-set from our samples with those of two different East Asian populations (Japanese and Singaporean). There were 279 common carrier variants between the Japanese and Taiwanese data sets, and 611 common carrier variants between the Singaporean and Taiwanese data sets . Monogenic disease risk in the offspring of 1.2–3.9 % of couples We compared our carrier data with that of the United States cohort , and another domestic data set generated using the array genotyping method (TWBv2) . The main results are listed in . The most striking difference relates to the GJB2 gene, a well-known causative gene for hearing impairment. The estimated carrier frequency in the Taiwan biobank NGS database was 16.7 %, more than 90 % of which was contributed by the GJB2 V37I variant (MAF = 8.6 %). GJB2 P/LP carrier frequency in the US cohort was only 6.25 %, less than half of the Taiwanese frequency. In contrast, the GJB2 carrier rate was estimated as a much lower 1.59 % with TWBv2 array genotyping. Another example is the SLC25A13 gene, related to citrullinemia. P/LP carrier frequency was 2.57 % in our series but only 0.40 % in the U.S. cohort and 1.9 % in the TWBv2 series. Another one is the PTS gene, defective variants of which lead to phenylketonuria (PKU). The P/LP carrier frequency of PTS was 0.66 % in our cohort and only 0.12 % in the U.S. cohort. The frequency in our samples is compatible with clinical observation of PKU patients in Taiwan, where BH4-deficiency (defective PTS ) PKU patients account for up to 1/4 of total PKU patients. In contrast, defective PTS PKU patients only account for 1–2 % of Caucasian cohorts. No PTS variants were shown in the TWBv2 array data. We applied the method used in Westemeyer et al. to the TWB 1496 NGS cohort data to calculate the combined at-risk couple rate in Taiwan. The employed 270 gene panel (274 genes in Westemeyer et al.) omits 4 genes ( HBA1/2 , DMD for Duchenne muscular dystrophy, SMN1 , and FMR1 ) that would identify the risk for a genetic disorder in the offspring of 1 in 28 couples (3.55 %). If DUOX2 for hypothyroidism and G6PD for glucose-6-phosphate dehydrogenase deficiency were added to the local carrier screening list, the risk ratio would be 1 in 25 couples (3.94 %) . For thalassemia, the SV-related hereditary disease in Taiwan, we used Manta and AnnotSV to identify HBA1/2 pathogenic variants in a subset of the TWB cohort (N = 494). Alpha thalassemia carrier frequency was 6.88 % (5.06 % 0 , 1.81 % + ) . For spinal muscular atrophy (SMA), another high-prevalence hereditary disease, the SMN1 / SMN2 copy number was analyzed by SMNCopyNumberCaller . Ten carriers with only one copy of SMN1 were identified among the 494 samples, equating to an SMA carrier frequency of 2.02 % . Validation of NGS-based SMA carrier screening methods in China , makes a convincing case that these assessments are as accurate as the traditional MLPA method. The comparison of alpha thalassemia carrier and SMA carrier rate in neighboring Asian regions are listed in . Variants with clinical significance and high allele frequency in Taiwan Initial data filtering with allele frequencies singled out 54 variants with MAF > 0.01 and 24 with MAF 0.005 –0.01 among the carrier call-set from 1496 TWB participants. To address the characteristics of the Taiwanese population, we further filtered these 78 variants by comparing their MAF with those in the gnomAD genome database (version 2.1.1) and the ExAC database, yielding 9 variants with clinical significance and relatively high MAF in Taiwan . High prevalence (MAF ≥ 0.01 ) clinical significance variants among Taiwan Biobank participants Splice-donor variant in CACNA1B : A splice-donor variant in CACNA1B (c.390 + 1 390 + 2insACGACACGGAGC) occurred in the TWB1496 data set with a MAF of 0.019, but was not reported in ExAC and the East Asian population of gnomAD. Stop-gained variant in DUOX2 : The stop-gained variant c.1588A > T in DUOX2 occurred with a MAF of 0.013 among TWB participants, while its MAF in the East Asian population in gnomAD was 0.0071. The protein product of DUOX2 is an oxidase and part of the peroxide-generating system located at the apical membrane of thyroid follicular cells , . Prematurely terminated protein products of DUOX2 of the noted pathogenic variant lead to a lower level of hydrogen peroxide and consequently insufficient thyroid hormone for normal human development. This variant has been identified in patients with transient or permanent hypothyroidism or iodide organification. Medium prevalence (0.005 ≤ MAF ≤ 0.01) clinically important variants among Taiwan Biobank participants Missense variant in OPN1MW : The missense variant c.989G > A in OPN1MW is reported to be causative of deuteranopia. OPN1MW , the medium-wave-sensitive opsin-1 gene, is mapped to chromosome Xq28 and encodes the green cone pigment, which is crucial to color vision. In the presence of this variant, the absorbance of the Arg330Gln mutant opsin decreased dramatically compared to normal green opsin, and it has been reported to be causative of deutan color blindness . A precise study on the prevalence of deuteranopia in Taiwan may help explain the relatively high allele frequency of this variant among Taiwan Biobank participants (MAF = 0.008) compared to that in the gnomAD East Asian population. It is also worth to note that less than 50 % of individuals in gnomAD v2.1.1 exomes covered this locus. Intronic variants in SLC25A13 : The variant c.615 + 5G > A in SLC25A13 occurred with a MAF of 0.006 in TWB participants. SLC25A13 is localized to chromosome 7q21.3 and encodes citrin, which serves as a mitochondrial solute transporter in the urea cycle. The variant is linked to neonatal intrahepatic cholestasis caused by citrin deficiency (NICCD) and adult-onset type II citrullinemia (CTLN2) , , . Frameshift variant in SERPINB7 : Splice-acceptor variant c.522dup in SERPINB7 occurred with a MAF of 0.005 among TWB participants, while its allele frequency in the East Asian population in gnomAD2 was 0.0032. This variant is causative of Nagashima-type palmoplantar keratoderma, a skin disorder characterized by hyperkeratosis of the palm and feet of affected individuals , . Frameshift variant in TTLL5 : The frameshift variant c.3177 3180del in TTL5 occurred with a MAF of 0.005 among TWB participants, while its MAF in ExAC and in the East Asian population of gnomAD was 0.0021 and 0.0045, respectively. TaiwanGenomes web browser TaiwanGenomes ( https://genomes.tw ) provides a user-friendly interface to access all variants reported in this study. The’MAIN’ tab links to the main table that contains 59,433,212 variants yielded by the joint calling of the 1496 WGS, including 51,135,411 variants passing the VQSR analysis (setting ‘VQSR = PASS’). The’SUPPLEMENT’ tab links to the supplementary table that provides information on read depth for 2,792,591,408 positions in the human genome. These positions are categorized into four classes. A: Reference allele (MAF = 0); B: Missing (MAF = n.a); C: Uncertain quality (MAF = n.a./0); and the 59,433,212 variants called that have links to the MAIN table. TaiwanGenomes provides users with the flexibility of selecting columns of interest to examine. Among the passed variants, 439,192 variants fall in the coding regions (setting ‘FILTER = PASS’ and ‘fun.refGene = exonic’), and 55,949 have a minor allele frequency ≥ 0.01 (setting ‘FILTER = PASS’ and ‘fun.refGene = exonic’ and ‘AF ≥ 0.01′). Another example of setting condition combinations of multiple selected columns is examination of all nonsynonymous variants in BRCA1 by setting ‘Gene.refGene = BRCA1′ and ‘ExonicFunc.refGene = nonsynonymous SNV’, resulting in 40 variants. To evaluate the accuracy of the joint calling, we used the same pipeline to analyze seven samples (HG001 ∼ HG007) from GIAB as the internal quality control. We jointly called 1496 TWB subjects with seven reference samples and stratified the joint call set according to the VQSR tranches into strata of 100.0, 99.9, 99.8, 99.7, 99.6, 99.5, and 99.0. We then confirmed both SNV and INDEL variant calling accuracy by comparing HG001 (NA12878) genetic variants as the ground truth to the subset of HG001 results in the joint called cohort for every variant. We defined VQSR tranche 99.7 as the pass criterion to balance the overall sensitivity and specificity (a) and S2(b). All variants can be accessed from the TaiwanGenomes database ( https://genomes.tw ) including their VQSR tranche and annotated information. We only included variants with pass quality in the downstream analyses. In addition to alternative alleles, we also analyzed reference alleles and used the call rate, allele number (AN), and depth of coverage (DP) filtration for quality classification. We classified all non-alternative positions into A, B, and C categories. If the call rate of a reference allele was above 80 % (AN > 2,400 & DP > 18,000), the reference allele was defined as category A (2,686,586,573 variants). Category B (16,201,674 variants) indicated that the call rate of an allele was below 10 % (AN < 300 & DP < 4,500). Category C (30,369,949 variants) represents the call rates between categories A and B. Researchers should be careful when interpreting the variants in categories B and C as having a MAF of zero in TaiwanGenomes . It actually means missing genotypes or needs more depth of coverage to determine. The quality information for each genomic locus can be easily browsed at https://genomes.tw/#/supplement Variant detection bias or variability in population structure may have driven MAF differences between the two cohorts. We then compared the allele frequencies of 18,624 variants on chromosome 1 in which the variants were identified in both TWB1496 and gnomAD East Asian data. To derive differences in allele frequencies distribution between the two databases, we used the Python package Scipy and Statsmodels.api to fit linear regressions and produce scatter plots. We tested the correlation of MAF between the two databases among different VQSR tranches . Greater VQSR specificity and greater correlation (r-squared) to the current gnomAD East Asian dataset was found. The differences between the jointly called VCF file and individual variant calls are worth noting. Theoretically, multi-sample joint calling and followed by VQSR were suggested by GATK best practices to rescue variants with moderate individual-level quality and improve the overall calling accuracy when the sample size increases ( https://gatk.broadinstitute.org/hc/en-us/articles/360035890431-The-logic-of-joint-alling-for-germline-short-variants,https://gatk.broadinstitute.org/hc/en- s/articles/360035890411 , https://gatk.broadinstitute.org/hc/en-us/articles/360035531612-Variant-Quality-Score-Recalibration-VQSR ). Hundreds of variants were checked by Sanger sequencing, but only 109 variants matching the following criteria were included in the accuracy analysis: 1) unambiguous biallelic variants; 2) clear Sanger results; 3) inconsistent results between jointly called and individually called variants. Note that the number of true negative variant calls (TN) was underestimated because the VCF file only included the positive variants. We defined VQSR tranche 99.7 as the pass criterion, which was a greater accuracy than that of the variant calls in all higher VQSR tranches, thereby reducing the false discovery rate (FDR) by 25 % (0.75 to 0.5). Based on the Sanger validation analysis, applying VQSR can remove false positive (FP) variants by 68.6 % (35/51) . Our final variant call set consisted of 59,433,212 variants from autosomes and sex chromosomes . Among these variants, there were 49,701,036 variants with MAF < 0.05 and 44,591,573 variants with MAF < 0.01. There were 480,784 variants in coding regions, and 274,265 variants were annotated as non-synonymous. For the variants satisfying the pass criterion (VQSR 99.7), the number decreased to 51,135,411. Of these, 42,557,774 variants had MAF < 0.05 and 38,599,450 had MAF < 0.01. There were 439,192 variants in coding regions and 250,940 variants annotated as nonsynonymous. To analyze the unique genetic characteristics of the Taiwanese population, we further filtered out variants that were not present in other databases (see Methods). This resulted in 16,520,159 variants classified as “globally novel.” To further investigate the population genetic structure, we sorted the samples by their total variant numbers and calculated the number of unique globally novel variants . Results from the last 50 samples suggested that a Taiwanese person has an average of 6,870.7 globally novel variants. Overall, there were 16,066,996 variants with MAF < 0.05 and 15,495,346 variants with MAF < 0.01. Most of these novel variants were in the intergenic regions. Of the novel variants, 26,664 were in coding regions, 700 were annotated as missense variants, and 4,159 were annotated as loss-of-function variants, including 151 nonsense variants, 3,174 frameshift variants, and 834 splicing variants. Loss-of-function and deleterious variants play a crucial role in Mendelian disorders. From the spectrum of variant numbers and the allele frequencies, we observed that there were far fewer loss-of-function variants than others under negative selection, which is in line with our expectations. We also observed that non-frameshift indels occurred at almost the same frequency as loss-of-function variants . We identified 1,136 variants with clinical evidence, which we termed a carrier call-set. Of these, 58 were among the 73 actionable genes recommended by ACMG 13 . These variants were further reviewed by practicing medical geneticists, resulting in 53 secondary findings. We also filtered the carrier call-set by allele frequency and found 78 variants with allele frequency ≥ 0.005. For further reduction of ambiguity, we then filtered out variants if the allele frequency was also ≥ −0.005 in any East Asian population in public databases, including ExAC EAS, gnomAD genome AF eas, and gnomAD exome AF eas. The final filtered carrier call-set consists of nine variants. On the TaiwanGenomes website, a user can set the filter combination as “CLNSIG: Pathogenic, Pathogenic/Likely pathogenic, Likely pathogenic,” resulting in these 1,136 variants. By adding the filter AF ≥ 0.005, 78 variants are retrieved, reduced to 54 by adding the filter AF ≥ 0.01. Most affected samples had only one variant from among the 53 variants considered secondary findings; average occurrence was about 5.54 % . However, the high occurrence mainly arises from three disease genes with autosomal recessive inheritance: MUTYH , ATP7B , and GAA . The ACMG guidelines suggest that athogenic or expected pathogenic variants from both alleles should be reported together. None of the samples in the TWB 1496 cohort harbored a second pathogenic variant, indicating an average occurrence of 1.7 %. Notably, one pathogenic variant with high occurrence is PTEN (NM 000314.6:c.802-2A > T, rs587782455), a splicing-acceptor variant, which did not pass the VQSR threshold and thus was excluded from further analysis. For PharmGKB-reported clinical variant data with a level of evidence of 1A/1B/2A, there were 13 HLA alleles and 17 clinically significant variant-drug pairs. We evaluated the frequencies of the HLA haplotypes as risk alleles with matched nomenclature fields from 1,103 TWB subjects . Most HLA haplotypes in PharmGKB are three or four fields, suggesting that high-resolution genotypes were necessary. Among the 17 variant-drug pairs, we found 16 pairs in the TWB cohort; 13 pairs had haplotype frequencies > 1 %. The most common haplotype was HLA-C*01:02:01 group alleles (16.05 %), followed by HLA-A*33:03 (12.93 %). Two alleles were associated with the adverse drug events (ADEs) of taking methazolamide and allopurinol, respectively. However, one known pharmacogenomics allele, HLA-DRB*08:03 (8 %, 178/2206), was absent from the PharmGKB clinical variant list, suggesting that a systematic review of this resource may be necessary. Among the 1,103 TWB subjects, we found 439 people who carried one risk allele (439/1103 = 39.8 %), 147 people who had at least two, 170 people who had three, 66 people who had four, and nine people who had five risk alleles. This haplotype frequency analysis suggests that approximately three out of four (831/1103 = 75.3 %) Taiwanese people have at least one risk allele, implying that a considerable proportion of people may benefit from HLA typing whenever a prescription for a corresponding drug is needed. In addition, we also evaluated the allele frequencies of PharmGKB-reported CYP variant-drug pairs in the sub-sample of 1,017 TWB volunteers (manuscript in preparation). CYP variants analysis revealed 89 pharmacogenetic variant-drug groups with a level of evidence of 1A/1B/2A . Of these groups, 54 had allele frequencies (excluding WT) ≥ 10 % in the TWB cohort. The most common variant was CYP3A5*3 (73.2 %), followed by haplotypes of the CYP2C19 group ( CYP2C19*2 , CYP2C19*3 , CYP2C19*9 , CYP2C19*10 , CYP2C19*17 , CYP2C19*24 , and CYP2C19*26 ) (36.2 %). The variants were associated with the abnormal drug metabolism of tacrolimus and omeprazole, implying that CYP typing is helpful for clinical medication safety. Since treatment alterations may occur when a genetic variant alters a treatment’s efficacy, dosage, metabolism, or pharmacokinetics or otherwise causes toxicity or an adverse drug reaction (ADR), such information is valuable for both clinicians and patients. To evaluate genomic characteristics specific to the Taiwanese, we compared the MAF of 1,136 carrier call-set from our samples with those of two different East Asian populations (Japanese and Singaporean). There were 279 common carrier variants between the Japanese and Taiwanese data sets, and 611 common carrier variants between the Singaporean and Taiwanese data sets . We compared our carrier data with that of the United States cohort , and another domestic data set generated using the array genotyping method (TWBv2) . The main results are listed in . The most striking difference relates to the GJB2 gene, a well-known causative gene for hearing impairment. The estimated carrier frequency in the Taiwan biobank NGS database was 16.7 %, more than 90 % of which was contributed by the GJB2 V37I variant (MAF = 8.6 %). GJB2 P/LP carrier frequency in the US cohort was only 6.25 %, less than half of the Taiwanese frequency. In contrast, the GJB2 carrier rate was estimated as a much lower 1.59 % with TWBv2 array genotyping. Another example is the SLC25A13 gene, related to citrullinemia. P/LP carrier frequency was 2.57 % in our series but only 0.40 % in the U.S. cohort and 1.9 % in the TWBv2 series. Another one is the PTS gene, defective variants of which lead to phenylketonuria (PKU). The P/LP carrier frequency of PTS was 0.66 % in our cohort and only 0.12 % in the U.S. cohort. The frequency in our samples is compatible with clinical observation of PKU patients in Taiwan, where BH4-deficiency (defective PTS ) PKU patients account for up to 1/4 of total PKU patients. In contrast, defective PTS PKU patients only account for 1–2 % of Caucasian cohorts. No PTS variants were shown in the TWBv2 array data. We applied the method used in Westemeyer et al. to the TWB 1496 NGS cohort data to calculate the combined at-risk couple rate in Taiwan. The employed 270 gene panel (274 genes in Westemeyer et al.) omits 4 genes ( HBA1/2 , DMD for Duchenne muscular dystrophy, SMN1 , and FMR1 ) that would identify the risk for a genetic disorder in the offspring of 1 in 28 couples (3.55 %). If DUOX2 for hypothyroidism and G6PD for glucose-6-phosphate dehydrogenase deficiency were added to the local carrier screening list, the risk ratio would be 1 in 25 couples (3.94 %) . For thalassemia, the SV-related hereditary disease in Taiwan, we used Manta and AnnotSV to identify HBA1/2 pathogenic variants in a subset of the TWB cohort (N = 494). Alpha thalassemia carrier frequency was 6.88 % (5.06 % 0 , 1.81 % + ) . For spinal muscular atrophy (SMA), another high-prevalence hereditary disease, the SMN1 / SMN2 copy number was analyzed by SMNCopyNumberCaller . Ten carriers with only one copy of SMN1 were identified among the 494 samples, equating to an SMA carrier frequency of 2.02 % . Validation of NGS-based SMA carrier screening methods in China , makes a convincing case that these assessments are as accurate as the traditional MLPA method. The comparison of alpha thalassemia carrier and SMA carrier rate in neighboring Asian regions are listed in . Initial data filtering with allele frequencies singled out 54 variants with MAF > 0.01 and 24 with MAF 0.005 –0.01 among the carrier call-set from 1496 TWB participants. To address the characteristics of the Taiwanese population, we further filtered these 78 variants by comparing their MAF with those in the gnomAD genome database (version 2.1.1) and the ExAC database, yielding 9 variants with clinical significance and relatively high MAF in Taiwan . ≥ 0.01 ) clinical significance variants among Taiwan Biobank participants Splice-donor variant in CACNA1B : A splice-donor variant in CACNA1B (c.390 + 1 390 + 2insACGACACGGAGC) occurred in the TWB1496 data set with a MAF of 0.019, but was not reported in ExAC and the East Asian population of gnomAD. Stop-gained variant in DUOX2 : The stop-gained variant c.1588A > T in DUOX2 occurred with a MAF of 0.013 among TWB participants, while its MAF in the East Asian population in gnomAD was 0.0071. The protein product of DUOX2 is an oxidase and part of the peroxide-generating system located at the apical membrane of thyroid follicular cells , . Prematurely terminated protein products of DUOX2 of the noted pathogenic variant lead to a lower level of hydrogen peroxide and consequently insufficient thyroid hormone for normal human development. This variant has been identified in patients with transient or permanent hypothyroidism or iodide organification. Missense variant in OPN1MW : The missense variant c.989G > A in OPN1MW is reported to be causative of deuteranopia. OPN1MW , the medium-wave-sensitive opsin-1 gene, is mapped to chromosome Xq28 and encodes the green cone pigment, which is crucial to color vision. In the presence of this variant, the absorbance of the Arg330Gln mutant opsin decreased dramatically compared to normal green opsin, and it has been reported to be causative of deutan color blindness . A precise study on the prevalence of deuteranopia in Taiwan may help explain the relatively high allele frequency of this variant among Taiwan Biobank participants (MAF = 0.008) compared to that in the gnomAD East Asian population. It is also worth to note that less than 50 % of individuals in gnomAD v2.1.1 exomes covered this locus. Intronic variants in SLC25A13 : The variant c.615 + 5G > A in SLC25A13 occurred with a MAF of 0.006 in TWB participants. SLC25A13 is localized to chromosome 7q21.3 and encodes citrin, which serves as a mitochondrial solute transporter in the urea cycle. The variant is linked to neonatal intrahepatic cholestasis caused by citrin deficiency (NICCD) and adult-onset type II citrullinemia (CTLN2) , , . Frameshift variant in SERPINB7 : Splice-acceptor variant c.522dup in SERPINB7 occurred with a MAF of 0.005 among TWB participants, while its allele frequency in the East Asian population in gnomAD2 was 0.0032. This variant is causative of Nagashima-type palmoplantar keratoderma, a skin disorder characterized by hyperkeratosis of the palm and feet of affected individuals , . Frameshift variant in TTLL5 : The frameshift variant c.3177 3180del in TTL5 occurred with a MAF of 0.005 among TWB participants, while its MAF in ExAC and in the East Asian population of gnomAD was 0.0021 and 0.0045, respectively. TaiwanGenomes ( https://genomes.tw ) provides a user-friendly interface to access all variants reported in this study. The’MAIN’ tab links to the main table that contains 59,433,212 variants yielded by the joint calling of the 1496 WGS, including 51,135,411 variants passing the VQSR analysis (setting ‘VQSR = PASS’). The’SUPPLEMENT’ tab links to the supplementary table that provides information on read depth for 2,792,591,408 positions in the human genome. These positions are categorized into four classes. A: Reference allele (MAF = 0); B: Missing (MAF = n.a); C: Uncertain quality (MAF = n.a./0); and the 59,433,212 variants called that have links to the MAIN table. TaiwanGenomes provides users with the flexibility of selecting columns of interest to examine. Among the passed variants, 439,192 variants fall in the coding regions (setting ‘FILTER = PASS’ and ‘fun.refGene = exonic’), and 55,949 have a minor allele frequency ≥ 0.01 (setting ‘FILTER = PASS’ and ‘fun.refGene = exonic’ and ‘AF ≥ 0.01′). Another example of setting condition combinations of multiple selected columns is examination of all nonsynonymous variants in BRCA1 by setting ‘Gene.refGene = BRCA1′ and ‘ExonicFunc.refGene = nonsynonymous SNV’, resulting in 40 variants. It is noteworthy that all TWB subjects were 20 year-old adults with no distinct developmental defects or malignant tumors at the time they joined the study. Unlike the genotyping array approach that focuses on detecting the associations between genomic haplotype blocks and phenotypic traits, personal genome sequences directly uncover all hereditary risks. In this study, we reanalyzed 1496 whole-genome sequence data sets from the Taiwan Biobank, which is one of the few existing valuable datasets for the east Asian population. Our reanalysis generated acomprehensive medical genomic profile, which complements previous Taiwan Biobank studies and reinforces the significance of WGS data. We utilized benchmarked data from GIAB and jointly called 1496 samples to determine a cut-off VQSR tranche of 99.7, and provided VQSR quality flags for each locus on the website ( TaiwanGenomes https://genomes.tw ). We further selected hundreds of variants irrespective of VQSR tranche and validated them by the Sanger sequencing. All variants that were removed in the joint calling step had negative Sanger sequencing results, indicating the effectiveness of joint calling to eliminate false positives in individual variant calling. Among variants that were retrieved by joint calling, only some had positive Sanger sequencing results, suggesting that further VQSR filtering criteria may be necessary. For the entile reference genome, we classified all non-alternative allele regions into three categories by overall call rate and depth of coverage. This allowed us to clearly distinguish reference homozygous regions or difficult-to-map regions in the genome reference for the TWB 1496 cohort. We found that our TWB samples yielded an average of 6,871 globally novel variants for a Taiwanese person. In this study, we utilized WGS data to estimate carrier risk in Taiwan. For the secondary findings, we checked for presence of the variant in ACMG SF v3.0 gene list. This is the first time ACMG SF v3.0 has been applied to the East Asian population. We found that 5.54 % of the 1496 participants were carriers of ACMG SF v3.0 genes, that 1.67 % carried autosomal dominant and X-linked diseases, and 3.87 % carried recessive diseases. Notably, one pathogenic variant NM000314.6:c.802-2A > T on the PTEN gene was caused by a technical error in the TWB 1496 cohort. In our analysis, this variant was located at the end of a polyT region with strand bias evidence, highlighting the necessity of applying VQSR and quality checks in any reanalyses of WGS data. We also analyzed a 270 gene panel to estimate the accumulated monogenic disease risk and found that 3.55 % of offspring would be carriers, implying that approximately 1 in 28 couples might benefit from carrier screening. Two main pathogenic variant contributors were GJB2 :NP_003995.2:p.Val37Ile (rs72474224, MAF: 8.6 %) and CFTR :NM_000492.4:c.1210-11_1210-10insG (rs551227135, MAF: 0.9 %). Both are considered “mild” pathogenic variants. Phenotype penetration of GJB2 (rs72474224) is highly variable in Taiwan's clinical experience. CFTR (rs551227135) is one of the classical CFTR IVS8-5 T variants responsible for non-classical cystic fibrosis presentation (CABVD, recurrent pancreatitis, late-onset CF). The regional normalization of variants is complicated and annotations are still controversial. Another caveat to carrier screening is the inability to detect long fragment variants (such as CNV, large deletion, trinucleotide repeats, and gene conversion) from short-reads data. Use of short-read data will often underestimate the carrier frequency of four important genes: HBA1/2 , DMD , FMR1 , and SMN1 , of which many known pathogenic variants are relatively large. Concerning long-fragment genetic variants, we randomly selected 494 samples for deciphering the population structure variation profile on the thalassemia genes HBA1/2 and the copy number on SMN1/2 genes. The overall population profile was similar to those found in previous studies using a target panel approach, suggesting that WGS can potentially replace target panels once an analysis pipeline has been built . Furthermore, our results showed that 75.3 % (831/1103) of the cohort was vulnerable to severe ADRs since they carried at least one risk HLA allele. Although several studies have implied that HLA types can be inferred from SNP genotyping results, full HLA haplotypes from sequence data were found to be unbiased for population-specific rare haplotypes. Similarly, SNVs on CYP genes could be detected by SNP genotyping, but WGS can reveal complete haplotype information, which is more accurate in assessing susceptibility. Our findings reveal that Taiwanese carry a high frequency (MAF > 10 %) of abnormal alleles in more than 60 % of the pharmacogenetic variant-drug groups. Whole-genome sequencing has become affordable for both research and clinical use. Since a person only needs to be sequenced once in a lifetime, a more comprehensive reanalysis of their genomic profile can be of great clinical value. Our study highlights the potential uses and benefits of a complete genomic profile with medical information for at both the population and individual level. This project was supported by the research grants from the Ministry of Science and Technology in Taiwan (MOST 109-2622-B-002-004-CC2, MOST 109-2221-E-002-162-MY3 and MOST 110-2320-B-002-078), National Science and Technology Council in Taiwan (NSTC 111-2320-B-002-091-MY3), and National Taiwan University Hospital (UN109-070). This work was also financially supported by the “Center for Advanced Computing and Imaging in Biomedicine (NTU-112L900701)” from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2008 (5). Informed consent was obtained from all patients for being included in the study. The Institutional Review Board Approval from Biomedical Science Research of Academia Sinica, Taiwan (IRB-BM) and the Ethics and Governance Council (EGC) of Taiwan Biobank, Taiwan, were obtained and gave ethical approval for this work (AS-IRB01-18041(N)). All raw sequence data used in this study were generated as part of the Taiwan Biobank project. This study did not reveal any individual participant’s information, and none of the results can be used to identify individual participants. Jacob Shujui Hsu: Conceptualization, Formal analysis, Funding acquisition, Writing – original draft, Writing – review & editing, Supervision. Dung-Chi Wu: Formal analysis ,Writing – original draft, Software. Shang-Hung Shih: Formal analysis, Writing – original draft. Jen-Feng Liu: Formal analysis, Writing – original draft, Data curation. Ya-Chen Tsai: Formal analysis, Software. Tung-Lin Lee: Formal analysis, Data curation. Wei-An Chen: Formal analysis. Yi- Hsuan Tseng: Formal analysis. Yi-Chung Lo: Software. Hong-Ye Lin: Software. Yi-Chieh Chen: Formal analysis. Jing-Yi Chen: Software. Ting-Hsuan Chou: Formal analysis. Darby Tienhao Chang: Software, Supervision. Ming Wei Su: Data curation. Wei-Hong Guo: Formal analysis. Hsin-Hsiang Mao: Formal analysis. Chien-Yu Chen: Conceptualization, Funding acquisition, Software, Writing – review & editing, Supervision. Pei-Lung Chen: Conceptualization, Formal analysis, Data curation, Funding acquisition, Writing – review & editing, Supervision. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. |
Tumour Marker Expression in Head and Neck Malignancies to Identify Potential Targets for Intraoperative Molecular Near-Infrared Imaging | 750c14d7-9617-406d-a2ef-fd3f7ae31773 | 11512873 | Anatomy[mh] | Oral and laryngeal squamous cell carcinoma (OSSC and LSSC, respectively) and papillary thyroid carcinoma (PTC) are prevalent head and neck cancers (HNCs). Surgical resection is their cornerstone treatment, yet the management of HNC remains challenging due to their intricate anatomy, proximity to vital structures, and the potential functional impairments following surgery. Precise excision of tumours is crucial in these cases to optimize patient outcomes. Inadequate resection margins [i.e., close (1–5 mm) and positive (≤ 1 mm) resection margins together] are a major negative prognostic factor for overall survival across these cancers , and are nonetheless common in HNCs with reports in up to 19%, 26% and 85% of cases for PTC, LSSC and OSSC, respectively . Enhanced intraoperative detection, including the ability to identify multifocality in PTC patients, is vital for guiding the extent of surgery, where it may significantly influence surgical decisions. In contrast to OSCC and LSSC, where the primary surgical goal is to achieve clear margins due to their aggressive nature and poorer prognosis, PTC generally has a better survival outcome. Therefore, the balance between complete tumour resection and maintaining quality of life post-surgery is particularly critical in PTC. FLI can aid in achieving this balance by precisely delineating tumour margins, potentially reducing the extent of unnecessary tissue removal and preserving function. Furthermore, FLI could identify patients who were mistakenly preoperatively selected for a hemithyroidectomy. If FLI reveals large extrathyroidal extension, a contralateral malignancy or lymph node involvement, one could directly perform a total thyroidectomy to enable treatment with radioactive iodine if deemed appropriate. This highlights the urgent clinical need for advanced intraoperative tools to facilitate precise tumour resection. Fluorescence imaging (FLI) of tumours in the near-infrared (NIR) region has shown significant potential and has been increasingly recognized as an effective approach for improving the precision of tumour excision during surgery . This technique involves the administration of a tumour-targeting fluorescent tracer before surgery, which binds to overexpressed molecular targets on cancer cells. One of various targeting strategies, such as using antibodies, small peptides, or activatable fluorophores, is employed to ensure precise tumour localization . Dedicated camera systems are used to intraoperatively detect the fluorescent signal, facilitating real-time imaging of cancer tissue. Fluorophores emitting in the NIR range (700–900 nm)—a spectrum invisible to the naked eye—are used to avoid interference in the surgeon's field of view, optimize tissue penetration, and minimize autofluorescence from surrounding tissues . Currently, there are no NIR imaging tracers available for HNCs that have been approved for clinical use . Identifying targets that are overexpressed in HNC cells is crucial for developing effective tumour-targeting tracers. Moreover, these targets ideally should not be expressed in healthy tissues, to minimize background signal and improve specificity. With a focus on potential rapid translation towards clinical applications, this study evaluated six targets with (pre-)clinically available NIR fluorescent tracers that are currently being investigated in various cancers: carcinoembryonic antigen (CEA) , epidermal growth factor receptor (EGFR), epithelial cell adhesion molecule (EpCAM) , vascular endothelial growth factor α (VEGF-α) , c-mesenchymal-epithelial transition factor (c-MET) , and integrin αvβ6 . While these targets have shown potential for tumour imaging, the literature regarding their overexpression in OSCC is incomplete, and nearly absent in LSCC and PTC . Consequently, our study aimed to assess the suitability of these six targets for future clinical trials on FLI-based tumour imaging in HNC. We examined their expression in OSCC, LSCC and PTC using immunohistochemical (IHC) staining, with an emphasis on their presence at the invasive front. These findings may expedite the initiation of FLI clinical trials in HNC, contributing to the development of targeted imaging tracers. Tissue and Tumour Marker Selection Formalin-fixed paraffin-embedded (FFPE) tissue specimens of patients with OSCC (n = 20), LSSC ( n = 10) and PTC ( n = 10) were obtained from the Erasmus Medical Centre Tissue Bank. These samples, collected from patients that were surgically treated between 2014 and 2023, were reviewed for selection by dedicated pathologists (SKol and FvK). Only specimens containing both tumour and (adjacent) healthy epithelium and muscle tissue were selected for OSCC and LSCC. For thyroid tissue, two specimens per patient were selected: one consisting of PTC tissue, and one consisting of healthy thyroid tissue. Seven out of ten LSCC patients had received radiotherapy, and one out of ten PTC patients had received chemotherapy. None of the OSCC patients had received pre-operative adjuvant therapy. The tissues were sectioned (4 µm thickness) and immunohistochemically stained for expression of the six tumour markers: CEA, EpCAM, integrin αvβ6, VEGF-α, EGFR and c-MET (see Online Supplementary Material (OSM) Table 1). Immunohistochemical Staining Protocol The tissue slides were deparaffinized using xylene, followed by rehydration. After rinsing with demineralized water, slides were treated with 0.3% hydrogen peroxide (Merck Millipore, Darmstadt, Germany) for 20 min at room temperature to prevent endogenous peroxidase activity. Antigen retrieval methods varied based on antibody protocols (see OSM Table 1). For CEA and EpCAM, heat induction at 95 °C for 10 min using Envision FLEX target retrieval solution with low-pH (pH 6.0, DAKO) was performed. For c-MET and EGFR, Envision FLEX target retrieval solution with high-pH (pH 9.0, DAKO) was utilized. For αvβ6, antigen retrieval was carried out with 0.4% pepsin at 37 °C using a water bath for 20 min. No antigen retrieval was used for VEGF-α. Primary antibody incubation was done overnight at room temperature, diluted in 1% w/v Bovine Serum Albumin/Phosphate-Buffered Saline (BSA/PBS). Negative control samples were incubated with 1% w/v BSA/PBS. Tissue sections that were known to express the specific marker(s) served as positive controls. The next day, the slides were washed with PBS and incubated with HRP-labeled secondary antibodies (Envision+ System anti-mouse or anti-rabbit, both from DAKO) for 30 min at room temperature. A liquid 3,3'-diaminobenzidine (DAB+) Substrate Chromogen System (DAKO) was applied for 10 min for staining, followed by a 15-s application of haematoxylin (VWR international, Amsterdam, the Netherlands) for counterstaining. Lastly, the tissue sections were dehydrated for 60 min at 37 °C and mounted in Pertex (Histolab, Askim, Sweden). Scoring Methods A specialized head and neck pathologist (SKop) evaluated all samples using the total immunostaining score (TIS) system, comprising intensity score (IS) and proportion score (PS), as has been previously described . The IS reflects the intensity of staining in cells, ranging from 0 to 3 (0: none; 1: weak; 2: moderate; 3: strong). The PS represents the percentage of cells that exhibited staining, ranging from 0 to 4 (0: no staining; 1: 1–10%; 2: 11–50%; 3; 51–80%; 4: 81–100%). The staining of background tissues, specifically healthy epithelium, muscle and thyroid tissue, was evaluated using the same scoring method. The TIS, a product of IS and PS, categorized marker expression into absent (TIS 0), low (TIS 1–5), moderate (TIS 6–8) or high (TIS 9–12). The expression rate was thus determined by calculating the proportion of markers exhibiting high, moderate and low expression. Furthermore, while this scoring system is commonly used in IHC evaluations for therapeutic targets, its application in this study is aimed at assessing the potential of these markers for FLI. The emphasis on contrast and precise delineation at the tumour's invasive front is particularly relevant for imaging purposes. The expression pattern at the tumour's invasive front was therefore assessed separately in each sample. Markers with strong localized staining there, despite a lower overall PS, can still be useful for FLI. This approach ensured that even markers with limited overall expression were evaluated for their potential to enhance tumour delineation in FLI. Statistics Statistical analysis was performed using the R-studio software (version 4.1.0, R Foundation for Statistical Computing, Vienna, Austria). Patient characteristics were reported using descriptive statistics. The Mann-Whitney U test compared marker expression between tumour and healthy tissues, considering a p -value < 0.05 as significant. Formalin-fixed paraffin-embedded (FFPE) tissue specimens of patients with OSCC (n = 20), LSSC ( n = 10) and PTC ( n = 10) were obtained from the Erasmus Medical Centre Tissue Bank. These samples, collected from patients that were surgically treated between 2014 and 2023, were reviewed for selection by dedicated pathologists (SKol and FvK). Only specimens containing both tumour and (adjacent) healthy epithelium and muscle tissue were selected for OSCC and LSCC. For thyroid tissue, two specimens per patient were selected: one consisting of PTC tissue, and one consisting of healthy thyroid tissue. Seven out of ten LSCC patients had received radiotherapy, and one out of ten PTC patients had received chemotherapy. None of the OSCC patients had received pre-operative adjuvant therapy. The tissues were sectioned (4 µm thickness) and immunohistochemically stained for expression of the six tumour markers: CEA, EpCAM, integrin αvβ6, VEGF-α, EGFR and c-MET (see Online Supplementary Material (OSM) Table 1). The tissue slides were deparaffinized using xylene, followed by rehydration. After rinsing with demineralized water, slides were treated with 0.3% hydrogen peroxide (Merck Millipore, Darmstadt, Germany) for 20 min at room temperature to prevent endogenous peroxidase activity. Antigen retrieval methods varied based on antibody protocols (see OSM Table 1). For CEA and EpCAM, heat induction at 95 °C for 10 min using Envision FLEX target retrieval solution with low-pH (pH 6.0, DAKO) was performed. For c-MET and EGFR, Envision FLEX target retrieval solution with high-pH (pH 9.0, DAKO) was utilized. For αvβ6, antigen retrieval was carried out with 0.4% pepsin at 37 °C using a water bath for 20 min. No antigen retrieval was used for VEGF-α. Primary antibody incubation was done overnight at room temperature, diluted in 1% w/v Bovine Serum Albumin/Phosphate-Buffered Saline (BSA/PBS). Negative control samples were incubated with 1% w/v BSA/PBS. Tissue sections that were known to express the specific marker(s) served as positive controls. The next day, the slides were washed with PBS and incubated with HRP-labeled secondary antibodies (Envision+ System anti-mouse or anti-rabbit, both from DAKO) for 30 min at room temperature. A liquid 3,3'-diaminobenzidine (DAB+) Substrate Chromogen System (DAKO) was applied for 10 min for staining, followed by a 15-s application of haematoxylin (VWR international, Amsterdam, the Netherlands) for counterstaining. Lastly, the tissue sections were dehydrated for 60 min at 37 °C and mounted in Pertex (Histolab, Askim, Sweden). A specialized head and neck pathologist (SKop) evaluated all samples using the total immunostaining score (TIS) system, comprising intensity score (IS) and proportion score (PS), as has been previously described . The IS reflects the intensity of staining in cells, ranging from 0 to 3 (0: none; 1: weak; 2: moderate; 3: strong). The PS represents the percentage of cells that exhibited staining, ranging from 0 to 4 (0: no staining; 1: 1–10%; 2: 11–50%; 3; 51–80%; 4: 81–100%). The staining of background tissues, specifically healthy epithelium, muscle and thyroid tissue, was evaluated using the same scoring method. The TIS, a product of IS and PS, categorized marker expression into absent (TIS 0), low (TIS 1–5), moderate (TIS 6–8) or high (TIS 9–12). The expression rate was thus determined by calculating the proportion of markers exhibiting high, moderate and low expression. Furthermore, while this scoring system is commonly used in IHC evaluations for therapeutic targets, its application in this study is aimed at assessing the potential of these markers for FLI. The emphasis on contrast and precise delineation at the tumour's invasive front is particularly relevant for imaging purposes. The expression pattern at the tumour's invasive front was therefore assessed separately in each sample. Markers with strong localized staining there, despite a lower overall PS, can still be useful for FLI. This approach ensured that even markers with limited overall expression were evaluated for their potential to enhance tumour delineation in FLI. Statistical analysis was performed using the R-studio software (version 4.1.0, R Foundation for Statistical Computing, Vienna, Austria). Patient characteristics were reported using descriptive statistics. The Mann-Whitney U test compared marker expression between tumour and healthy tissues, considering a p -value < 0.05 as significant. Tissue samples ( n = 40) of patients with OSCC ( n = 20), LSCC ( n = 10) and PTC ( n = 10) were included in this study. Patient and tumour characteristics are presented in Table . Target Expression in Oral Tissues In OSCC, integrin αvβ6 expression was significantly higher than in healthy epithelium ( p < 0.001) and muscle tissue ( p < 0.001). Moderate to high expression was observed in 80% of tumour samples (median TIS: 12), whereas expression in epithelium and muscle tissue was low to absent (median TIS: 2.5 and 2, respectively). EGFR expression was moderate to high in 87.5% of the tumour samples (median TIS: 8), significantly surpassing that in epithelial and muscle tissue (median TIS: 5 and 0, p = 0.028 and p < 0.001 respectively). VEGF-α expression was moderate to high in 87.5% of both tumour and muscle tissue samples (median TIS: 8 vs. 8, p = 0.701). Expressions of c-MET, CEA and EpCAM were negligible in all tumour samples (median TIS: 0.5, 1.5, and 0, respectively). Marker expression in representative OSCC is shown in Fig. and OSM Fig. 1a. Comparisons of TIS between OSCC, epithelium and muscle tissue for all six tumour markers are presented in Fig. and Table a. Expression rates are presented in OSM Table 2a and OSM Fig. 2. Target Expression in Laryngeal Tissues In laryngeal tissues, integrin αvβ6 expression was moderate to high in 90% of tumour samples (median TIS: 8), low to moderate in epithelium (median TIS: 6), and predominantly absent in muscle tissue (median TIS: 0). TIS was significantly higher in tumour than in muscle tissue (p = 0.002). EGFR expression was moderate to high in all tumour samples (100%; median TIS: 12), significantly exceeding that in epithelial and muscle tissue (median TIS: 7 and 0; p = 0.007 and p = 0.001, respectively). VEGF-α expression was moderate to high in both tumour and muscle tissue (median TIS: 10 vs. 8, p = 0.118). While absent in the majority of samples, there was moderate to high expression of EpCAM in 30% of tumours (median TIS: 0). Expressions of c-MET and CEA were (near-)absent in all tumour samples (median TIS: 0 and 0.5, respectively). See Fig. , OSM Fig. 1b, Fig. , Table b, OSM Table 2b and OSM Fig. 2. Target Expression in Thyroid Tissues In thyroid tissues, EpCAM was moderately to highly expressed in 90% of PTC samples, but this was not significantly higher than in healthy thyroid tissue (median TIS 12 vs. 8, p = 0.057). EGFR and VEGF-α expressions were low to moderate in all tumour samples (100%; median TIS: 6 and 6, respectively), and significantly higher than in healthy thyroid (median TIS: 2 and 1, p = 0.006 and p < 0.001, respectively). Integrin αvβ6, c-MET and CEA expressions ranged from absent to low in the majority of tumour samples (90%, 80% and 100%, respectively). However, despite being predominantly low, integrin αvβ6 and c-MET expression in PTC were significantly higher than in healthy thyroid tissue (median TIS 3.5 vs. 0, p = 0.002 and median TIS 2 vs. 0, p = 0.002, respectively). See Fig. , OSM Fig. 1c, Fig. , Table c, OSM Table c, OSM Fig. 2. Expression Pattern in Tumour Tissue For all markers, no notable staining patterns were observed. More specifically, none of the targets were solely overexpressed at the invasive front. In OSCC, integrin αvβ6 expression was significantly higher than in healthy epithelium ( p < 0.001) and muscle tissue ( p < 0.001). Moderate to high expression was observed in 80% of tumour samples (median TIS: 12), whereas expression in epithelium and muscle tissue was low to absent (median TIS: 2.5 and 2, respectively). EGFR expression was moderate to high in 87.5% of the tumour samples (median TIS: 8), significantly surpassing that in epithelial and muscle tissue (median TIS: 5 and 0, p = 0.028 and p < 0.001 respectively). VEGF-α expression was moderate to high in 87.5% of both tumour and muscle tissue samples (median TIS: 8 vs. 8, p = 0.701). Expressions of c-MET, CEA and EpCAM were negligible in all tumour samples (median TIS: 0.5, 1.5, and 0, respectively). Marker expression in representative OSCC is shown in Fig. and OSM Fig. 1a. Comparisons of TIS between OSCC, epithelium and muscle tissue for all six tumour markers are presented in Fig. and Table a. Expression rates are presented in OSM Table 2a and OSM Fig. 2. In laryngeal tissues, integrin αvβ6 expression was moderate to high in 90% of tumour samples (median TIS: 8), low to moderate in epithelium (median TIS: 6), and predominantly absent in muscle tissue (median TIS: 0). TIS was significantly higher in tumour than in muscle tissue (p = 0.002). EGFR expression was moderate to high in all tumour samples (100%; median TIS: 12), significantly exceeding that in epithelial and muscle tissue (median TIS: 7 and 0; p = 0.007 and p = 0.001, respectively). VEGF-α expression was moderate to high in both tumour and muscle tissue (median TIS: 10 vs. 8, p = 0.118). While absent in the majority of samples, there was moderate to high expression of EpCAM in 30% of tumours (median TIS: 0). Expressions of c-MET and CEA were (near-)absent in all tumour samples (median TIS: 0 and 0.5, respectively). See Fig. , OSM Fig. 1b, Fig. , Table b, OSM Table 2b and OSM Fig. 2. In thyroid tissues, EpCAM was moderately to highly expressed in 90% of PTC samples, but this was not significantly higher than in healthy thyroid tissue (median TIS 12 vs. 8, p = 0.057). EGFR and VEGF-α expressions were low to moderate in all tumour samples (100%; median TIS: 6 and 6, respectively), and significantly higher than in healthy thyroid (median TIS: 2 and 1, p = 0.006 and p < 0.001, respectively). Integrin αvβ6, c-MET and CEA expressions ranged from absent to low in the majority of tumour samples (90%, 80% and 100%, respectively). However, despite being predominantly low, integrin αvβ6 and c-MET expression in PTC were significantly higher than in healthy thyroid tissue (median TIS 3.5 vs. 0, p = 0.002 and median TIS 2 vs. 0, p = 0.002, respectively). See Fig. , OSM Fig. 1c, Fig. , Table c, OSM Table c, OSM Fig. 2. For all markers, no notable staining patterns were observed. More specifically, none of the targets were solely overexpressed at the invasive front. Our study investigated six potential imaging targets for various head and neck tumours. For oral and laryngeal cancer, integrin αvβ6 emerged as a key marker, exhibiting moderate to high overexpression in 80% and 90% of cases, respectively, as corroborated by previous research . Its absence in muscle and low expression in epithelium results in a strong contrast for imaging of both superficial and deep tumour margins, further supported by its successful targeting in colorectal cancer with cRGD-ZW800-1, a zwitterionic, integrin-targeted NIR tracer, currently under investigation in OSCC (NCT 04191460) . Although EGFR shows promise as an imaging target due to its negligible expression in muscle tissue, its inconsistent expression in oral and laryngeal epithelium limits its effectiveness in tumour detection and epithelial delineation . Indeed, while antibody-dye conjugates cetuximab-IRDye800CW and panitumumab-IRDye800CW have shown promise in OSCC, their ability to provide clear contrast is compromised by the sequestration of labelled antibodies by off-target, normal tissues . Consequently, pre-dosing with the unlabelled antibody has been proposed to enhance the tumour-specificity of these tracers . Next, although FLI with VEGF-α targeted tracer bevacizumab-800CW has shown promise in other tumour types, such as breast and pancreatic cancer , the widespread presence of VEGF-α in the surrounding muscle tissues of oral and laryngeal cancer undermines its utility in detecting deep surface resection margins. This limitation is particularly relevant considering that inadequate surgical margins often occur at the deeper, muscle-adjacent planes. Furthermore, the requirement for early administration of antibody-based fluorescent tracers targeting EGFR and VEGF-α—necessitating an additional hospital visit—highlights the need for more logistically feasible fluorescent tracers in HNC and beyond, to alleviate patient inconvenience and improve clinical workflow. Finally, our findings indicate negligible expressions of c-MET, CEA and EpCAM in tumour samples, suggesting they are unsuitable as FLI targets in OSCC and LSCC. Indeed, while CEA expression has not been studied in HNC before, CEA-targeted fluorescent tracer SGM-101 is being evaluated in two phase III trials of FLI in colorectal cancer, indicating the utility of targeting CEA in other cancers . Contrarily, preliminary and preclinical studies challenge our findings on c-MET and EpCAM in OSCC . Notably, c-MET-targeted FLI, demonstrated with EMI-137 and cMBP-ICG, alongside immunohistochemical confirmation of c-MET overexpression in OSCC, showed high sensitivity and specificity in oral cancer specimens . Similarly, an anti-EpCAM tracer conjugated with IRDye800CW efficiently delineated tumours in HNC models . These findings underscore an opportunity for re-evaluating c-MET and EpCAM as feasible FLI targets in OSCC and potentially other HNCs. For thyroid cancer, EGFR and VEGF-α are moderately expressed in PTCs and significantly overexpressed when compared to healthy tissue, making them viable candidates for NIR FLI targeting. Similarly, c-MET, though expressed at lower levels in PTC, exhibited significant overexpression compared to healthy tissue. This aligns with recent studies on the effectiveness of FLI using MET-receptor targeted EMI-137 in enhancing the detection of multifocal PTC . Conversely, the high expression of EpCAM was noteworthy, but not significantly higher than normal thyroid tissue, limiting its utility for tumour detection and delineation. Although expression of integrin αvβ6 was significantly higher in PTC than in healthy thyroid tissue, TIS scores were absent to low, yielding no practical benefits. Expression of CEA was negligible in most PTC tissues. Research on NIR FLI for the thyroid has mainly focused on using ICG or autofluorescence for parathyroid gland identification, aiming to minimize postoperative hypoparathyroidism risks . The necessity for clear surgical margins in PTC is less critical due to the ability of PTC cells to absorb iodine, allowing for post-surgery radioactive iodine (RAI) ablation to target residual disease . Yet, with half of metastatic thyroid cancers becoming RAI-refractory, there's an emerging need for novel, iodine-independent targeting methods to ensure complete removal of malignant tissue. In PTC, the goal is to achieve optimal tumour resection while preserving vital structures and function to maintain quality of life, given the generally favourable prognosis. FLI can aid in this balance by providing precise tumour delineation, thus helping to avoid unnecessary tissue removal and associated complications. Although exploring NIR FLI for PTC surgery margin assessment is in its infancy, its proven effectiveness in preserving parathyroid glands suggests wider applicability . This underscores the need for further investigation into PTC-specific FLI targets to improve tumour delineation. It is important to note that IHC is typically used to assess overall tumour expression for therapeutic potential—such as using cetuximab in chemotherapy—rather than imaging utility. For delineating resection margins through FLI, however, the presence of the target at the invasive front of the tumour is key. In theory, a target that might have relatively lower overall expression of the tumour with high expression at the invasive front could be less suited for therapeutic applications, yet could still be very valuable for intraoperative FLI. Although no distinct tumour-staining patterns were observed for the markers in our study, future assessments or marker suitability for FLI should continue to include this aspect in their evaluations. Additionally, our scoring method, which is based on the total staining pattern, includes both intracellular and membranous staining. This could lead to an over-representation of the portion of membranous receptors available for binding, affecting the suitability of certain markers for receptor-targeted imaging. Future studies should consider distinguishing between membranous and intracellular receptor expression to better evaluate each marker's potential for FLI. Moreover, the major limitation of our study is the size of the patient cohort, with 20 OSCC, ten LSCC and ten PTC cases included. This relatively small cohort size, primarily constrained by sample availability, limits the generalizability of our findings and the precision of biomarker expression estimates. Our selection of OSCC, LSCC and PTC was driven by their prevalence and the clinical challenges they present in achieving clear surgical margins due to their anatomical complexity and proximity to vital structures. These specific tumour types also represent a range of tissue types (squamous cell and glandular), providing broader insights into the applicability of FLI across different HNCs. Our study differs from existing clinical studies by including this diverse selection of tumour types, evaluating some of the same markers in different and additional tumour types, and exploring markers that have been limitedly tested or not tested at all in HNCs. This comprehensive approach provides a deeper understanding of the potential and applicability of these markers across various HNCs. Furthermore, it should be noted that seven out of ten LSCC patients had received radiotherapy, and one out of ten PTC patients had received chemotherapy. None of the OSCC patients had received pre-operative adjuvant therapy. Although we did not specifically analyze the impact of these treatments on target expression, our initial observations did not indicate any apparent effect. Future studies should consider these variables to comprehensively assess their potential impact on biomarker expression. Based on our findings, integrin αvβ6, EGFR, c-MET and VEGF-α hold promise for clinical studies with tumour-specific FLI tracers in HNC, depending on the specific tumour type. While the ideal scenario would be to identify a universal marker suitable for a wide variety of tumours in general, our findings suggest a present necessity to utilize a range of different targets for FLI, given that a singular, universal target remains elusive. The efficacy of tracer binding to these markers is influenced by factors such as pharmacokinetics, target expression in non-tumour tissues, tracer size, dose and timing . Additionally, clinical fluorescent tracers may bind to different epitopes than those stained in our study, underscoring the need for clinical evaluation of tracer binding under varied conditions. Our study underscores the potential of NIR FLI in HNC surgery, particularly highlighting integrin αvβ6 and EGFR as notable targets for OSCC and LSCC, and EGFR, c-MET and VEGF-α for PTC, paving the way for further research into molecular imaging tracers to enhance surgical outcomes. Our results identify integrin αvβ6 and EGFR as key overexpressed markers in OSCC and LSCC, suggesting their suitability as targets for FLI to improve tumour resection precision. Integrin αvβ6, in particular, offers a strong contrast for tumour margin delineation in OSCC and LSCC. In PTC, EGFR, c-MET and VEGF-α showed notable overexpression compared to normal tissue, making them viable candidates for FLI targets. These findings are instrumental for guiding future clinical trials in the development of tumour-specific FLI tracers, aiming to enhance surgical precision and outcomes in HNC. Below is the link to the electronic supplementary material. Supplementary file1 (PDF 604 KB) |
Physician-to-Physician eConsultations to Ophthalmologists at an Academic Medical Center | 17c67216-3d50-4c76-ab90-d4dbe4741b67 | 11463702 | Ophthalmology[mh] | Electronic consultations (eConsults) are asynchronous, provider-to-provider exchanges that occur within shared electronic medical record (EMR) systems or secure online platforms. In eConsults, referring providers submit a clinical question via an EMR, and consulting providers review that question, along with available information in the patient's medical record, and document clinical impressions and recommendations in a clinical note; there is no direct consultant/patient interaction. This asynchronous method of telecommunication confers a range of advantages to patients, providers, and healthcare systems, including increased timely access to specialist consultation, decreased resource waste, improved care coordination, greater patient and provider satisfaction, and cost savings. – Through eConsults, providers can determine whether patients require in-person evaluation, which can decrease the number of unnecessary referrals. When patient referral is deemed beneficial, eConsults can streamline the process by identifying the appropriate subspecialty for referral. , Furthermore, eConsults can determine necessary pre-visit diagnostic workup or imaging, thus optimizing clinical evaluation during the visit and reducing further delays in patient care. , In addition to care coordination, eConsults can ensure that subspecialty evaluation is achieved. The eConsults can bypass many of the hurdles associated with completing specialist referrals, increase physician follow-up and awareness of referral completion, and remove the burden of seeking specialty care from patients. By improving the efficiency and effectiveness of interprofessional clinical exchange, eConsults can also improve access to and quality of care. , However, eConsults also have drawbacks. Namely, there are risks of diagnostic errors and concerns for patient safety given the lack of an in-person examination. Ophthalmology, like many other medical specialties, has increased use of telemedical modalities including eConsults during the COVID-19 pandemic as a means of continuing access to specialty care while limiting in-person exposures. , Many subspecialties of medicine have reported on the benefits of eConsults. , Despite the many known advantages of eConsults, the role of eConsults in ophthalmology has yet to be explored. In this study, we describe the use of ophthalmology eConsults in an academic medical center. We report specifically on the diagnostic accuracy, outcomes, and response timeliness of eConsults, as well as the different ophthalmic conditions inquired about through eConsults. In doing so, we characterize the types of clinical questions asked and identify the learning needs of nonophthalmic providers. Finally, we assess the utility of eConsults to safely and comprehensively manage nonurgent ophthalmic conditions remotely. Unlike other literature published on various modalities of ophthalmic telehealth, this study is the first to describe the feasibility and potential impact of eConsults in ophthalmology. Study Design We conducted a retrospective cohort study of all eConsults submitted to the Ophthalmology Department of Massachusetts Eye and Ear (MEE) from February 11, 2019 through August 18, 2021. The eConsult Program Massachusetts General Brigham (MGB) is a tertiary academic health care system with multiple inpatient and ambulatory health centers throughout the state of Massachusetts. The MGB eConsult program was established in 2013 and expanded to include ophthalmology eConsults in 2019. This program enables providers across all specialties within the MGB healthcare system to submit a clinical question to an ophthalmologist through a shared EMR system, and the ophthalmologist can respond via the EMR without evaluating the patient in person. The eConsult is requested through an EMR order and consists of three fields for referring providers to complete: (1) the reason for eConsult, (2) specific patient care questions, and (3) any additional comments, pertinent patient history, and attachments. Clinical images taken by the referring provider can be uploaded into the EMR. In this study, the term “referring provider” refers to a non-ophthalmology healthcare provider who submits an eConsult to an ophthalmologist on behalf of a patient. The terms “consulting provider” and “eConsultant” are used interchangeably and refer to the ophthalmologist receiving and reviewing the eConsult. The term “referral request” refers to the act of placing an electronic order in the EMR to refer a patient for an in-person evaluation by an ophthalmologist. All ophthalmology eConsults were assigned to and addressed by a comprehensive ophthalmologist, who additionally reviewed the patient's medical history, ophthalmic history, recent progress notes, associated images, and diagnostic tests. The general ophthalmologist asked for clinical advice from appropriate ophthalmology subspecialists to answer eConsult questions as needed. Clinical recommendations were returned to the referring provider through an “eConsult note” within the EMR. The referring provider then determined next steps, including relaying the eConsultant's recommendations to the patient, scheduling appropriate follow-up, and determining which aspects of the eConsultant recommendations act upon. A workflow diagram of the eConsult process is included in . There was no direct ophthalmologist-to-patient exchange, and all eConsults included standardized language indicating that ongoing management of the patient's clinical problem is the responsibility of the referring provider and other members of the patient's care team, that eConsults are conducted in the absence of an examination or direct conversation with the patient, and that recommendations made are solely based on the information provided by the referring provider in the eConsult and information available in the EMR. The eConsult program is funded internally by MGB and does not involve charges to patient insurance; rather, MGB pays eConsultants a standard fixed fee for their consulting service. Data Extraction and Analysis All MEE ophthalmology eConsults were retrospectively reviewed through manual review of eConsult encounters in the EMR system, including the reason for eConsult, the date and time an eConsult order was placed and subsequently responded to, the presence of external images of patient eyes, results of diagnostic and screening tests (e.g., fundus photography, newborn vision screens, computed tomography, magnetic resonance imaging, radiography), eConsultant diagnosis, and clinical recommendation . Patient demographic data, including patient age at time of eConsult, sex, race, ethnicity, and insurance coverage status, were collected. Demographic data of referring providers (medical specialty, academic degree, and location) were also collected . Whether patients had known past ocular history or an established ophthalmologist, defined as any visit with an ophthalmologist in the five years before eConsult, was also recorded. Each eConsult was classified into clinically meaningful diagnostic categories based on the literature. The eConsult questions underwent thematic review to identify the types of clinical questions asked by providers and each question was assigned a category type. Seven question types were identified, including issues of diagnosis, treatment, management, triage/referral, workup, risk assessment and timing of routine screenings, and medication management . The eConsults were assessed for whether follow-up with an ophthalmologist was recommended, which party (eConsultant or referring provider) was responsible for placing the referral, whether the referral was placed, and whether a follow-up appointment occurred. Patient charts were reviewed up to six-months following the initial eConsult to assess for subsequent emergency department (ED) encounters associated with the eConsult concern, evidence of eConsult recommendations being relayed to the patient, and patient presentation to an ophthalmologist for in-person evaluation within the MGB system. Clinical details of subsequent in-person ophthalmology visits were assessed, including the diagnoses made during in-person ophthalmology visits which were assessed for concordance with any diagnoses made during eConsults. Institutional level data on ICD-10 codes for patient visits to the MEE ophthalmology-specific ED during the period of this study were obtained through automated EMR reports. This data was used to identify the most common eye diagnoses presenting to the MEE ED, which were used as a comparison for the eConsult eye concerns. Descriptive statistics and chi-square analyses were executed using R version 4.1.0 (R Foundation for Statistical Computing). A P value of α <0.05 was considered to be statistically significant. Institutional Review Board This study protocol was reviewed by the MGB Institutional Review Board and determined to be a Quality Improvement study. Therefore approval was not required. Data Availability The data used in this study is available on request. We conducted a retrospective cohort study of all eConsults submitted to the Ophthalmology Department of Massachusetts Eye and Ear (MEE) from February 11, 2019 through August 18, 2021. Massachusetts General Brigham (MGB) is a tertiary academic health care system with multiple inpatient and ambulatory health centers throughout the state of Massachusetts. The MGB eConsult program was established in 2013 and expanded to include ophthalmology eConsults in 2019. This program enables providers across all specialties within the MGB healthcare system to submit a clinical question to an ophthalmologist through a shared EMR system, and the ophthalmologist can respond via the EMR without evaluating the patient in person. The eConsult is requested through an EMR order and consists of three fields for referring providers to complete: (1) the reason for eConsult, (2) specific patient care questions, and (3) any additional comments, pertinent patient history, and attachments. Clinical images taken by the referring provider can be uploaded into the EMR. In this study, the term “referring provider” refers to a non-ophthalmology healthcare provider who submits an eConsult to an ophthalmologist on behalf of a patient. The terms “consulting provider” and “eConsultant” are used interchangeably and refer to the ophthalmologist receiving and reviewing the eConsult. The term “referral request” refers to the act of placing an electronic order in the EMR to refer a patient for an in-person evaluation by an ophthalmologist. All ophthalmology eConsults were assigned to and addressed by a comprehensive ophthalmologist, who additionally reviewed the patient's medical history, ophthalmic history, recent progress notes, associated images, and diagnostic tests. The general ophthalmologist asked for clinical advice from appropriate ophthalmology subspecialists to answer eConsult questions as needed. Clinical recommendations were returned to the referring provider through an “eConsult note” within the EMR. The referring provider then determined next steps, including relaying the eConsultant's recommendations to the patient, scheduling appropriate follow-up, and determining which aspects of the eConsultant recommendations act upon. A workflow diagram of the eConsult process is included in . There was no direct ophthalmologist-to-patient exchange, and all eConsults included standardized language indicating that ongoing management of the patient's clinical problem is the responsibility of the referring provider and other members of the patient's care team, that eConsults are conducted in the absence of an examination or direct conversation with the patient, and that recommendations made are solely based on the information provided by the referring provider in the eConsult and information available in the EMR. The eConsult program is funded internally by MGB and does not involve charges to patient insurance; rather, MGB pays eConsultants a standard fixed fee for their consulting service. All MEE ophthalmology eConsults were retrospectively reviewed through manual review of eConsult encounters in the EMR system, including the reason for eConsult, the date and time an eConsult order was placed and subsequently responded to, the presence of external images of patient eyes, results of diagnostic and screening tests (e.g., fundus photography, newborn vision screens, computed tomography, magnetic resonance imaging, radiography), eConsultant diagnosis, and clinical recommendation . Patient demographic data, including patient age at time of eConsult, sex, race, ethnicity, and insurance coverage status, were collected. Demographic data of referring providers (medical specialty, academic degree, and location) were also collected . Whether patients had known past ocular history or an established ophthalmologist, defined as any visit with an ophthalmologist in the five years before eConsult, was also recorded. Each eConsult was classified into clinically meaningful diagnostic categories based on the literature. The eConsult questions underwent thematic review to identify the types of clinical questions asked by providers and each question was assigned a category type. Seven question types were identified, including issues of diagnosis, treatment, management, triage/referral, workup, risk assessment and timing of routine screenings, and medication management . The eConsults were assessed for whether follow-up with an ophthalmologist was recommended, which party (eConsultant or referring provider) was responsible for placing the referral, whether the referral was placed, and whether a follow-up appointment occurred. Patient charts were reviewed up to six-months following the initial eConsult to assess for subsequent emergency department (ED) encounters associated with the eConsult concern, evidence of eConsult recommendations being relayed to the patient, and patient presentation to an ophthalmologist for in-person evaluation within the MGB system. Clinical details of subsequent in-person ophthalmology visits were assessed, including the diagnoses made during in-person ophthalmology visits which were assessed for concordance with any diagnoses made during eConsults. Institutional level data on ICD-10 codes for patient visits to the MEE ophthalmology-specific ED during the period of this study were obtained through automated EMR reports. This data was used to identify the most common eye diagnoses presenting to the MEE ED, which were used as a comparison for the eConsult eye concerns. Descriptive statistics and chi-square analyses were executed using R version 4.1.0 (R Foundation for Statistical Computing). A P value of α <0.05 was considered to be statistically significant. This study protocol was reviewed by the MGB Institutional Review Board and determined to be a Quality Improvement study. Therefore approval was not required. The data used in this study is available on request. A total of 100 pediatric and adult ophthalmic eConsults were ordered and completed between February 11, 2019, and August 18, 2021. Ophthalmic eConsults were most frequently ordered by internal medicine providers (67%), followed by family medicine providers (13%), pediatricians (9%), medicine-pediatrics providers (5%), and six other types of specialists . Nearly one quarter (24%) of eConsults asked two or more clinical questions. Regarding ophthalmic diagnoses, hordeola and chalazia accounted for 14% (14) of eConsults, with a total of 45 different ophthalmic diagnoses being asked about . All eConsults were considered nonurgent. Of the 100 eConsults ordered, 52% included a picture, and 8% included diagnostic imaging from previous visits. The average response time and standard deviation (SD) from when the eConsult order was placed to the time the consultation was completed was 1.6 days (SD ±1.9). The average number of days from eConsult response to in-person follow-up was 28.9 (SD ± 27.4) days. Of the 100 eConsults, in-person evaluation was recommended in 62% (n = 62) of cases; roughly 13% of these were due to medication follow-up (e.g., initiation of ophthalmic steroid drops). For patients requiring referral for in person evaluation by an ophthalmic subspecialty, MEE staff volunteered to submit the referral request for 41.9% (n = 26) of patients to ensure rapid referral placement and patient follow-up; referring providers were responsible for submitting a referral request for the remaining 58.1% (n = 36) of patients. However, the rate of actually submitting those referrals varied where MEE staff submitted the referral request 76.9% (n = 20/26) of the time whereas the referring provider submitted the referral request 66.7% (n = 24/36) of the time. No statistically significant difference was found between referring providers and eConsultants in the likelihood of placing the referral request ( P = 0.3799). Of the 62 patients recommended for in-person evaluation, 48.4% (n = 30) presented to an ophthalmologist for evaluation. Agreement in diagnostic concordance between the eConsultant clinical diagnosis and in-person diagnosis occurred in 93.3% (n = 28) of cases. Two cases of non-concordant diagnoses were documented (6.9%). The first was a case of light sensitivity and blurred vision secondary to central retinal vein occlusion initially attributed to post-concussion syndrome; the second was a case of orbital fat prolapse initially thought to be a hordeola. Referring providers proceeded to follow eConsult recommendations in 91% (n = 91) of cases. Evidence that the referring provider relayed the ophthalmologist plan to the patient was documented in 75% (n = 75) of cases based on documentation anywhere in the EMR (e.g., visit note, telephone encounter). Only 5% (n = 5) of patients presented to any MGB ED for issues related to the eConsult after the eConsult was placed. Seventy-nine percent of patients presented to any MGB clinic within the six months after the eConsult request, which indicated they were not lost to follow-up in the system. Before the time of eConsult placement, the vast majority of patients (73%, n = 73) had never seen an ophthalmologist or had any documented visit from an ophthalmologist accessible in the EMR and had no known ocular history (62%, n = 62) . Patient demographics including age, race/ethnicity, and sex are presented in . In this study we found that eConsults in ophthalmology provide for high diagnostic accuracy, are useful across a range of clinical diagnoses and concerns, and result in timely responses. Additionally, this study thematically describes eConsult questions and topic areas. Although eConsults have not previously been reported within the field of ophthalmology, this study supports the efficacy and feasibility of eConsults to provide asynchronous specialty advice to non-ophthalmic providers. , Clinical Outcomes The results of this study demonstrate the feasibility of eConsults to accurately diagnose nonurgent ophthalmic conditions. Of the patients recommended for an in-person evaluation with an ophthalmologist and subsequently followed up, concordance between the clinical assessment of the eConsultant and in-person ophthalmologist occurred in 93.1% of cases, with only two cases of missed diagnoses identified (central retinal vein occlusion and orbital fat prolapse as described in the results section). The first of the two cases of missed diagnosis deserves special attention. This case was a patient who presented two months after a motor vehicle accident that resulted in a concussion with light sensitivity and blurred vision. The patient’s symptoms were initially thought to be related to post-concussion syndrome; however, the eConsultant recommended the patient be evaluated to rule out other serious and vision-threatening etiologies. This patient was given an in-person appointment within three weeks of the eConsult response, during which this patient was found to have multiple peripheral retinal hemorrhages and elevated intraocular pressure and was referred for evaluation of a central retinal vein occlusion by the retina service. This scenario highlights the importance of referral for in-person evaluation to fully evaluate concerning clinical presentations and to rule out serious pathology. The growing practice of eConsultations will help build a body of evidence that can be leveraged to provide guidance in the future on such cases. The rate of diagnostic accuracy found in this study is similar to prior reports of high diagnostic accuracy when eConsults were used in other subspecialties such as dermatology. This is encouraging because access to ophthalmic care has become increasingly challenging given the growing shortage of ophthalmologists, persistent geographic barriers to ophthalmic care, and widening socioeconomic inequities in accessing specialty care. – Ophthalmology Referrals to Subspecialty Services The eConsults allow prompt ophthalmology evaluation to be achieved for patients presenting with ocular concerns in non-ophthalmic ambulatory care settings. Rates of completion for referrals to in-person specialty care vary tremendously from 30% to 80% across specialties, making eConsults an attractive alternative. , Among ophthalmology referrals for in-person evaluation specifically, reported rates of completion are similarly variable, ranging from 5% to 75% within the recommended referral period. , Patients also report that the cost of visits, insurance status, distance to clinics, lack of transportation, language barriers and limited health literacy, and work schedule conflicts lead to delayed or complete inaccessibility to ophthalmic care. , , The eConsults address many of these barriers through the elimination of time, cost, or travel to achieve an initial subspecialty evaluation. Additionally, eConsults create a direct line of communication among all members of the patient’s care team, delivering well-coordinated, high-quality patient-centered care. In the present study, 100% of submitted eConsults were completed, and 75% of cases had evidence that the ophthalmologist's recommendations were communicated to the patient. This reflects the suitability of eConsults as a mechanism to ensure patient access to ophthalmic care. Timing of eConsults and Subsequent Evaluation Importantly, eConsults have a demonstrated benefit of decreasing excessive wait times to reach specialty care. , , In the literature, the estimated median time from when a specialist referral is placed to the completion of an in-person visit is 7.5 to 8.7 weeks. , In our study, the average response time to complete an eConsult was 1.6 days. This is consistent with the results of other eConsult programs, including dermatology, allergy/immunology, endocrinology, and rheumatology, where eConsult were completed within three days. , – This is an especially notable benefit of eConsults because lengthy waiting periods are a common deterrent to seeking follow-up care, and the wait times for specialty appointments are steadily increasing. , The eConsults also improve referral quality, ensuring that patients requiring in-person evaluation are scheduled with the appropriate subspecialist. , In our study, 50% of subsequent referrals were made to ophthalmology subspecialties, including retina, pediatrics, neuro-ophthalmology, and oculoplastics. The eConsult program thus bypassed the standard practice of an initial in-person evaluation by a comprehensive ophthalmologist prior to receiving a subspecialist appointment. Of the many benefits that can be inferred from this, key among them is expedition of time to care, decreased unnecessary patient visits, and associated patient cost savings. The ability of eConsults to decrease unnecessary specialist referrals and potential economic advantages to the healthcare system merits further investigation in future studies. Advantages of Ophthalmology eConsults The eConsults also confer potential advantages across other care settings. An average of two million eye-related ED visits occur annually, with nearly half of those visits being for nonurgent conditions that can safely be managed in the outpatient setting. , EDs across the country are poorly positioned to provide optimal ophthalmic care because of inadequate or nonexistent ophthalmology service coverage and insufficient ophthalmic training for ED providers—limitations that are exacerbated in rural and underserved populations. , These limitations have also been associated with ED provider discomfort and inaccuracy evaluating ocular concerns. At the national level, hordeola are the second-most common cause of nonurgent ocular presentation to an ED. At the institutional level, 6% of all ophthalmic related presentations to the MEE ED (approximately 935 visits annually) from May 2019 to June 2021 had a primary diagnosis of hordeola and chalazia. In our study, hordeola and chalazia were the most common ophthalmic conditions inquired about in eConsults, accounting for 14% of the submitted eConsults. All eConsults related to hordeola and chalazia were managed conservatively with symptom improvement or resolution documented in patients’ charts, except in the case of a missed diagnosis described above. None of these eConsults were followed by an ED presentation, which suggests that nonurgent ocular conditions can safely and accurately be diagnosed and managed remotely and are a potential source of avoidable ED visits. Evidence suggests that the cost of managing nonurgent ophthalmic conditions in the ED can be anywhere from two to four times higher than in an ambulatory care setting. , Ophthalmic eConsults can help patients and healthcare systems avoid unnecessary costs for nonurgent, non-vision-threatening ophthalmic issues. Additionally, reduction in nonurgent ocular ED visits can help decrease overall ED crowding and allow resources to be directed toward patients with urgent ophthalmic and medical conditions. The eConsult offers educational advantages to providers and medical educators. – By communicating a plan of care back to the referring provider, eConsults create an opportunity for immediate feedback and education of the referring provider as it relates to ophthalmic issues. , This is in contrast to the traditional referral process where providers may not be aware of the specialist recommendation or plan of care for prolonged periods of time. Furthermore, providers will be able to apply information learned in the eConsult to similar cases in the future. Understanding the types of clinical problems and content areas providers most commonly ask questions about is necessary to better guide educational efforts. For example, this study suggests that further education on hordeolum/chalazion identification and management may be beneficial to physicians in ambulatory care settings. Limitations There are a few notable limitations in the use of eConsults in our study. First, the present study included only 100 eConsults over a 30-month period submitted by 74 individual providers, which suggests that many of the nearly 1400 primary care providers in the MGB system were unaware of or did not use the service. Understanding attitudes toward ophthalmic eConsult practices and seeking a broader sample size would be advantageous to inform widespread implementation and adoption of such programs. Second, the role of eConsults in decreasing or exacerbating disparities in ophthalmic care has yet to be established. An understanding of how eConsults can be used to decrease disparities in the provision of healthcare would be useful. Another important limitation is the retrospective nature of this study, because follow-up of patient outcomes was limited by chart review and documentation. Additionally, the percentage of patients who presented to an ophthalmologist after receiving a referral for an in-person evaluation was low, thus limiting our ability to assess the true rates of diagnostic concordance. Furthermore, all eConsults were answered by a single ophthalmologist at a single academic institution. Patients who sought subsequent eye care outside of this institution were unidentifiable in this study, because they would not be identifiable in the institutional EMR. Future research should evaluate the use of eConsults in various practice sites and locations, which would improve the generalizability of the findings. This study did not collect survey data on patient perspectives regarding confidence in, comfort with, or perceived benefits of eConsults. As a result, this study is unable to comment on patient perspectives of the ophthalmology eConsult program and the potential impact it could have on seeking care elsewhere or delaying care if symptoms progress. Future studies should consider a patient survey component. Last, this study focuses solely on eConsults and is therefore unable to compare the advantages and limitations of eConsults to other synchronous and asynchronous forms of telehealth. In this study, ophthalmic eConsults were associated with timely response back to the referring provider, and with a high rate of diagnostic accuracy among a subset of patients subsequently seen for an in-person ophthalmology visit. Our results support the use of eConsults as an effective telehealth modality to obtain timely diagnosis, access, and management of nonurgent eye conditions. Our study also demonstrates that eConsults enhance patient quality of care by soliciting specialist input, ensuring that timely ophthalmology evaluation is achieved, coordinating appropriate referral management and triage, and collaborating across interdisciplinary care teams. The results of this study demonstrate the feasibility of eConsults to accurately diagnose nonurgent ophthalmic conditions. Of the patients recommended for an in-person evaluation with an ophthalmologist and subsequently followed up, concordance between the clinical assessment of the eConsultant and in-person ophthalmologist occurred in 93.1% of cases, with only two cases of missed diagnoses identified (central retinal vein occlusion and orbital fat prolapse as described in the results section). The first of the two cases of missed diagnosis deserves special attention. This case was a patient who presented two months after a motor vehicle accident that resulted in a concussion with light sensitivity and blurred vision. The patient’s symptoms were initially thought to be related to post-concussion syndrome; however, the eConsultant recommended the patient be evaluated to rule out other serious and vision-threatening etiologies. This patient was given an in-person appointment within three weeks of the eConsult response, during which this patient was found to have multiple peripheral retinal hemorrhages and elevated intraocular pressure and was referred for evaluation of a central retinal vein occlusion by the retina service. This scenario highlights the importance of referral for in-person evaluation to fully evaluate concerning clinical presentations and to rule out serious pathology. The growing practice of eConsultations will help build a body of evidence that can be leveraged to provide guidance in the future on such cases. The rate of diagnostic accuracy found in this study is similar to prior reports of high diagnostic accuracy when eConsults were used in other subspecialties such as dermatology. This is encouraging because access to ophthalmic care has become increasingly challenging given the growing shortage of ophthalmologists, persistent geographic barriers to ophthalmic care, and widening socioeconomic inequities in accessing specialty care. – The eConsults allow prompt ophthalmology evaluation to be achieved for patients presenting with ocular concerns in non-ophthalmic ambulatory care settings. Rates of completion for referrals to in-person specialty care vary tremendously from 30% to 80% across specialties, making eConsults an attractive alternative. , Among ophthalmology referrals for in-person evaluation specifically, reported rates of completion are similarly variable, ranging from 5% to 75% within the recommended referral period. , Patients also report that the cost of visits, insurance status, distance to clinics, lack of transportation, language barriers and limited health literacy, and work schedule conflicts lead to delayed or complete inaccessibility to ophthalmic care. , , The eConsults address many of these barriers through the elimination of time, cost, or travel to achieve an initial subspecialty evaluation. Additionally, eConsults create a direct line of communication among all members of the patient’s care team, delivering well-coordinated, high-quality patient-centered care. In the present study, 100% of submitted eConsults were completed, and 75% of cases had evidence that the ophthalmologist's recommendations were communicated to the patient. This reflects the suitability of eConsults as a mechanism to ensure patient access to ophthalmic care. Importantly, eConsults have a demonstrated benefit of decreasing excessive wait times to reach specialty care. , , In the literature, the estimated median time from when a specialist referral is placed to the completion of an in-person visit is 7.5 to 8.7 weeks. , In our study, the average response time to complete an eConsult was 1.6 days. This is consistent with the results of other eConsult programs, including dermatology, allergy/immunology, endocrinology, and rheumatology, where eConsult were completed within three days. , – This is an especially notable benefit of eConsults because lengthy waiting periods are a common deterrent to seeking follow-up care, and the wait times for specialty appointments are steadily increasing. , The eConsults also improve referral quality, ensuring that patients requiring in-person evaluation are scheduled with the appropriate subspecialist. , In our study, 50% of subsequent referrals were made to ophthalmology subspecialties, including retina, pediatrics, neuro-ophthalmology, and oculoplastics. The eConsult program thus bypassed the standard practice of an initial in-person evaluation by a comprehensive ophthalmologist prior to receiving a subspecialist appointment. Of the many benefits that can be inferred from this, key among them is expedition of time to care, decreased unnecessary patient visits, and associated patient cost savings. The ability of eConsults to decrease unnecessary specialist referrals and potential economic advantages to the healthcare system merits further investigation in future studies. The eConsults also confer potential advantages across other care settings. An average of two million eye-related ED visits occur annually, with nearly half of those visits being for nonurgent conditions that can safely be managed in the outpatient setting. , EDs across the country are poorly positioned to provide optimal ophthalmic care because of inadequate or nonexistent ophthalmology service coverage and insufficient ophthalmic training for ED providers—limitations that are exacerbated in rural and underserved populations. , These limitations have also been associated with ED provider discomfort and inaccuracy evaluating ocular concerns. At the national level, hordeola are the second-most common cause of nonurgent ocular presentation to an ED. At the institutional level, 6% of all ophthalmic related presentations to the MEE ED (approximately 935 visits annually) from May 2019 to June 2021 had a primary diagnosis of hordeola and chalazia. In our study, hordeola and chalazia were the most common ophthalmic conditions inquired about in eConsults, accounting for 14% of the submitted eConsults. All eConsults related to hordeola and chalazia were managed conservatively with symptom improvement or resolution documented in patients’ charts, except in the case of a missed diagnosis described above. None of these eConsults were followed by an ED presentation, which suggests that nonurgent ocular conditions can safely and accurately be diagnosed and managed remotely and are a potential source of avoidable ED visits. Evidence suggests that the cost of managing nonurgent ophthalmic conditions in the ED can be anywhere from two to four times higher than in an ambulatory care setting. , Ophthalmic eConsults can help patients and healthcare systems avoid unnecessary costs for nonurgent, non-vision-threatening ophthalmic issues. Additionally, reduction in nonurgent ocular ED visits can help decrease overall ED crowding and allow resources to be directed toward patients with urgent ophthalmic and medical conditions. The eConsult offers educational advantages to providers and medical educators. – By communicating a plan of care back to the referring provider, eConsults create an opportunity for immediate feedback and education of the referring provider as it relates to ophthalmic issues. , This is in contrast to the traditional referral process where providers may not be aware of the specialist recommendation or plan of care for prolonged periods of time. Furthermore, providers will be able to apply information learned in the eConsult to similar cases in the future. Understanding the types of clinical problems and content areas providers most commonly ask questions about is necessary to better guide educational efforts. For example, this study suggests that further education on hordeolum/chalazion identification and management may be beneficial to physicians in ambulatory care settings. There are a few notable limitations in the use of eConsults in our study. First, the present study included only 100 eConsults over a 30-month period submitted by 74 individual providers, which suggests that many of the nearly 1400 primary care providers in the MGB system were unaware of or did not use the service. Understanding attitudes toward ophthalmic eConsult practices and seeking a broader sample size would be advantageous to inform widespread implementation and adoption of such programs. Second, the role of eConsults in decreasing or exacerbating disparities in ophthalmic care has yet to be established. An understanding of how eConsults can be used to decrease disparities in the provision of healthcare would be useful. Another important limitation is the retrospective nature of this study, because follow-up of patient outcomes was limited by chart review and documentation. Additionally, the percentage of patients who presented to an ophthalmologist after receiving a referral for an in-person evaluation was low, thus limiting our ability to assess the true rates of diagnostic concordance. Furthermore, all eConsults were answered by a single ophthalmologist at a single academic institution. Patients who sought subsequent eye care outside of this institution were unidentifiable in this study, because they would not be identifiable in the institutional EMR. Future research should evaluate the use of eConsults in various practice sites and locations, which would improve the generalizability of the findings. This study did not collect survey data on patient perspectives regarding confidence in, comfort with, or perceived benefits of eConsults. As a result, this study is unable to comment on patient perspectives of the ophthalmology eConsult program and the potential impact it could have on seeking care elsewhere or delaying care if symptoms progress. Future studies should consider a patient survey component. Last, this study focuses solely on eConsults and is therefore unable to compare the advantages and limitations of eConsults to other synchronous and asynchronous forms of telehealth. In this study, ophthalmic eConsults were associated with timely response back to the referring provider, and with a high rate of diagnostic accuracy among a subset of patients subsequently seen for an in-person ophthalmology visit. Our results support the use of eConsults as an effective telehealth modality to obtain timely diagnosis, access, and management of nonurgent eye conditions. Our study also demonstrates that eConsults enhance patient quality of care by soliciting specialist input, ensuring that timely ophthalmology evaluation is achieved, coordinating appropriate referral management and triage, and collaborating across interdisciplinary care teams. |
Association between left atrial function and pulmonary vein stump thrombus after left upper lobectomy: insights from cine-MRI | c4f6c4ba-efd8-4a39-a375-c0f0ae6ad12d | 11850599 | Surgical Procedures, Operative[mh] | Pulmonary vein stump thrombus (PVST) has been increasingly recognized as a frequent complication following surgical treatment for lung cancer, particularly after left upper lobectomy (LUL) – . The occurrence rate of PVST following LUL surgery ranges from 13.5 to 34.0% , , . This complication can lead to embolism in various organs, including the brain , , spleen , and kidney . Therefore, it is vitally important to understand the mechanism that underlies PVST after LUL. Although the exact pathogenesis of PVST is unclear, several factors potentially contribute to its development. Some recent studies have suggested that blood stasis and turbulent blood flow near the pulmonary vein (PV) stump may contribute to the development of thrombus – . However, it is clinically challenging to assess the state of blood flow in the left atrium (LA) and the PV stump after LUL. The LA appendage, the most common site of LA thrombus formation—particularly in patients with atrial fibrillation—shares structural similarities with the post-LUL PV stump, suggesting that its pathogenesis may similarly apply to PVST. LA enlargement has been linked to appendage dysfunction and thrombus formation , , thereby elevating the risk of stroke and systemic embolization , . Although LA size and function are routinely assessed by transthoracic echocardiography (TTE), cardiac MR offers more reliable evaluations , . We hypothesized that LA size and function would be associated with PVST after LUL. The purpose of this study was to clarify the association between LA function as evaluated by cine-magnetic resonance imaging (cine-MRI) and the development of PVST after LUL. The novelty of this study lies in using cine-MRI to comprehensively evaluate LA function and clarify its relationship with PVST formation after LUL. By leveraging cine MRI’s superior structural and functional assessment capabilities over conventional echocardiography, our approach offers unique insights into the hemodynamic mechanisms underlying PVST and may contribute to more effective risk stratification and patient management. Interobserver agreement was excellent for all quantitative measurements (ICC = 0.99 for LAESV, ICC = 0.96 for LAEDV, ICC = 0.99 for LAESVI, ICC = 0.96 for LAEDVI, ICC = 0.85 for LASV, ICC = 0.90 for LAEF, and ICC = 0.98 for length of PV stump). Measured values were significantly greater in patients who developed PVST than in those without PVST in terms of LAESV (74.2 ± 34.3 vs. 58.7 ± 18.6mL, p = 0.009), LAEDV (49.5 ± 30.0 vs. 33.1 ± 16.3 mL, p < 0.001), LAESVI (45.8 ± 19.8 vs. 36.5 ± 10.2 mL/m 2 , p = 0.004), and LAEDVI (30.5 ± 17.6 vs. 20.4 ± 9.6 mL/m 2 , p < 0.001) (Table ). LAEF was significantly lower in patients with PVST (35.0 ± 8.5%) than in those without PVST (45.0 ± 10.1%, p < 0.001). There was no significant difference in LASV between patients with and without development of PVST (24.6 ± 8.2 vs. 25.7 ± 8.3 mL/m 2 , p = 0.627). PV stump length was significantly longer in patients with PVST than in those without PVST (22.9 ± 4.5 vs. 20.6 ± 5.0 mm, p = 0.033). The AUC value for predicting the development of PVST was 0.668 for LAESV (95% confidence interval (CI), 0.562–0.764), 0.769 for LAEDV (95%CI, 0.669–0.851), 0.688 for LAESVI (95%CI, 0.582–0.781), 0.792 for LAEDVI (95%CI, 0.694–0.870), and 0.803 for LAEF (95%CI, 0.706–0.879) (Fig. ). AUC values for LAEDV and LAEDVI, and LAEF were significantly greater than those for LAESV ( p = 0.001, 0.002, and 0.016, respectively) and LAESVI ( p = 0.016, 0.002, and 0.031, respectively). ROC analysis yielded a cut-off value of 66.4 mL for LAESV, 35.3 mL for LAEDV, 35.9 mL/m 2 for LAESVI, 23.3 mL/m 2 for LAEDVI, and 40.2% for LAEF. In the univariate logistic regression analysis, PV stump length (OR: 1.103; 95%CI: 1.003–1.213), LAESV (OR: 1.028; 95%CI: 1.005–1.051), LAEDV (OR: 1.044; 95%CI: 1.013–1.075), LAESVI (OR: 1.054; 95%CI: 1.011–1.099), LAEDVI (OR: 1.082; 95%CI: 1.025–1.141), and LAEF (OR: 0.896; 95%CI: 0.846–0.950) were significantly associated with PVST (Table ). In the multivariate analysis, variables such as LAESV, LAEDV, LAESVI, LAEDVI, and SV exhibited high VIF values (VIF = 1591.596, 1418.090, 342.474, 821.579, and 166.538, respectively) and were therefore excluded from the analysis. In the multivariate analysis with stepwise selection, only LAEF remained significantly associated with PVST (OR: 0.896; 95%CI: 0.846–0.950). Representative cases are shown in Fig. . This is the first report to reveal an association between LA function and development of PVST after LUL. The major finding of the present study is that LA enlargement and dysfunction, as evaluated by cine-MRI, were significantly associated with the development of PVST after LUL. These results suggest that assessing LA function using cine-MRI may be a critical step in identifying patients at an increased risk for development of PVST. Understanding the relationship between LA function and PVST will potentially enable clinicians and surgeons to improve patient outcomes through early intervention and more targeted treatment approaches. Virchow described three critically important factors in the development of venous thrombosis: blood stasis, activation of blood coagulation, and venous intimal damage . It is important to note that activation of blood coagulation and venous intimal damage occur to some degree in all lung surgeries and are not exclusive to LUL. In other words, it is possible that blood stasis might play a significant role in PVST formation after LUL. Certain distinct regional blood flow patterns (blood stasis or turbulent blood flow) around the PV stump have been observed more frequently after LUL than after other lung lobectomies, which could be related to thrombus formation – . Recent studies using 4D flow MRI have suggested that LUL probably induces blood turbulence around the PV stump by complicated blood streams in the LA, and have also suggested that PVST is prone to develop under certain hemodynamic conditions , . Ohtaka et al. found that slow blood flow and the presence of spontaneous echo contrast (SEC) in the left superior PV stump on ultrasonography were associated with PVST . A long PV stump after LUL has been proposed as a significant risk factor for PVST as it can create a procoagulant environment characterized by turbulent flow or blood stasis . The left superior PV stump is reported to be significantly longer than other PV stumps, primarily because it has the longest intrapericardial segment . In our results, PV stump length was significantly longer in patients with thrombus formation than in those without PVST formation, which is consistent with their findings. However, in the multivariate analysis conducted in this study, it did not remain as an independent predictor. As shown in several reports, attempts to prevent PVST after lobectomy by shortening the PV stump have not always been effective, and PVST has still occurred in patients who underwent proximal PV ligation after LUL , . Accordingly, factors other than long PV stump length might cause blood stasis near the PV stump and development of PVST. LA blood stasis is associated with LA enlargement and dysfunction, which have been reported to be associated with an increased incidence of cardiovascular events such as LA appendage thrombus formation, atrial fibrillation, ischemic heart disease, heart failure, stroke, and cardiovascular death , , . Low LAEF has been reported to have a significant association with SEC in LA and with LA appendage thrombus development , , in agreement with the present finding of an association between LAEF and PVST development. SEC in the LA has also been observed in patients with a large LA , . LAESVI and LAEDVI have been reported to be associated with LA appendage thrombus . LAESVI, as an indicator of the largest LA volume, is a known predictor of cardiovascular outcomes and is the recommended measure of LA size . LA enlargement can promote blood stasis, which facilitates thrombus formation in the LA appendage . Similarly, PVST formation after LUL may be affected by blood stasis associated with LA dilation. A previous study has reported that LA volume as measured by preoperative non-ECG gated CT was significantly greater in patients who developed PVST than in those without thrombus . In the present study, LA volume and function were evaluated using ECG-gated cine-MRI, which provides more accurate assessment than conventional non-ECG gated CT and does not require injection of contrast material. LAESVI > 32 mL/m 2 and LA volume index (LAVI) ≥ 34 mL/m 2 are widely recognized risk factors for cardiovascular events , . In the present study, LAESVI > 35.9 mL/m 2 was significantly associated with PVST development, which is similar to those reported values. According to a recent report, LAEDVI is also a strong predictor of cardiovascular events . LAEDVI has been reported to be significantly higher in patients with LA appendage thrombus than in those without thrombus . To address potential biases and confounding factors, we conducted a multivariate analysis that included age, gender, and underlying conditions such as hypertension, diabetes, atrial fibrillation, and medication of anticoagulant or antiplatelet drugs. These factors were selected based on their known influence on LA function and thrombus formation. The analysis demonstrated that LAEF remained the only independent predictor significantly associated with PVST. This finding suggests that the observed relationship between LAEF and PVST is robust and not confounded by these variables. There are some limitations in the present study. First, there were fewer patients with PVST than those without. Second, the present evaluation focused solely on the presence of PVST development on the seventh postoperative day, and long-term assessment was not performed. Third, this study evaluated the LA function using only LAEF. However, this approach does not account for other important aspects of left atrial function, such as conduit function (left atrial passive ejection fraction, LAPEF) and pump function (left atrial active ejection fraction, LAAEF), nor does it include advanced techniques such as strain analysis. Future research should incorporate a more comprehensive assessment of left atrial function using these additional parameters and advanced imaging techniques to provide deeper insights into the relationship between left atrial function and PVST. Finally, we recognize that our study only evaluated the relationship between postoperative LA function and PVST development. From a clinical perspective, it would be highly beneficial to determine patients’ risk of thrombus formation preoperatively. Nevertheless, this is the first study to focus on the relationship between LA function and PVST after LUL, and we believe that our findings offer an important starting point and will encourage future large-scale multicenter, prospective studies that include preoperative LA function assessments, ultimately leading to improved risk stratification and patient management. In conclusion, this study demonstrated a significant association of LA enlargement and LA dysfunction with PVST after LUL, suggesting that LA dysfunction may contribute to the development of PVST. Therefore, assessment of LA function by cine-MRI could be useful for predicting PVST development after LUL. This study was approved and the requirement for informed consent from the study subjects was waived by the institutional review board (Ethics Committee on Epidemiological and its related Studies, Sakuragaoka Campus, Kagoshima University; approval number, revised edition 1 of 220009) due to the retrospective study design. This study was conducted in accordance with the Declaration of Helsinki and Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan. Patients Our institutional ethics review board approved this retrospective study and the requirement for patients’ informed consent was waived. We reviewed the imaging database of our radiology department and patients’ electronic medical records to identify those who had undergone surgical treatment for lung neoplasm as well as a postoperative MR examination between January 2018 and April 2024. Among the 294 consecutive patients identified, 91 (36 women and 55 men; mean age, 69.3 ± 9.2 years; age range, 41 to 85 years) met the following inclusion criteria and were enrolled in the study: (1) treated with LUL, and (2) evaluated by cine-MRI for the presence or absence of PVST at 7 days after surgery. One case was excluded due to poor image quality. PVST was defined as a structure situated in the PV stump, possessing boundaries that are discernible from the PV wall and can be distinguished from artifacts. For all cases of PVST suspected on cine-MRI, contrast-enhanced ECG-gated CT was performed to confirm the existence of PVST. PVST was confirmed in 30 of the 91 patients. Table summarizes the patients’ clinical characteristics according to the presence or absence of PVST after LUL. Cine-MRI protocol and evaluation of left atrial function MRI examinations were performed on the seventh postoperative day, using a 3T system (Prisma, Siemens Medical Systems, Erlangen, Germany) with a 30-channel body array coil. At our institute, cardiac cine-MRI is included in the routine clinical protocol for assessing potential PVST formation following LUL because it involves no radiation exposure or contrast material injection, and has been recognized as a reliable diagnostic tool for assessing thrombi in the LA or LA appendage – . Cine-MR images were obtained in the coronal plane using a balanced steady-state free precession (bSSFP) sequence with short periods of breath-holding. The imaging parameters were as follows: field-of-view = 360 × 360 mm, in-plane spatial resolution = 1.9 × 1.9 mm, slice thickness = 5 mm, number of slices = 27, repetition time = 40–80 ms, echo time = 1.1 ms, flip angle = 48°, number of cardiac phases = 10–15, and number of signals averaged = 1. All MR images were transferred and analyzed using a 3D image-analysis system (SYNAPSE VINCENT; Fujifilm Medical Co., Tokyo, Japan). Two radiologists with 21 and 2 years of chest radiology experience performed manual LA segmentation. They were blinded to the final results concerning the presence or absence of PVST. LA functional parameters (LA end-systolic volume [LAESV], LA end-diastolic volume [LAEDV], LAESV index [LAESVI], LAEDV index [LAEDVI], LA stroke volume [LASV], and LA ejection fraction [LAEF]) were calculated. LAEF was calculated using the formula: (the maximum LA volume – the minimum LA volume) / the maximum LA volume×100. LAESVI and LAEDVI were calculated according to the body surface area (BSA) of each patient, as LAESV/BSA and LAEDV/BSA, respectively. The radiologists also measured the length of the left superior PV stump. Statistical analysis Intraclass correlation coefficient (ICC) was calculated to assess inter-observer agreement of all LA functional parameters (κ = 0.00–0.20, poor correlation; κ = 0.21–0.40, fair correlation; κ = 0.41–0.60, moderate correlation; κ = 0.61–0.80, good correlation; κ = 0.81–1.00, excellent correlation). Comparisons of all LA functional parameters and PV stump length between patients with and without development of PVST were conducted using the Mann–Whitney U test. Receiver-operating characteristic (ROC) curve analysis was performed to evaluate the predictive accuracy of LA functional parameters for PVST. Optimal cutoff values were selected based on the maximum Youden index for predicting PVST. Areas under the ROC curve (AUC) were compared using DeLong’s test. Clinical and LA functional parameters were analyzed by univariate and multivariate logistic regression models to determine predictors of PVST. For the multivariate analysis, the variance inflation factor (VIF) was calculated to assess multicollinearity among the predictor variables. Variables with VIF values exceeding 10 were excluded. Subsequently, a stepwise regression analysis was performed to identify the best predictor variables for PVST. Continuous variables are displayed as the mean ± standard deviation (SD). A P- value < 0.05 was considered to indicate statistical significance in all analyses. Statistical analyses were performed using MedCalc version 20.211 (MedCalc Software, Mariakerke, Belgium) and SPSS version 28.0 (SPSS, Chicago, IL). Our institutional ethics review board approved this retrospective study and the requirement for patients’ informed consent was waived. We reviewed the imaging database of our radiology department and patients’ electronic medical records to identify those who had undergone surgical treatment for lung neoplasm as well as a postoperative MR examination between January 2018 and April 2024. Among the 294 consecutive patients identified, 91 (36 women and 55 men; mean age, 69.3 ± 9.2 years; age range, 41 to 85 years) met the following inclusion criteria and were enrolled in the study: (1) treated with LUL, and (2) evaluated by cine-MRI for the presence or absence of PVST at 7 days after surgery. One case was excluded due to poor image quality. PVST was defined as a structure situated in the PV stump, possessing boundaries that are discernible from the PV wall and can be distinguished from artifacts. For all cases of PVST suspected on cine-MRI, contrast-enhanced ECG-gated CT was performed to confirm the existence of PVST. PVST was confirmed in 30 of the 91 patients. Table summarizes the patients’ clinical characteristics according to the presence or absence of PVST after LUL. MRI examinations were performed on the seventh postoperative day, using a 3T system (Prisma, Siemens Medical Systems, Erlangen, Germany) with a 30-channel body array coil. At our institute, cardiac cine-MRI is included in the routine clinical protocol for assessing potential PVST formation following LUL because it involves no radiation exposure or contrast material injection, and has been recognized as a reliable diagnostic tool for assessing thrombi in the LA or LA appendage – . Cine-MR images were obtained in the coronal plane using a balanced steady-state free precession (bSSFP) sequence with short periods of breath-holding. The imaging parameters were as follows: field-of-view = 360 × 360 mm, in-plane spatial resolution = 1.9 × 1.9 mm, slice thickness = 5 mm, number of slices = 27, repetition time = 40–80 ms, echo time = 1.1 ms, flip angle = 48°, number of cardiac phases = 10–15, and number of signals averaged = 1. All MR images were transferred and analyzed using a 3D image-analysis system (SYNAPSE VINCENT; Fujifilm Medical Co., Tokyo, Japan). Two radiologists with 21 and 2 years of chest radiology experience performed manual LA segmentation. They were blinded to the final results concerning the presence or absence of PVST. LA functional parameters (LA end-systolic volume [LAESV], LA end-diastolic volume [LAEDV], LAESV index [LAESVI], LAEDV index [LAEDVI], LA stroke volume [LASV], and LA ejection fraction [LAEF]) were calculated. LAEF was calculated using the formula: (the maximum LA volume – the minimum LA volume) / the maximum LA volume×100. LAESVI and LAEDVI were calculated according to the body surface area (BSA) of each patient, as LAESV/BSA and LAEDV/BSA, respectively. The radiologists also measured the length of the left superior PV stump. Intraclass correlation coefficient (ICC) was calculated to assess inter-observer agreement of all LA functional parameters (κ = 0.00–0.20, poor correlation; κ = 0.21–0.40, fair correlation; κ = 0.41–0.60, moderate correlation; κ = 0.61–0.80, good correlation; κ = 0.81–1.00, excellent correlation). Comparisons of all LA functional parameters and PV stump length between patients with and without development of PVST were conducted using the Mann–Whitney U test. Receiver-operating characteristic (ROC) curve analysis was performed to evaluate the predictive accuracy of LA functional parameters for PVST. Optimal cutoff values were selected based on the maximum Youden index for predicting PVST. Areas under the ROC curve (AUC) were compared using DeLong’s test. Clinical and LA functional parameters were analyzed by univariate and multivariate logistic regression models to determine predictors of PVST. For the multivariate analysis, the variance inflation factor (VIF) was calculated to assess multicollinearity among the predictor variables. Variables with VIF values exceeding 10 were excluded. Subsequently, a stepwise regression analysis was performed to identify the best predictor variables for PVST. Continuous variables are displayed as the mean ± standard deviation (SD). A P- value < 0.05 was considered to indicate statistical significance in all analyses. Statistical analyses were performed using MedCalc version 20.211 (MedCalc Software, Mariakerke, Belgium) and SPSS version 28.0 (SPSS, Chicago, IL). |
Patient-derived cell-based pharmacogenomic assessment to unveil underlying resistance mechanisms and novel therapeutics for advanced lung cancer | c3e742e8-8901-4f2a-8d54-89334dcea15b | 9885631 | Pharmacology[mh] | Lung cancer is the leading cause of cancer-related mortality worldwide . The development of targeted therapies, such as epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors (TKIs), have helped to extend the survival time of patients; however, the improvement in progression-free survival eventually fails in cases of advanced lung cancer owing to resistance development . Thus, treatment tailored to overcome resistance is need to improve prognosis. To develop novel therapeutics, several cellular, organoid, and mouse models are available for use in pharmacological platforms . Although organoid and mouse models recapitulate the heterogeneous molecular characteristics of patient biopsies, their establishment is labor-intensive and they have relatively low tumor-formation rates . Models based on patient-derived immortalized cells provide good reproducibility to the cohort, but have less heterogeneity, are more mesenchymal in nature, and have distinct chemical and genetic dependencies . Fortunately, models derived from short-term–cultured patient-derived cells (PDCs) retain the genomic characteristics of solid tumor biopsy better than immortalized cell lines . Thus, the use of PDC models can help to enlarge the scope of drug screening to include more chemicals and multiple doses. To date, PDC pharmacogenomic platforms have been successfully used in treating glioblastoma and gynecologic and gastric cancers . A PDC platform was also established to screen drugs for non-small cell lung cancer (NSCLC) . However, this platform required cell culture for 2–6 months and demonstrated a 50% success rate. These numbers represent an obstacle for its application in the medical field to fill the need of immediate patient-tailored drug prediction. Therefore, establishing a PDC platform for refractory lung cancers would satisfy the clinically unmet needs of therapeutic and drug-resistance research. Advances in large-scale lung cancer genomics approaches have helped to tackle the challenges posed by tumor heterogeneity, therapeutic evolution, and histology . EGFR-TKI therapies possess good EGFR -mutation selectivity, but resistance occasionally evolves via activation of non-targetable bypass pathways such as neuronal differentiation or KRAS amplification . Treatment of small-cell lung cancer (SCLC) is associated with an additional challenge of the lack of further effective therapies in the face of rapid resistance acquisition to platinum-based chemotherapy . Moreover, information available on driver genes and various (TKI-resistant) variants of refractory lung cancers have not been effectively translated into clinical targeted therapies. To bridge this gap, in this study, we developed PDC models using mainly the pleural effusions of refractory lung cancer patients. Using our platform, we explored drug candidates and target regulatory mechanisms according to genomic features and lung cancer molecular types. We further assessed the pharmacogenomic characteristics of the cells by screening their drug responses and performing next-generation sequencing. To establish integrative analysis for drug and genomic characteristics, statistical and machine-learning methods were employed to investigate the sensitivity to each drug according to molecular subtype, cancer types, therapeutic groups, and variants. This PDC platform and associated analysis can highlight novel drug candidates and target pathways to improve the personalized treatment of advanced lung cancer and facilitate further research in this regard. Lung cancer sample acquisition and PDC establishment Cancer samples were collected from patients with advanced or refractory lung cancer diagnosed and treated at the National Cancer Center in Korea between December 2016 and February 2020. The histological types were determined according to the 2015 World Health Organization classification of lung tumors. This study was approved by the National Cancer Center Institutional Review Board (approval number NCC2019-0082). All patients provided written informed consent. PDC establishment and drug screening details are described in Additional file (Supplementary methods). Drug sensitivity screening using PDCs and cell lines Stabilized PDCs were seeded in 384-well plates (1000 cells/20 μl/well) in quadruplicate for each treatment. A total of 16 or 48 compounds were used for screening each PDC sample (Additional file : Table S1, S2). After overnight incubation, the cells were treated with one drug at a 5-fold serial dilution for a total of 6 doses (50 μM ~ 16 nM). Cell viability was measured after 72 hrs of treatment using the CellTiter-Glo Luminescent Cell Viability Assay kit (Promega, Madison, WI, USA) and an Infinite 200 Pro system (TECAN, Mannedorf, Switzerland). Each screening plate contained a dimethyl sulfoxide (DMSO)-only vehicle to calculate relative cell viability and normalize the data. Dose response curve (DRC) fitting and area under the curve (AUC) values were assessed using GraphPad Prism 5.3 (GraphPad Software Inc., San Diego, CA, USA). A screening compound library was newly prepared every month and tested for the preservation of chemical activities using NSCLC cancer cell lines (A549, PC9 and H1299). All library compounds were purchased from Selleckchem (Houston, TX, USA). Targeted next-generation sequencing (NGS) dataset and molecular subtype identification Mutation and copy number variant (CNV) calling were performed using NGS and targeted sequencing. Gene expression profiles and fusion genes were identified from RNA-sequencing data. To identify somatic mutations from target-seq data, we started with preprocessing with quality check, read trimming using Trimmomatic 0.39, and alignment using BWA 0.7.17 to hg19 . Alignment bam files were recalibrated and realigned for target regions using Picard 1.119 and GATK 4.1.3.0 . Next, we called somatic mutations using Mutect2, referring to the panel-of-normal from the 1000 Genomes PON and gnomAD VCF files . The somatic mutation results were annotated using Oncotator . Additionally, we eliminated germline variants registered in the 1000 Genomes Project. The CNV profile was identified using GATK 4.0.4.0, and CNV peak calling was performed by GISTIC2 . Additionally, we selected CNV genes as those with high amplification (log2 CNV > 2), high amplification rate (> 5%), and a correlation with mRNA expression ( P value < 0.01). The tumor mutation burden (TMB) of the mutation count per mega base pair (perMbp) values was calculated by Maftools . We classified TMB values into low (TMB < 0.1), middle (0.1 ≤ TMB < 0.2), and high (0.2 ≤ TMB) groups. To additionally classify the functionality of the genes with the most recurrent mutations ( TP53 and EGFR) , we categorized previously reported hotspot and non-hotspot mutations. TP53 mutations were classified into hotspot and non-hotspot mutations . EGFR variants were divided into four types (TARGET: exon 19 deletion, L858R), T790M acquisition (T790M aq ), not otherwise specified mutations (NOS), and NOS acquisition with target (NOS aq ) . The co-occurrence of mutated genes was tested by Fisher’s exact test for genes with recurrence > 3 and MutSig P value < 0.05 . The double mutation pairs were referred for drug sensitivity test as well as single mutation cases. RNA-seq analysis also proceeded with similar preprocessing with quality check, read trimming using Trimmomatic, and alignment using STAR v2.7.0a to hg19 referring to gene model ENSEMBL release 75 . The gene expression profile was extracted from RPKM quantified from bam files using RSEM v1.2.31 . Fusion genes were identified from RNA-Seq and merged with the results from three callers, Defuse, PRADA v1.2 and STAR-Fusion v1.7.0 . Fusions were annotated using Pegasus, and we extracted drug-targetable candidates involved in kinase or oncogenic signaling from published databases from TCGA and COSMIC . To classify the RNA molecular subtype of the samples, we performed nonnegative matrix factorization (NMF) clustering using the RPKM gene expression profile. We assessed optimal cluster size from 2 to 5. Finally, cluster size for molecular subtype was chosen to computationally present a clear consensus plot ( n = 4). To extract the transcriptomic characteristics of RNA subtypes, pathway activities for each sample were estimated using gene set variation analysis (GSVA) according to HALLMARK gene set collections . Next, the difference in pathway activities according to RNA subtypes was tested by limma . Additionally, the up-regulation of stemness-associated gene signatures were assessed using GSVA . The gene signatures were collected from microarray, and ChIP-seq of two human embryonic stem cell gene sets, target genes of transcription factors (NANOG, OCT4, SOX2, and MYC), and Polycomb targets (Suz12, Eed, H2K27, and PRC2) to be under-expression to embryonic stem cells . The difference of collected signature scores for subtype was tested by Wilcoxon rank-sum test. Comparison of PDCs with other lung cancer datasets We evaluated the characteristics of PDCs and survival using external lung cancer cohorts. First, to evaluate the genomic concordance of PDCs with solid tumor biopsies, we assessed the similarity of mutations and expression profiles between our PDCs and The Cancer Genome Atlas (TCGA) lung adenocarcinoma (LUAD) datasets . Before comparison, the transcriptome batch effect was eliminated using ComBat . The expression profiles were verified from a principal component analysis plot. The frequencies of the most recurrent mutated genes in PDCs were also compared with those in TCGA dataset. Next, prognostic significance according to distinct molecular subtypes determined based on gene expression profiles was evaluated using meta-transcriptome datasets: five National Center for Biotechnology Information Gene Expression Omnibus lung cancer datasets and TCGA-LUAD datasets ( n = 1587 patients; see Additional file , Fig. S1). The batch effect among multiple datasets was also eliminated using ComBat. We respectively calculated the four subtype signature scores from the meta-transcriptome using GSVA to obtain differentially expressed gene (DEG) sets ( n = 300). The subtype DEGs were acquired using the limma test from our PDCs. Next, the log-rank test and Cox proportional hazard analysis were performed to compare overall survival according to high (> 25%) and low (≤ 25%) scores for each subtype. Drug sensitivity test according to genomic variants and groups To evaluate the drug sensitivity of our PDCs according to genomic characteristics, we performed tests for multiple conditions, including lung cancer histologic type, EGFR-TKI therapy group, mutations, CNVs, fusions, RNA subtypes, and co-occurring mutation pairs. TP53 and EGFR mutations were additionally categorized (details are described in the Results). We performed the Wilcoxon rank-sum test using area under the dose response curve (AUC) values between the two groups. P -values were adjusted using the Benjamini–Hochberg method. To compare alpelisib response of RB1 / TP53 cells between PDCs and cell lines, we additionally collected drug screening dataset of CCLE lung cancer cell lines ( n = 70) to include SCLC ( n = 7) . Alpelisib sensitivity test was also performed by Wilcoxon rank-sum test for mutated cases, and SCLC lung cancer type. Evaluation of cell cycle inhibitors effective for SCLC and the associated gene FOXM1 We investigated the transcriptome characteristics of SCLC from differentially expressed gene (DEG) analysis to compare SCLC with NSCLC using limma . Next, gene set enrichment analysis (GSEA) was performed using the upregulated DEGs for each group by referring to WikiPathways To evaluate drugs and genes for SCLC, we additionally investigated drug sensitivity data for SCLC versus NSCLC from cell lines. Our drug response screening result for AZD7762 was acquired from 13 NSCLC cell lines and 5 SCLC cell lines, like our PDCs. To assess the similarity of drug signature genes between cell line and PDC, we additionally calculated Pearson correlation coefficient . The siRNAs targeting FOXM1 (Hs_FOXM1_6 and Hs_FOXM1_7) and the control siRNA were purchased from QIAGEN (Foster City, CA, USA). H69 and H209 cells were transiently transfected with siRNAs using a NEPA21 electroporator (NEPA GENE, Chiba, Japan). Suspension cells (1 × 10 6 ) with 100 pmole siRNA in 100 μl OPTI MEM media per cuvette were subjected to electroporation with program No. 5 following the manufacturer’s instructions. Drug response-associated gene signature extracted using machine learning To extract gene sets related to the response to each drug, we used the expression profiles to filter out genes according to mean expression level ≤ 1 and standard deviation ≤2. AUC values were transformed to a log 2 scale, and gene expression values were converted to z scores of log 2 -scaled RPKM values. First, we selected the top 500 correlated genes with the AUC values for each drug. Next, we performed elastic net regularization to extract the gene feature importance for each drug response . The glmnet R package was applied to optimize parameters from nine values of α ∈ [0.1,0.9] and 50 values of λ ∈ [0.01,100] to minimize the root mean squared error using 10-fold cross-validation . Bootstrapping was performed 500 times using the R boot package to extract drug gene signatures from features . Finally, gene feature importance was extracted for each drug, and genes were ranked to define drug signature gene sets. Next, we performed GSVA to investigate pathways enriched for each drug gene signature using Reactome with significant enrichment was assessed at P < 0.1 . Molecular characteristics and drug identification according to EGFR-TKI therapeutics To identify the molecular features related to EGFR-TKI therapeutics, we categorized therapeutic groups from 27 PDC samples. We categorized these patients into four groups (Fig. A): (1) BASELINE, PDCs acquired from EGFR -mutated patients that did not receive any treatment; (2) POST1, PDCs without EGFR T790M mutation, acquired after disease progression to the first-line use of first- or second-generation EGFR-TKIs; (3) POST2, PDCs with EGFR T790M, acquired from patients after first- or second-generation EGFR-TKI treatment; and (4) POST3, T790M-positive PDCs, acquired after disease progression to second-line use of the third-generation TKI osimertinib. To identify gene regulation according to these four treatment groups, we identified upregulated DEGs ( P < 0.01) for each group and associated pathways using limma and GSEA ( P < 0.1 ). Additionally, we collected known EGFR-TKI resistance pathway gene sets: MAPK, PI3K-AKT, JAK-STAT, Wnt β-catenin, plasminogen activation, neuroendocrine activation, YAP/TAZ, MET, HER2, RAS, ERK, KRAS, and TAM (TYRO3-AXL-MERTK) family genes . Resistance pathway scores were calculated using GSVA for each PDC. The score difference of the four therapeutics groups was evaluated using the Wilcoxon rank-sum test . When exploring sensitivity to drugs for each EGFR-TKI group, the Wilcoxon test was also performed to determine the AUC value difference. We additionally demonstrate our “sensitive-drug candidates” assessed from 27 PDCs extended to PDC pools ( n = 70). To improve the statistical reliability of the current dataset, we also collected additional drug screening results of extended EGFR -mutated PDCs ( n = 70) acquired from patients previously treated with EGFR-TKIs based only on the clinical treatment profile without the corresponding NGS profile. The generation of osimertinib-resistant cell lines and experiments are described in Additional file (Supplementary methods). Cancer samples were collected from patients with advanced or refractory lung cancer diagnosed and treated at the National Cancer Center in Korea between December 2016 and February 2020. The histological types were determined according to the 2015 World Health Organization classification of lung tumors. This study was approved by the National Cancer Center Institutional Review Board (approval number NCC2019-0082). All patients provided written informed consent. PDC establishment and drug screening details are described in Additional file (Supplementary methods). Stabilized PDCs were seeded in 384-well plates (1000 cells/20 μl/well) in quadruplicate for each treatment. A total of 16 or 48 compounds were used for screening each PDC sample (Additional file : Table S1, S2). After overnight incubation, the cells were treated with one drug at a 5-fold serial dilution for a total of 6 doses (50 μM ~ 16 nM). Cell viability was measured after 72 hrs of treatment using the CellTiter-Glo Luminescent Cell Viability Assay kit (Promega, Madison, WI, USA) and an Infinite 200 Pro system (TECAN, Mannedorf, Switzerland). Each screening plate contained a dimethyl sulfoxide (DMSO)-only vehicle to calculate relative cell viability and normalize the data. Dose response curve (DRC) fitting and area under the curve (AUC) values were assessed using GraphPad Prism 5.3 (GraphPad Software Inc., San Diego, CA, USA). A screening compound library was newly prepared every month and tested for the preservation of chemical activities using NSCLC cancer cell lines (A549, PC9 and H1299). All library compounds were purchased from Selleckchem (Houston, TX, USA). Mutation and copy number variant (CNV) calling were performed using NGS and targeted sequencing. Gene expression profiles and fusion genes were identified from RNA-sequencing data. To identify somatic mutations from target-seq data, we started with preprocessing with quality check, read trimming using Trimmomatic 0.39, and alignment using BWA 0.7.17 to hg19 . Alignment bam files were recalibrated and realigned for target regions using Picard 1.119 and GATK 4.1.3.0 . Next, we called somatic mutations using Mutect2, referring to the panel-of-normal from the 1000 Genomes PON and gnomAD VCF files . The somatic mutation results were annotated using Oncotator . Additionally, we eliminated germline variants registered in the 1000 Genomes Project. The CNV profile was identified using GATK 4.0.4.0, and CNV peak calling was performed by GISTIC2 . Additionally, we selected CNV genes as those with high amplification (log2 CNV > 2), high amplification rate (> 5%), and a correlation with mRNA expression ( P value < 0.01). The tumor mutation burden (TMB) of the mutation count per mega base pair (perMbp) values was calculated by Maftools . We classified TMB values into low (TMB < 0.1), middle (0.1 ≤ TMB < 0.2), and high (0.2 ≤ TMB) groups. To additionally classify the functionality of the genes with the most recurrent mutations ( TP53 and EGFR) , we categorized previously reported hotspot and non-hotspot mutations. TP53 mutations were classified into hotspot and non-hotspot mutations . EGFR variants were divided into four types (TARGET: exon 19 deletion, L858R), T790M acquisition (T790M aq ), not otherwise specified mutations (NOS), and NOS acquisition with target (NOS aq ) . The co-occurrence of mutated genes was tested by Fisher’s exact test for genes with recurrence > 3 and MutSig P value < 0.05 . The double mutation pairs were referred for drug sensitivity test as well as single mutation cases. RNA-seq analysis also proceeded with similar preprocessing with quality check, read trimming using Trimmomatic, and alignment using STAR v2.7.0a to hg19 referring to gene model ENSEMBL release 75 . The gene expression profile was extracted from RPKM quantified from bam files using RSEM v1.2.31 . Fusion genes were identified from RNA-Seq and merged with the results from three callers, Defuse, PRADA v1.2 and STAR-Fusion v1.7.0 . Fusions were annotated using Pegasus, and we extracted drug-targetable candidates involved in kinase or oncogenic signaling from published databases from TCGA and COSMIC . To classify the RNA molecular subtype of the samples, we performed nonnegative matrix factorization (NMF) clustering using the RPKM gene expression profile. We assessed optimal cluster size from 2 to 5. Finally, cluster size for molecular subtype was chosen to computationally present a clear consensus plot ( n = 4). To extract the transcriptomic characteristics of RNA subtypes, pathway activities for each sample were estimated using gene set variation analysis (GSVA) according to HALLMARK gene set collections . Next, the difference in pathway activities according to RNA subtypes was tested by limma . Additionally, the up-regulation of stemness-associated gene signatures were assessed using GSVA . The gene signatures were collected from microarray, and ChIP-seq of two human embryonic stem cell gene sets, target genes of transcription factors (NANOG, OCT4, SOX2, and MYC), and Polycomb targets (Suz12, Eed, H2K27, and PRC2) to be under-expression to embryonic stem cells . The difference of collected signature scores for subtype was tested by Wilcoxon rank-sum test. We evaluated the characteristics of PDCs and survival using external lung cancer cohorts. First, to evaluate the genomic concordance of PDCs with solid tumor biopsies, we assessed the similarity of mutations and expression profiles between our PDCs and The Cancer Genome Atlas (TCGA) lung adenocarcinoma (LUAD) datasets . Before comparison, the transcriptome batch effect was eliminated using ComBat . The expression profiles were verified from a principal component analysis plot. The frequencies of the most recurrent mutated genes in PDCs were also compared with those in TCGA dataset. Next, prognostic significance according to distinct molecular subtypes determined based on gene expression profiles was evaluated using meta-transcriptome datasets: five National Center for Biotechnology Information Gene Expression Omnibus lung cancer datasets and TCGA-LUAD datasets ( n = 1587 patients; see Additional file , Fig. S1). The batch effect among multiple datasets was also eliminated using ComBat. We respectively calculated the four subtype signature scores from the meta-transcriptome using GSVA to obtain differentially expressed gene (DEG) sets ( n = 300). The subtype DEGs were acquired using the limma test from our PDCs. Next, the log-rank test and Cox proportional hazard analysis were performed to compare overall survival according to high (> 25%) and low (≤ 25%) scores for each subtype. To evaluate the drug sensitivity of our PDCs according to genomic characteristics, we performed tests for multiple conditions, including lung cancer histologic type, EGFR-TKI therapy group, mutations, CNVs, fusions, RNA subtypes, and co-occurring mutation pairs. TP53 and EGFR mutations were additionally categorized (details are described in the Results). We performed the Wilcoxon rank-sum test using area under the dose response curve (AUC) values between the two groups. P -values were adjusted using the Benjamini–Hochberg method. To compare alpelisib response of RB1 / TP53 cells between PDCs and cell lines, we additionally collected drug screening dataset of CCLE lung cancer cell lines ( n = 70) to include SCLC ( n = 7) . Alpelisib sensitivity test was also performed by Wilcoxon rank-sum test for mutated cases, and SCLC lung cancer type. We investigated the transcriptome characteristics of SCLC from differentially expressed gene (DEG) analysis to compare SCLC with NSCLC using limma . Next, gene set enrichment analysis (GSEA) was performed using the upregulated DEGs for each group by referring to WikiPathways To evaluate drugs and genes for SCLC, we additionally investigated drug sensitivity data for SCLC versus NSCLC from cell lines. Our drug response screening result for AZD7762 was acquired from 13 NSCLC cell lines and 5 SCLC cell lines, like our PDCs. To assess the similarity of drug signature genes between cell line and PDC, we additionally calculated Pearson correlation coefficient . The siRNAs targeting FOXM1 (Hs_FOXM1_6 and Hs_FOXM1_7) and the control siRNA were purchased from QIAGEN (Foster City, CA, USA). H69 and H209 cells were transiently transfected with siRNAs using a NEPA21 electroporator (NEPA GENE, Chiba, Japan). Suspension cells (1 × 10 6 ) with 100 pmole siRNA in 100 μl OPTI MEM media per cuvette were subjected to electroporation with program No. 5 following the manufacturer’s instructions. To extract gene sets related to the response to each drug, we used the expression profiles to filter out genes according to mean expression level ≤ 1 and standard deviation ≤2. AUC values were transformed to a log 2 scale, and gene expression values were converted to z scores of log 2 -scaled RPKM values. First, we selected the top 500 correlated genes with the AUC values for each drug. Next, we performed elastic net regularization to extract the gene feature importance for each drug response . The glmnet R package was applied to optimize parameters from nine values of α ∈ [0.1,0.9] and 50 values of λ ∈ [0.01,100] to minimize the root mean squared error using 10-fold cross-validation . Bootstrapping was performed 500 times using the R boot package to extract drug gene signatures from features . Finally, gene feature importance was extracted for each drug, and genes were ranked to define drug signature gene sets. Next, we performed GSVA to investigate pathways enriched for each drug gene signature using Reactome with significant enrichment was assessed at P < 0.1 . To identify the molecular features related to EGFR-TKI therapeutics, we categorized therapeutic groups from 27 PDC samples. We categorized these patients into four groups (Fig. A): (1) BASELINE, PDCs acquired from EGFR -mutated patients that did not receive any treatment; (2) POST1, PDCs without EGFR T790M mutation, acquired after disease progression to the first-line use of first- or second-generation EGFR-TKIs; (3) POST2, PDCs with EGFR T790M, acquired from patients after first- or second-generation EGFR-TKI treatment; and (4) POST3, T790M-positive PDCs, acquired after disease progression to second-line use of the third-generation TKI osimertinib. To identify gene regulation according to these four treatment groups, we identified upregulated DEGs ( P < 0.01) for each group and associated pathways using limma and GSEA ( P < 0.1 ). Additionally, we collected known EGFR-TKI resistance pathway gene sets: MAPK, PI3K-AKT, JAK-STAT, Wnt β-catenin, plasminogen activation, neuroendocrine activation, YAP/TAZ, MET, HER2, RAS, ERK, KRAS, and TAM (TYRO3-AXL-MERTK) family genes . Resistance pathway scores were calculated using GSVA for each PDC. The score difference of the four therapeutics groups was evaluated using the Wilcoxon rank-sum test . When exploring sensitivity to drugs for each EGFR-TKI group, the Wilcoxon test was also performed to determine the AUC value difference. We additionally demonstrate our “sensitive-drug candidates” assessed from 27 PDCs extended to PDC pools ( n = 70). To improve the statistical reliability of the current dataset, we also collected additional drug screening results of extended EGFR -mutated PDCs ( n = 70) acquired from patients previously treated with EGFR-TKIs based only on the clinical treatment profile without the corresponding NGS profile. The generation of osimertinib-resistant cell lines and experiments are described in Additional file (Supplementary methods). Establishment and molecular characteristics of lung cancer PDCs We established a PDC collection of 102 samples (National Cancer Center; see Additional file , Table S1) from patients with advanced lung cancer enrolled in this study (Fig. A). PDCs were primarily collected from pleural effusions (92.2%), and secondarily from pericardial effusions (4.9%), ascites (2.0%), and tissues (1.0%; Additional file , Table S1). The pathologic type was adenocarcinoma (ADC; 84.3%), SCLC (5.9%), and miscellaneous types (squamous cell carcinoma, 4.9%; sarcomatoid carcinoma, 3.9%; and not otherwise specified, 1%). The drug panel used for response screening included 48 anti-cancer compounds of seven classes targeting angiogenesis ( n = 3), the cell cycle ( n = 8), DNA damage ( n = 6), MAPK ( n = 3), PI3K/AKT/mTOR ( n = 3), protein tyrosine kinase ( n = 7), and others ( n = 18; see Additional file , Table S2). Among the 48 drugs, 16 were screened in all PDCs and 32 were screened in 38 PDCs. To investigate the genomic characteristics, we identified somatic mutations and CNVs using target-seq ( n = 98) and classified four molecular subtypes (C1–C4) from gene expression profiles ( n = 102; Additional file , Table S1). Next, we checked whether the PDCs effectively recapitulate the mutations and expression profiles of the solid tumors. Owing to the dominance of LUAD in our cohort (> 80%), we compared PDC gene expression profiles with TCGA-LUAD dataset ( n = 230). As expected, the PDC gene expression profile resembled with tumor samples and was separated from normal adjacent tissues (Fig. B) . Constitutive somatic gene mutations were similar in PDCs and TCGA samples. The recurrence of TP53 , RB1 , and BRAF mutations was highly preserved in both PDC and TCGA samples. The EGFR mutation frequency was higher in PDCs, whereas the recurrence of KRAS , KEAP1 , and STK11 mutations was lower than that of TCGA samples not shown in Fig. B. Thus, somatic mutations in TP53 (47%), EGFR (29%), and RB1 (8%) were frequently observed in PDC models (Fig. C). Moreover, MET (10%), CDK4 (6%), and MDM2 (6%) variants, as well as EML4 - ALK (4%) and CD74 - ROS (2%) fusion genes, were detected. Before in-depth analysis, the most frequent TP53 and EGFR mutations were additionally categorized according to selectivity or functionality . TP53 mutations were divided into hotspot (known as gain-of-function; 5.9%; R175, G245, R248, R249, R273, R282) and non-hotspot (unknown or loss-of-function; 41.2%) mutations. EGFR mutations were categorized into (1) TKI-targetable single variant (TARGET: L858R and exon 19 deletion = 19.6%), (2) T790M acquisition (T790M aq : TARGET and T790M = 2.9%), (3) non-other-specified (NOS) single variant except TARGET (4.9%), and (4) NOS acquisition (NOS aq ; TARGET and NOS = 2%; Table ) . Molecular subtype classification and targeted drug candidate identification PDC samples could be classified into four molecular subtypes using gene expression profile by NMF clustering. To extrapolate the clinical characteristics for each subtype, we interrogated patient clinical profile encompassing histologic type, survival, smoking, and EGFR-TKI therapy record (Fig. and Table ). To uncover regulatory program for each molecular subtype, variant enrichment, and pathway regulation scores were assessed from multi-omics profile. In brief, subtype C1 was associated with a good outcome shown in Fig. A, and was activated in inflammatory and IL6-JAK-STAT3 signaling pathways; subtype C2 was associated with a modest outcome, dominance of TP53 / EGFR wild-type, upregulation of epithelial-to-mesenchymal transition (EMT), and enrichment of osimertinib-resistant group (POST3) PDCs; subtype C3 was associated with the worst outcome, long-term smoking males, SCLC, TP53 hotspot mutation, MYC activation, and fasten G2M checkpoint; by contrast, subtype C4 exhibited the best survival, frequent TP53 non-hotspot mutation, the dominance of EGFR-TKI TARGET mutation, and NOTCH signaling activation (Fig. A, B). We also demonstrated the survival significance corresponding to C1–C4 subtype gene sets using additional transcriptome datasets ( n = 1587; see Fig. S1A-B in Additional file and Table S3 in Additional file ). Upregulation of the C3 gene set concurrently exhibited the worst outcomes ( P < 0.001 and hazard ratio [HR] = 2.6). When additionally demonstrating from known up-regulated genes of embryonic stem cell, we could observe the activation of human embryonic stem cell genes, and target genes’ upregulation of transcription factor MYC and SOX2 (Fig. S2 in Additional file ). The activation of C1 and C2 subtype genes significantly showed better prognosis ( P < 0.05 and HR < 0.79). Our subtype classification sustained a global prognostic signature for lung cancer. Finally, we could summarize the subtypes based on the following regulatory pathways: C1, inflammatory; C2, EMT-like; C3, stemness; and C4, EGFR -dominant. Sensitive drug candidates exhibited remarkable concordance with the previously identified regulatory pathways for each molecular subtype ( P < 0.05 and |log 2 fold change (FC)| < 0.2; FC was assessed to compare average drug AUC values between corresponding subtype and another group; Fig. C). The C1 inflammatory subtype was sensitive to only ruxolitinib (JAK1/2) that targets the JAT/STAT pathway (Fig. B-D). The C2 EMT-like subtype showed MAPK class drug resistance and sensitivity to dasatinib (angiogenesis and SRC inhibitor), cabozantinib (VEGFA inhibitor), miscellaneous class repotrectinib (ROS inhibitor), and XAV939 (WNT-TNKS-β-catenin inhibitor). The C3 stemness subtype was sensitive to the top five-ranked cell cycle inhibitors. The C4 EGFR -dominant subtype exhibited the strongest resistance to most drug classes except MAPK inhibitors (selumetinib and trametinib) and EGFR-TKIs (gefitinib and afatinib). Notably, targets of predicted drugs belonged to pathways that were found to be activated in each subtype (Fig. B). Predicting drugs for variants and dissecting the molecular subtype of EGFR-mutated PDCs To interrogate drug candidates for variants, we tested the difference in drug responses of single and co-occurrent mutation cases ( P < 0.005 and |log 2 FC| < 0.2; Fig. A). Co-mutated cases were also explored using the Fisher’s exact test ( P < 0.25; see Additional file , Fig. S3). Unexpectedly, EGFR -TARGET mutated PDCs showed a relatively modest response to three EGFR-TKIs (afatinib P = 0.17, gefitinib P = 0.19, osimertinib P = 0.08). To uncover EGFR -mutated cells’ molecular features interfering with the EGFR-TKI response, we dissected TKI TARGET mutations according to our four molecular subtypes (Fig. B). The C4 EGFR -dominant subtype was the most sensitive to all EGFR-TKIs, and the C1 inflammatory subtype also showed an especially good response to afatinib and osimertinib. Mutated cases ( n = 2) of the C3 stemness subtype was insufficient for statistical test. Finally, mutated cases ( n = 6) of the C2 EMT-like subtype did not respond to any EGFR-TKIs. Additionally, both T790M aq and NOS aq ( EGFR R776G and I744M with L858R) PDCs were TKI-sensitive. In the EGFR NOS type mutations, G719A was observed, and it comprised 11.5% among the NOS mutations in previous NSCLC study, and patient-derived xenografts demonstrated that the mutation was resistant to osimertinib . The remaining NOS mutations excluded exon 18–21 and had a low possibility of finding a structure-based therapeutic target . Therefore, we concluded that the remaining NOS group mutations were not targetable by EGFR-TKIs. Especially, EGFR TARGET mutated PDCs classified to EMT-like subtype exhibited low response to EGFR-TKIs. Thus, our molecular subtype of EGFR mutations revealed that these PDC models can be used to verify the heterogeneous tumor environment affecting drug responses. When interrogating drug candidates for mutations, BRAF variants were sensitive to trametinib (Fig. A). The TP53 non-hotspot exhibited resistance to olaparib (PARP inhibitor), repotrectinib, and dasatinib. RB1 variants responded well to alpelisib. Although TP53 mutation was not associated with any drug response ( P < 0.28), RB1 / TP53 co-mutated samples exhibited certain sensitivity that was stronger than RB1 (Fig. A). RB1 / TP53 mutations were mostly enriched in SCLC ( n = 4; P < 0.001) but were also observed in NSCLC ( n = 3; Additional file , Fig. S3). We additionally investigated alpelisib sensitivity comparing PDCs with cell lines ( n = 70; SCLC n = 7). Notably, although alpelisib inhibited SCLC (Fig. C), RB1 / TP53 variants ( P < 0.001) were more strongly inhibited than SCLC ( P = 0.02). However, in contrast to RB1/TP53 ( P < 0.01), we failed to identify SCLC sensitivity to alpelisib ( P = 0.47; Fig. C) using PDCs. Meanwhile, another PIK3-AKT target drug capivasertib (AKT1) was sensitive in RB1 / EGFR -NOS and EGFR -NOS cases. When dissecting details, two of the five ( RB1 / EGFR -NOS) PDCs was classified to SCLC type additionally harboring TP53 mutations (Fig. S3B in Additional file ) . Therefore, we could infer that RB1 / EGFR -NOS PDC cases accompanied the dependency to SCLC type. Therefore, PI3K-AKT class drugs detected from SCLC or RB1 / TP53 cells also affect EGFR -NOS mutations. Collectively, mutation-based drug detection result emphasize that the genomic characteristics could predict both the cancer drug response and histological lung cancer type SCLC (Additional file , Table S1). Furthermore, our PDC model predicted alpelisib as an appropriate treatment for SCLCs, which was not identified in cell lines (Fig. C). To delineate the drug susceptibility of alpelisib from transcriptome, we extracted the transcriptome signature for each drug using machine-learning approaches. The response to alpelisib was modulated by a transcriptome signature enriched “MITOTIC G1 PHASE AND G1 S TRANSITION” pathway gene set (Fig. D, and Additional file , Table S4). Upregulation of MYC , E2F1 , and CCNB1 expression led to resistance to alpelisib. Previously, we uncovered that stem-like C3 subtype exhibited both MYC up-regulation, and SCLC enrichment (Additional file , Fig. S2). MYC facilitates SCLC tumorigenesis . Moreover, our SCLC PDCs predicted the response to alpelisib better than cell lines (Fig. C). Therefore, transcriptome gene signature extracted by machine-learning determined alpelisib response for SCLC like RB1 / TP53 . FOXM1 over-expression in SCLC type and sensitivity to cell cycle inhibitors For further in-depth exploration of the candidate drugs for treating different lung cancer types, we investigated the drug responses of ADC, SCLC, and miscellaneous cancers. SCLC was sensitive to drugs targeting cell cycle pathways (adavosertib, barasertib, berzosertib, and AZD7762) and DNA damage (vorinostat; P < 0.1 and |log 2 FC| < 0.3; Fig. A). The SCLC sensitivity to alpelisib was relatively lower than that to cell cycle drugs according to the FC value. The machine-learning signatures of three drugs (adavosertib, AZD7762, and berzosertib) consistently contained cell cycle pathways (Fig. B, and Additional file , Table S4). The transcriptome profile revealed the activation of cell cycle genes in SCLC. Moreover, the DNA damage response and RAS signal were downregulated in NSCLC (Additional file , Fig. S4). Collectively, these comprehensive pharmacogenomics result suggest the role of cell cycle inhibition in treating SCLC types. PDCs and cell lines showing FOXM1 upregulation exhibited AZD7762 sensitivity (Fig. C, D; PDC R = –0.77; cell line R = –0.51). FOXM1 was clearly over-expressed in SCLC compared to that in LUAD (Fig. E) . To evaluate whether FOXM1 is a plausible target to AZD7762 for SCLC, we transfected H69 cells with small interfering RNAs (siRNAs) against FOXM1 (#6 and #7, siFOXM1) and control siRNA and measured the FOXM1 protein level using western blotting. The FOXM1 expression level was decreased in siFOXM1 (Fig. F). siFOXM1 cells also exhibited decreased cell proliferation compared to siControl-transfected cells. However, G2-M phase arrest was detected in siFOXM1 H69 cells (Fig. G). siFOXM1 became more resistant to AZD7762 (AUC FC > 1.1) than siControl-transfected cells (Fig. H). Another cell lines H209 also presented the same result (see Additional file , Fig. S5). Therefore, we concluded that FOXM1 plays essential role to SCLC participating in the G2-M phase, and can be targeted by cell cycle inhibitors. Pharmacogenomic analysis according to EGFR-TKI treatment status Among our PDCs, we identified 27 EGFR-TKI-treated cases with corresponding NGS results (independently obtained in the clinic using tissue samples). As described in methods, we these patients into four groups (Fig. A). Primarily, POST3 samples (86%) were enriched in the C2 EMT-like subtype (Table ). EGFR mutation-calling failure was observed in the clinical tumor biopsy NGS record of some patients from which BASELINE ( n = 2) and POST2 ( n = 1) PDCs were derived. The mutation-calling failure likely originated from differences in the platform between the clinical biopsy and laboratory PDC testing. Moreover, TP53 non-hotspot mutations were present in all of the POST2 PDCs and were markedly absent in POST3 PDCs. EGFR T790M aq was identified in only POST2 PDCs (Fig. B red asterisk); interestingly, BRAF mutations highly co-occurred with EGFR T790M aq mutations (Fig. B and Additional file , Fig. S3). Thus, our PDC models significantly reflected EGFR mutation, acquisition, and extinction status according to EGFR-TKI therapies. The four therapeutic groups exhibited distinct regulatory programs. To identify the resistance pathway, we applied two different methods. First, activated pathways were inferred from upregulated DEGs for each group (see Additional file , Table S5). The activation of ATM signaling and the DNA damage response according to CDK5 upregulation were enriched in the BASELINE PDCs. POST1 PDCs predisposed to RAS, IL-6, WNT, and PI3K-AKT-mTOR signals were enriched with upregulated genes ( BRAP , KRAS , and NF1 ). POST2 PDCs were enriched in NOTCH and STAT3 signaling, including DLL4 , EP300 , and STAT3 over-expression. Finally, POST3 PDCs exhibited activated TGF-β signaling regulated by PDK1 , SMAD2 , TGFBR3 , and ZEB1 . The over-expression of TGF-β signaling and ZEB1 demonstrated that osimertinib-resistant cancer exhibited EMT pathway activation (Fig. C). We next assessed the activities of 12 knowledge-based therapeutic resistance pathways for each sample according to the four EGFR-TKI therapeutic groups (Additional file , Table S6). The SERPINE1 signature was only activated in the BASELINE PDCs (Additional file , Fig. S6). The WNT, ERBB2, and MET pathway scores were increased in POST2 PDCs but were decreased in POST3 PDCs. The YAP/TAZ and PI3K-AKT pathways were elevated in POST3 PDCs. MET and ERBB2 gene expression also showed positive correlations with activation in the POST2 PDCs ( R = 0.62), whereas the POST3 PDCs exhibited inactivation in both pathways (Fig. D). The EMT-like POST3 PDCs exhibited the over-expression of both YAP/TAZ and AXL ( R = 0.86; Fig. D). Among the four EGFR-TKI groups, we explored the sensitivity of POST3 PDCs to etoposide and XAV939 (Fig. E). To predict drugs for 27 PDCs, we prepared an additional drug screening dataset by including the extended PDC set without available sequencing data ( n = 70; Additional file , Fig. S7A). To make up for the lack of genomic profiles, we collected EGFR-TKI therapy and clinical NGS test medical records. We could assign these samples to the four groups (BASELINE, n = 12; POST1, n = 23; POST2, n = 17; POST3, n = 18). The extended PDC dataset showed remarkable concordance in patient categorization obtained using NGS results and medical records. Finally, POST3 PDCs (PDC P < 0.2, extended PDC P < 0.083; Fig. E) were sensitive to etoposide and XAV939. We found that POST3 PDCs exhibited higher sensitivity to XAV939 than etoposide. The machine-learning XAV939 gene signature revealed that the response to XAV939 also involved EMT-like molecular features (i.e., collagen formation and extracellular matrix organization; Additional file , Table S4). To evaluate the previous finding about XAV939 selectivity and pathways, we created two cell lines— EGFR -T790M aq (H1975) cells using POST2 PDCs and XAV939-resistant (H1975_OR3, H1975_OR4) cells using POST3 PDCs—with osimertinib exposure. Both cell lines acquired osimertinib resistance (FC > 1.7; Fig. F). TCGA pan-cancer and reverse transcription-quantitative polymerase chain reaction analyses (RT-PCR) revealed that osimertinib-resistant cells concurrently exhibited over-expression of YAP/TAZ target genes . Ten target genes, including YAP1 and AXL, were considerably upregulated in the osimertinib-resistant cell lines compared to that in parent H1975 cells (Fig. G). In particular, AXL was elevated with the highest FC (5.1–9.0), and YAP1 and AXL proteins showed consistent over-expression in osimertinib-resistant cell lines (Fig. H). Next, we identified osimertinib resistance-related molecular subtypes and cellular alterations in POST3 PDCs. Since POST3 PDCs were classified as an EMT-like subtype (Fig. C), we assessed the EMT features of osimertinib-resistant PDCs. Two osimertinib-resistant PDCs showed decreased E-cadherin and EpCAM (epithelial markers) and increased N-cadherin and Vimentin (mesenchymal markers; Fig. H) expression levels. Subsequently, using flow cytometry, we established that cell surface EpCAM expression was considerably decreased in these cell lines (Fig. I). We further conducted a migration assay to verify these EMT-associated molecular changes (Fig. J). As expected, the two osimertinib-resistant PDCs had greater migration ability than H1975 cells. Collectively, these findings demonstrated that the two successfully generated osimertinib-resistant cell lines were representative of POST3 PDCs. Next, we checked if XAV939 had similar inhibitory effects on POST3 PDCs and H1975, H1975_OR3, and H1975_OR4 cells. XAV939 was not cytotoxic to H1975 cells (AUC = 105.5); however, H1975_OR3 (93.2) and OR4 (87.8) cells were clearly more sensitive to XAV939 (Fig. K). Overall, we validated that osimertinib resistance facilitates YAP/TAZ and AXL activation, and an EMT-like phenotype. Thus, XAV939 may be used to treat osimertinib-resistant tumors. We established a PDC collection of 102 samples (National Cancer Center; see Additional file , Table S1) from patients with advanced lung cancer enrolled in this study (Fig. A). PDCs were primarily collected from pleural effusions (92.2%), and secondarily from pericardial effusions (4.9%), ascites (2.0%), and tissues (1.0%; Additional file , Table S1). The pathologic type was adenocarcinoma (ADC; 84.3%), SCLC (5.9%), and miscellaneous types (squamous cell carcinoma, 4.9%; sarcomatoid carcinoma, 3.9%; and not otherwise specified, 1%). The drug panel used for response screening included 48 anti-cancer compounds of seven classes targeting angiogenesis ( n = 3), the cell cycle ( n = 8), DNA damage ( n = 6), MAPK ( n = 3), PI3K/AKT/mTOR ( n = 3), protein tyrosine kinase ( n = 7), and others ( n = 18; see Additional file , Table S2). Among the 48 drugs, 16 were screened in all PDCs and 32 were screened in 38 PDCs. To investigate the genomic characteristics, we identified somatic mutations and CNVs using target-seq ( n = 98) and classified four molecular subtypes (C1–C4) from gene expression profiles ( n = 102; Additional file , Table S1). Next, we checked whether the PDCs effectively recapitulate the mutations and expression profiles of the solid tumors. Owing to the dominance of LUAD in our cohort (> 80%), we compared PDC gene expression profiles with TCGA-LUAD dataset ( n = 230). As expected, the PDC gene expression profile resembled with tumor samples and was separated from normal adjacent tissues (Fig. B) . Constitutive somatic gene mutations were similar in PDCs and TCGA samples. The recurrence of TP53 , RB1 , and BRAF mutations was highly preserved in both PDC and TCGA samples. The EGFR mutation frequency was higher in PDCs, whereas the recurrence of KRAS , KEAP1 , and STK11 mutations was lower than that of TCGA samples not shown in Fig. B. Thus, somatic mutations in TP53 (47%), EGFR (29%), and RB1 (8%) were frequently observed in PDC models (Fig. C). Moreover, MET (10%), CDK4 (6%), and MDM2 (6%) variants, as well as EML4 - ALK (4%) and CD74 - ROS (2%) fusion genes, were detected. Before in-depth analysis, the most frequent TP53 and EGFR mutations were additionally categorized according to selectivity or functionality . TP53 mutations were divided into hotspot (known as gain-of-function; 5.9%; R175, G245, R248, R249, R273, R282) and non-hotspot (unknown or loss-of-function; 41.2%) mutations. EGFR mutations were categorized into (1) TKI-targetable single variant (TARGET: L858R and exon 19 deletion = 19.6%), (2) T790M acquisition (T790M aq : TARGET and T790M = 2.9%), (3) non-other-specified (NOS) single variant except TARGET (4.9%), and (4) NOS acquisition (NOS aq ; TARGET and NOS = 2%; Table ) . PDC samples could be classified into four molecular subtypes using gene expression profile by NMF clustering. To extrapolate the clinical characteristics for each subtype, we interrogated patient clinical profile encompassing histologic type, survival, smoking, and EGFR-TKI therapy record (Fig. and Table ). To uncover regulatory program for each molecular subtype, variant enrichment, and pathway regulation scores were assessed from multi-omics profile. In brief, subtype C1 was associated with a good outcome shown in Fig. A, and was activated in inflammatory and IL6-JAK-STAT3 signaling pathways; subtype C2 was associated with a modest outcome, dominance of TP53 / EGFR wild-type, upregulation of epithelial-to-mesenchymal transition (EMT), and enrichment of osimertinib-resistant group (POST3) PDCs; subtype C3 was associated with the worst outcome, long-term smoking males, SCLC, TP53 hotspot mutation, MYC activation, and fasten G2M checkpoint; by contrast, subtype C4 exhibited the best survival, frequent TP53 non-hotspot mutation, the dominance of EGFR-TKI TARGET mutation, and NOTCH signaling activation (Fig. A, B). We also demonstrated the survival significance corresponding to C1–C4 subtype gene sets using additional transcriptome datasets ( n = 1587; see Fig. S1A-B in Additional file and Table S3 in Additional file ). Upregulation of the C3 gene set concurrently exhibited the worst outcomes ( P < 0.001 and hazard ratio [HR] = 2.6). When additionally demonstrating from known up-regulated genes of embryonic stem cell, we could observe the activation of human embryonic stem cell genes, and target genes’ upregulation of transcription factor MYC and SOX2 (Fig. S2 in Additional file ). The activation of C1 and C2 subtype genes significantly showed better prognosis ( P < 0.05 and HR < 0.79). Our subtype classification sustained a global prognostic signature for lung cancer. Finally, we could summarize the subtypes based on the following regulatory pathways: C1, inflammatory; C2, EMT-like; C3, stemness; and C4, EGFR -dominant. Sensitive drug candidates exhibited remarkable concordance with the previously identified regulatory pathways for each molecular subtype ( P < 0.05 and |log 2 fold change (FC)| < 0.2; FC was assessed to compare average drug AUC values between corresponding subtype and another group; Fig. C). The C1 inflammatory subtype was sensitive to only ruxolitinib (JAK1/2) that targets the JAT/STAT pathway (Fig. B-D). The C2 EMT-like subtype showed MAPK class drug resistance and sensitivity to dasatinib (angiogenesis and SRC inhibitor), cabozantinib (VEGFA inhibitor), miscellaneous class repotrectinib (ROS inhibitor), and XAV939 (WNT-TNKS-β-catenin inhibitor). The C3 stemness subtype was sensitive to the top five-ranked cell cycle inhibitors. The C4 EGFR -dominant subtype exhibited the strongest resistance to most drug classes except MAPK inhibitors (selumetinib and trametinib) and EGFR-TKIs (gefitinib and afatinib). Notably, targets of predicted drugs belonged to pathways that were found to be activated in each subtype (Fig. B). To interrogate drug candidates for variants, we tested the difference in drug responses of single and co-occurrent mutation cases ( P < 0.005 and |log 2 FC| < 0.2; Fig. A). Co-mutated cases were also explored using the Fisher’s exact test ( P < 0.25; see Additional file , Fig. S3). Unexpectedly, EGFR -TARGET mutated PDCs showed a relatively modest response to three EGFR-TKIs (afatinib P = 0.17, gefitinib P = 0.19, osimertinib P = 0.08). To uncover EGFR -mutated cells’ molecular features interfering with the EGFR-TKI response, we dissected TKI TARGET mutations according to our four molecular subtypes (Fig. B). The C4 EGFR -dominant subtype was the most sensitive to all EGFR-TKIs, and the C1 inflammatory subtype also showed an especially good response to afatinib and osimertinib. Mutated cases ( n = 2) of the C3 stemness subtype was insufficient for statistical test. Finally, mutated cases ( n = 6) of the C2 EMT-like subtype did not respond to any EGFR-TKIs. Additionally, both T790M aq and NOS aq ( EGFR R776G and I744M with L858R) PDCs were TKI-sensitive. In the EGFR NOS type mutations, G719A was observed, and it comprised 11.5% among the NOS mutations in previous NSCLC study, and patient-derived xenografts demonstrated that the mutation was resistant to osimertinib . The remaining NOS mutations excluded exon 18–21 and had a low possibility of finding a structure-based therapeutic target . Therefore, we concluded that the remaining NOS group mutations were not targetable by EGFR-TKIs. Especially, EGFR TARGET mutated PDCs classified to EMT-like subtype exhibited low response to EGFR-TKIs. Thus, our molecular subtype of EGFR mutations revealed that these PDC models can be used to verify the heterogeneous tumor environment affecting drug responses. When interrogating drug candidates for mutations, BRAF variants were sensitive to trametinib (Fig. A). The TP53 non-hotspot exhibited resistance to olaparib (PARP inhibitor), repotrectinib, and dasatinib. RB1 variants responded well to alpelisib. Although TP53 mutation was not associated with any drug response ( P < 0.28), RB1 / TP53 co-mutated samples exhibited certain sensitivity that was stronger than RB1 (Fig. A). RB1 / TP53 mutations were mostly enriched in SCLC ( n = 4; P < 0.001) but were also observed in NSCLC ( n = 3; Additional file , Fig. S3). We additionally investigated alpelisib sensitivity comparing PDCs with cell lines ( n = 70; SCLC n = 7). Notably, although alpelisib inhibited SCLC (Fig. C), RB1 / TP53 variants ( P < 0.001) were more strongly inhibited than SCLC ( P = 0.02). However, in contrast to RB1/TP53 ( P < 0.01), we failed to identify SCLC sensitivity to alpelisib ( P = 0.47; Fig. C) using PDCs. Meanwhile, another PIK3-AKT target drug capivasertib (AKT1) was sensitive in RB1 / EGFR -NOS and EGFR -NOS cases. When dissecting details, two of the five ( RB1 / EGFR -NOS) PDCs was classified to SCLC type additionally harboring TP53 mutations (Fig. S3B in Additional file ) . Therefore, we could infer that RB1 / EGFR -NOS PDC cases accompanied the dependency to SCLC type. Therefore, PI3K-AKT class drugs detected from SCLC or RB1 / TP53 cells also affect EGFR -NOS mutations. Collectively, mutation-based drug detection result emphasize that the genomic characteristics could predict both the cancer drug response and histological lung cancer type SCLC (Additional file , Table S1). Furthermore, our PDC model predicted alpelisib as an appropriate treatment for SCLCs, which was not identified in cell lines (Fig. C). To delineate the drug susceptibility of alpelisib from transcriptome, we extracted the transcriptome signature for each drug using machine-learning approaches. The response to alpelisib was modulated by a transcriptome signature enriched “MITOTIC G1 PHASE AND G1 S TRANSITION” pathway gene set (Fig. D, and Additional file , Table S4). Upregulation of MYC , E2F1 , and CCNB1 expression led to resistance to alpelisib. Previously, we uncovered that stem-like C3 subtype exhibited both MYC up-regulation, and SCLC enrichment (Additional file , Fig. S2). MYC facilitates SCLC tumorigenesis . Moreover, our SCLC PDCs predicted the response to alpelisib better than cell lines (Fig. C). Therefore, transcriptome gene signature extracted by machine-learning determined alpelisib response for SCLC like RB1 / TP53 . For further in-depth exploration of the candidate drugs for treating different lung cancer types, we investigated the drug responses of ADC, SCLC, and miscellaneous cancers. SCLC was sensitive to drugs targeting cell cycle pathways (adavosertib, barasertib, berzosertib, and AZD7762) and DNA damage (vorinostat; P < 0.1 and |log 2 FC| < 0.3; Fig. A). The SCLC sensitivity to alpelisib was relatively lower than that to cell cycle drugs according to the FC value. The machine-learning signatures of three drugs (adavosertib, AZD7762, and berzosertib) consistently contained cell cycle pathways (Fig. B, and Additional file , Table S4). The transcriptome profile revealed the activation of cell cycle genes in SCLC. Moreover, the DNA damage response and RAS signal were downregulated in NSCLC (Additional file , Fig. S4). Collectively, these comprehensive pharmacogenomics result suggest the role of cell cycle inhibition in treating SCLC types. PDCs and cell lines showing FOXM1 upregulation exhibited AZD7762 sensitivity (Fig. C, D; PDC R = –0.77; cell line R = –0.51). FOXM1 was clearly over-expressed in SCLC compared to that in LUAD (Fig. E) . To evaluate whether FOXM1 is a plausible target to AZD7762 for SCLC, we transfected H69 cells with small interfering RNAs (siRNAs) against FOXM1 (#6 and #7, siFOXM1) and control siRNA and measured the FOXM1 protein level using western blotting. The FOXM1 expression level was decreased in siFOXM1 (Fig. F). siFOXM1 cells also exhibited decreased cell proliferation compared to siControl-transfected cells. However, G2-M phase arrest was detected in siFOXM1 H69 cells (Fig. G). siFOXM1 became more resistant to AZD7762 (AUC FC > 1.1) than siControl-transfected cells (Fig. H). Another cell lines H209 also presented the same result (see Additional file , Fig. S5). Therefore, we concluded that FOXM1 plays essential role to SCLC participating in the G2-M phase, and can be targeted by cell cycle inhibitors. Among our PDCs, we identified 27 EGFR-TKI-treated cases with corresponding NGS results (independently obtained in the clinic using tissue samples). As described in methods, we these patients into four groups (Fig. A). Primarily, POST3 samples (86%) were enriched in the C2 EMT-like subtype (Table ). EGFR mutation-calling failure was observed in the clinical tumor biopsy NGS record of some patients from which BASELINE ( n = 2) and POST2 ( n = 1) PDCs were derived. The mutation-calling failure likely originated from differences in the platform between the clinical biopsy and laboratory PDC testing. Moreover, TP53 non-hotspot mutations were present in all of the POST2 PDCs and were markedly absent in POST3 PDCs. EGFR T790M aq was identified in only POST2 PDCs (Fig. B red asterisk); interestingly, BRAF mutations highly co-occurred with EGFR T790M aq mutations (Fig. B and Additional file , Fig. S3). Thus, our PDC models significantly reflected EGFR mutation, acquisition, and extinction status according to EGFR-TKI therapies. The four therapeutic groups exhibited distinct regulatory programs. To identify the resistance pathway, we applied two different methods. First, activated pathways were inferred from upregulated DEGs for each group (see Additional file , Table S5). The activation of ATM signaling and the DNA damage response according to CDK5 upregulation were enriched in the BASELINE PDCs. POST1 PDCs predisposed to RAS, IL-6, WNT, and PI3K-AKT-mTOR signals were enriched with upregulated genes ( BRAP , KRAS , and NF1 ). POST2 PDCs were enriched in NOTCH and STAT3 signaling, including DLL4 , EP300 , and STAT3 over-expression. Finally, POST3 PDCs exhibited activated TGF-β signaling regulated by PDK1 , SMAD2 , TGFBR3 , and ZEB1 . The over-expression of TGF-β signaling and ZEB1 demonstrated that osimertinib-resistant cancer exhibited EMT pathway activation (Fig. C). We next assessed the activities of 12 knowledge-based therapeutic resistance pathways for each sample according to the four EGFR-TKI therapeutic groups (Additional file , Table S6). The SERPINE1 signature was only activated in the BASELINE PDCs (Additional file , Fig. S6). The WNT, ERBB2, and MET pathway scores were increased in POST2 PDCs but were decreased in POST3 PDCs. The YAP/TAZ and PI3K-AKT pathways were elevated in POST3 PDCs. MET and ERBB2 gene expression also showed positive correlations with activation in the POST2 PDCs ( R = 0.62), whereas the POST3 PDCs exhibited inactivation in both pathways (Fig. D). The EMT-like POST3 PDCs exhibited the over-expression of both YAP/TAZ and AXL ( R = 0.86; Fig. D). Among the four EGFR-TKI groups, we explored the sensitivity of POST3 PDCs to etoposide and XAV939 (Fig. E). To predict drugs for 27 PDCs, we prepared an additional drug screening dataset by including the extended PDC set without available sequencing data ( n = 70; Additional file , Fig. S7A). To make up for the lack of genomic profiles, we collected EGFR-TKI therapy and clinical NGS test medical records. We could assign these samples to the four groups (BASELINE, n = 12; POST1, n = 23; POST2, n = 17; POST3, n = 18). The extended PDC dataset showed remarkable concordance in patient categorization obtained using NGS results and medical records. Finally, POST3 PDCs (PDC P < 0.2, extended PDC P < 0.083; Fig. E) were sensitive to etoposide and XAV939. We found that POST3 PDCs exhibited higher sensitivity to XAV939 than etoposide. The machine-learning XAV939 gene signature revealed that the response to XAV939 also involved EMT-like molecular features (i.e., collagen formation and extracellular matrix organization; Additional file , Table S4). To evaluate the previous finding about XAV939 selectivity and pathways, we created two cell lines— EGFR -T790M aq (H1975) cells using POST2 PDCs and XAV939-resistant (H1975_OR3, H1975_OR4) cells using POST3 PDCs—with osimertinib exposure. Both cell lines acquired osimertinib resistance (FC > 1.7; Fig. F). TCGA pan-cancer and reverse transcription-quantitative polymerase chain reaction analyses (RT-PCR) revealed that osimertinib-resistant cells concurrently exhibited over-expression of YAP/TAZ target genes . Ten target genes, including YAP1 and AXL, were considerably upregulated in the osimertinib-resistant cell lines compared to that in parent H1975 cells (Fig. G). In particular, AXL was elevated with the highest FC (5.1–9.0), and YAP1 and AXL proteins showed consistent over-expression in osimertinib-resistant cell lines (Fig. H). Next, we identified osimertinib resistance-related molecular subtypes and cellular alterations in POST3 PDCs. Since POST3 PDCs were classified as an EMT-like subtype (Fig. C), we assessed the EMT features of osimertinib-resistant PDCs. Two osimertinib-resistant PDCs showed decreased E-cadherin and EpCAM (epithelial markers) and increased N-cadherin and Vimentin (mesenchymal markers; Fig. H) expression levels. Subsequently, using flow cytometry, we established that cell surface EpCAM expression was considerably decreased in these cell lines (Fig. I). We further conducted a migration assay to verify these EMT-associated molecular changes (Fig. J). As expected, the two osimertinib-resistant PDCs had greater migration ability than H1975 cells. Collectively, these findings demonstrated that the two successfully generated osimertinib-resistant cell lines were representative of POST3 PDCs. Next, we checked if XAV939 had similar inhibitory effects on POST3 PDCs and H1975, H1975_OR3, and H1975_OR4 cells. XAV939 was not cytotoxic to H1975 cells (AUC = 105.5); however, H1975_OR3 (93.2) and OR4 (87.8) cells were clearly more sensitive to XAV939 (Fig. K). Overall, we validated that osimertinib resistance facilitates YAP/TAZ and AXL activation, and an EMT-like phenotype. Thus, XAV939 may be used to treat osimertinib-resistant tumors. Our PDC platform was established with a lower labor burden and offers advantages of cost effectiveness and fast data acquisition compared with other platforms developed using PDXs, organoids, and cell lines. These advantages could increase the case number and, consequently, guaranteed the statistical significance and tumor heterogeneity of the test data. Our PDC platform currently includes 327 lung cancer cases and exhibits 77.7% success rates for an average 16.6 days duration time. Moreover, the PDC collection from 2020 achieved even better performance with an 83.9% success rate and 14.5 days duration time. Our platform performance exceeded that of a previous lung cancer screening . These characteristics indicate that our platform is effective to operate in the medical field for patient-tailored precision medicine. Moreover, in EGFR-TKI therapeutic groups, we observed approximately 11% EGFR -mutation calling loss compared with the medical record and patient NGS profile. This is likely due to multiple factors contributing to differences among platforms, such as the biopsy type, tumor clonality, purity, and sequencing depth . Meanwhile, our results were demonstrated using lung cancer cell lines. In further study, we have a plan to evaluate experimental demonstration from immortalized cells generated from PDCs. When adjusting the platform pipeline in the next collection, we expect to improve the accuracy of pharmacogenomic analysis. Among molecular subtypes, stem-like type C3 exhibited the most significant up-regulation in SOX2 than other transcription factors. In previous reports, stemness-related transcription factors play a role to depend on cancer tissue type. SOX2 was required in early-stage ADC and SCLC, whereas OCT3, KLF4, and NANOG participate other cancer types . Despite low frequency of SCLC, these were also enriched in this subtype. RB1/TP53 co-mutation was SCLC driver gene, and its depletion facilitate aberrant cell-cycle . Additionally, MYC up-regulated in C3 type drives dynamic evolution of SCLC . Therefore, our molecular subtype successfully classified these comprehensive regulatory features. Through detailed EGFR-TKI response analysis, we showed that the PDC molecular subtype uncovers the clinical characteristics and resistance factors of lung cancer patients. Our subtype classification suggested therapeutic candidates according to regulatory program. In additional classification of EGFR TARGET mutations, we could uncover that molecular subtype implicated in EGFR-TKI responses of PDCs harboring EGFR mutations. In particular, the C2 subtype designated as EMT-like exhibited FGFR2 over-expression. The FGF7-FGFR2 over-expressing lung cancer-associated fibrosis (CAF) type robustly protects EGFR -mutated cancer cells to maintain osimertinib resistance . CAF and EGFR -mutated co-cultured cells exhibit osimertinib resistance. Here, using our PDC models, we could directly assess EMT interference via EGFR-TKI resistance from spontaneous patient tumor clonal status. Moreover, our data showed that both the EMT-like subtype and osimertinib-resistant patients are sensitive to XAV939. Machine-learning gene signatures also revealed that regulatory programs that induce EMT-like characteristics were sensitive to XAV939. The genomic variant RB1 / TP53 showed more sensitivity to alpelisib than SCLC, whereas SCLC exhibited sensitivity to cell cycle inhibitors. Interestingly, NSCLC to SCLC transformation co-occurred with EGFR-TKI resistance, and RB1 / TP53 loss-of-function mutation occurred earlier than expected in the cancer cell cycle . Upregulation of PIK3CA mutation and PI3K/AKT pathway genes also occurred earlier during the transformation . A patient-derived EGFR -mutant xenograft model verified that early PI3K/AKT pathway inhibition delays tumor growth in SCLC or NSCLC undergoing transformation. Importantly, our PDC model concurrently selected alpelisib, a PI3K/AKT inhibitor, for treating early-stage SCLC. Consequently, our results emphasize that progressive and de novo SCLC phenotypes are more sensitive to cell cycle inhibitors . We suggest that strategic combinatorial therapy with anti-cancer drugs belonging to two different classes can block the progression of early-stage SCLC. EGFR -mutated lung cancer patients responding well to TKIs eventually develop resistance. Particularly, osimertinib resistance develops via heterogeneous and complex mechanisms, making establishment of an effective therapeutic strategy difficult. Our therapeutic follow-up delineated resistance pathways activated by EGFR-TKI treatment . ERBB2 and MET were activated in POST1 PDCs. BRAF mutations significantly co-occurred with EGFR -T790M. Furthermore, YAP/TAZ was activated in POST4 PDCs. Our results suggest XAV939 for POST3 PDCs. Interestingly, BASELINE PDCs, similar to POST3 PDCs, responded to XAV939 (Additional file , Fig. S7B). Thus, first-line combinatorial therapy with EGFR-TKI and XAV939 seems plausible. Gefitinib and XAV939 acted synergistically in the combinatorial therapy of the EGFR -mutated H1975 cell line (combination index (CI) = 0.388; synergism CI < 0.9). We expect effective translation of these results into treatments for patients with osimertinib-resistant lung cancer. Precision oncology in lung cancer is mainly based on gene-targeted chemotherapy; however, evasive mutations in target genes confound the prognosis. The PDCs developed in this study offer an advantage in tailoring patient-specific drugs and understanding the comprehensive molecular features of cancer. Importantly, our molecular subtypes reflected the PDC heterogeneity and recapitulated the drug response-mediated interference of EMT. However, as a limitation, we observed mutation-calling failure in a few EGFR-TKI treatment cases (11%), which was attributed multiple factors such as biopsy differences, tumor clonality, purity, and sequencing depth . In the future, we expect to improve the accuracy of pharmacogenomic analysis by adjusting the platform pipeline. Our well-established pharmacogenomic platform effectively predicted drugs and response mechanisms for refractory lung cancer. Following this pilot study, we expect consistent use of the established PDC bank in unveiling comprehensive drug–target associations of clinical relevance. Additional file 1. Supplemental methods. Additional file 2: Table S1. Clinical information and molecular subtypes of 102 lung cancer patients used to obtain PDCs. Table S2. The 48 drugs employed in the screening panel. We refer to the chemical or generic name of the drugs, and the target and class for each drug were classified. Table S3. Differentially expressed gene sets for the four molecular subtypes. Table S4. GSEA results ( P < 0.1) using drug-associated gene signatures extracted with a machine-learning approach. Table S5. GSEA results ( P < 0.1) acquired from analysis of the upregulated DEGs from four EGFR-TKI treatment groups. Table S6. EGFR-TKI resistance pathway and related gene sets. Additional file 3: Fig. S1. Molecular subtype evaluation using six lung cancer cohorts ( n = 1587). (A) Overall survival plots for each molecular subtype according to the subtype gene signature score. High and low groups were selected from the upper and lower quartiles. P -values and hazard ratios (HRs) were calculated using the log-rank test and Cox model. (B) HR forest plots for each subtype across six lung cancer cohorts. Fig. S2. Additional assessment of stemness scores for each molecular subtype. (A) A heatmap of average stemness scores according to molecular subtype. Stemness scores were assessed using GSVA from gene signatures: embryonic stem (ES) cell up-regulated genes (ES exp1, and ES exp2), and five transcription factors’ target genes as well as four PRC2 complex target signatures as control sets. (B) Boxplots of stemness scores according to subtypes. P -values were calculated by Wilcoxon rank-sum test to compare C3 and others. Fig. S3. Co-occurrent mutation case investigation. (A) Significant co-mutation pairs extracted using the Fisher’s exact test (recurrence > 5%). (B) The status of RB1/TP53 mutation and SCLC type. Fig. S4. Heatmap of DEGs between SCLC and NSCLC. The rows of the heatmap are the samples and the columns show the genes. Genes were extracted using limma (adjusted P < 0.005, |log2 FC| > 0.5) and the GSEA results (Table S3, Additional file ) are summarized on the left. Fig. S5. Validation of the correlation of FOXM1 expression and AZD7762 sensitivity in H209 cells. (A) The protein levels of FOXM1 and β-actin in siRNA-transfected H209 cells (siFOXM1-#6, #7) analyzed using western blotting. Ctrl represents the control. H209 cells transfected with siRNAs were incubated for 48 h. (B) After incubation, the cells were analyzed using flow cytometry to evaluate the DNA content. Representative DNA content profiles from three independent experiments are shown. The graphs show the proportion of cells in each cell cycle phase. (C) Drug response curve of siRNA-transfected H209 cells treated with AZD7762 (x-axis). The area under the receiver operating characteristic curve (AUC) values of AZD7762 in siRNA-transfected cells are shown in the panel. All experiments were performed in quadruplicate. The values represent the mean ± SEM (Student’s t-test, * P < 0.05; *** P < 0.001). Fig. S6. Pathway score bar plots of six EGFR-TKI resistance signatures according to four EGFR-TKI therapeutic groups. The numbers on each bar plot indicate the P values obtained using the Wilcoxon rank-sum test. Fig. S7. Drug candidates extracted from EGFR-TKI group PDCs ( n = 27) and extended EGFR PDCs ( n = 70). (A) Volcano plots for EGFR-TKI groups of two datasets. The x-axis indicates the log2-fold change between drug responses and the y-axis shows the log-scale adjusted P -value. Red circles indicate sensitivity whereas blue circles indicate resistance. (B) Bar plots for etoposide and XAV939 AUC in both datasets. The x-axis indicates EGFR-TKI groups and the y-axis indicates AUC values. P values were obtained using the Wilcoxon rank-sum test. |
Is Postoperative Nasal Stenting Necessary After Primary Cleft Lip and Nose Repair? | 5a47f9ef-62b1-4bd1-aa94-8e5ed3374de2 | 11830958 | Surgical Procedures, Operative[mh] | There is variability in surgeon's preferences for how the cleft lip nasal deformity is treated as part of the primary cleft repair (cheilorhinoplasty), which includes no universally accepted protocol for nostril retainer stents. Some surgeons perform a primary rhinoplasty at the time of cleft lip repair, with little consensus in using open versus endonasal approaches, early septal repositioning, and degree of overcorrection in lower lateral cartilages repositioning. Following primary cleft lip and nose repair, some relapse of the cleft nasal deformity is common due to cartilage memory and contraction. Surgeons who advocate for postoperative nasal stenting argue that the soft tissues can be conformed to the desired shape. Ideally, long‐lasting aesthetic and functional cleft lip and nasal outcomes could be achieved with codified surgical choices with or without nasal stenting. Due to the variance in the types of nasal stents (Fig. ), duration of use, and surgical technique, there is an opportunity to outline the indications for nasal stents following primary lip and nose repair. Numerous studies report generally favorable outcomes with the use of nostril stents following primary cleft lip and nose repair; however, nearly all reports are based on empiric assessments of nasal anatomy and symmetry. Despite a paucity of evidence‐based studies, two retrospective reports demonstrate favorable nasal outcomes based on objective measurements in comparison with a control group. Yeow et al. compared two groups of 3‐month‐old infants ( n = 30) who were either treated with nasal stents for 6 months or no stenting after repair of a unilateral complete cleft lip with primary rhinoplasty. NAM was not used. Photographs were graded at 5–8 years of age, demonstrating improved nasal appearance, symmetry, alar cartilage slump, alar base level, and columella tilt in the nostril stent group compared with controls. In a more recent 2022 comparative retrospective study, Al‐Qatami et al. divided 50 infants undergoing unilateral cleft lip repair into a stenting and no stenting group. In contrast to the study by Yeow et al., all individuals also completed preoperative NAM, followed by primary lip repair and rhinoplasty at 3 months of age. Cephalometric measurements were compared on nasal casts created during the palatoplasty. Most of the specific nasal measurements and nasal symmetry were improved in the nasal stenting group. A systematic review of the literature on the role of postoperative nasal stents following cleft rhinoplasty was published in 2023 by Nguyen et al. Nine studies included 269 patients with nasal stent use following primary cleft rhinoplasty. Unfortunately, none of the included studies were of high‐level evidence. There was heterogeneity among the materials used for stent fabrication, with some repurposing durable medical equipment (e.g. nasal oxygen cannulas), but the majority utilized prefabricated, commercially produced nasal stents. Regarding the stent duration recommended, most studies report at least 6 months of use. The aggregate of data presented in this review suggests that nasal stenting is safe with a low rate of minor complications and a likelihood of favorably preserving nasal shape and symmetry. As cleft surgeons, we recognize the burden that the parents face in keeping nasal stents in for weeks or months. To address this concern of parental quality of life, a survey study by Hennocq et al. asked parents ( n = 72) to recall their preoperative understanding of the stents and the infant's tolerance of the nasal stents after surgery. The survey results did not find any negative effect on quality of life, yet in practice, the ability of the caregiver to commit to the stent placement is imperative and should be considered. The evidence suggests that postoperative nasal stents are beneficial if used for 3–6 months by a hypothesized mechanism of hindering the contraction and regression of the surgically repositioned lower lateral cartilages. The advantages of stents must also be contrasted with the potential for parental stress in caring for the nasal stents, feeding difficulties with obstructed stents and a small risk of complications, such as pressure‐related skin injury. Although the duration of nasal stent use after primary cleft rhinoplasty was between 3 and 6 months in the above studies, partial benefit is likely gained from shorter stent durations, which the surgeon can consider in the context of the cleft lip nasal deformity severity, use of preoperative NAM, and nasal overcorrection performed with primary rhinoplasty techniques. Nguyen et al. is a level 2 study. Murali et al., Yeow et al., Al‐Qatami et al., and Hennocq et al. are level 3 studies. |
The Effect of Exosomes Derived from Human Umbilical Cord Mesenchymal Stromal/Stem Cells on the Regeneration of Human Pulpectomized Tooth: A Case Report | 1a429497-1b65-4562-93f5-7db6fd690a79 | 11829066 | Dentistry[mh] | The dental pulp is the innermost part of the tooth, comprising connective tissue, blood vessels, and nerves, which are essential for maintaining the tooth healthy and functional. When the dental pulp becomes infected or necrotic due to caries, trauma, or other pathologies, conventional treatment options include root canal therapy (RCT) or pulpectomy. These procedures aim to remove damaged pulp tissue, preventing further infection and restoring tooth function. However, removing pulp tissue also eliminates the tooth’s vitality, which may compromise its structural integrity and increase the risk of fracture over time. A novel treatment approach is regenerative endodontic procedures (REPs). The field of REPs aims to overcome these limitations by promoting pulp tissue regeneration, thereby preserving the vitality of the teeth and increasing their longevity. Traditional approaches to pulp regeneration included scaffolds, growth factors, and stem cells. Among these, mesenchymal stromal/stem cells (MSCs) have shown significant promise due to their multipotency, ability to differentiate into various cell types, and immunomodulatory properties. Human umbilical cord mesenchymal stromal/stem cells (hUCMSCs) are particularly attractive for regenerative applications. Unlike MSCs derived from bone marrow or adipose tissue, hUCMSCs are harvested from the umbilical cord, which is a medical waste product, making them readily available and ethically non-controversial. Additionally, hUCMSCs have a robust proliferative capacity and low immunogenicity, which makes them ideal for therapeutic utilization. Recent research has focused on the paracrine effects of MSCs, which are mediated through the secretion of extracellular vesicles, specifically exosomes. Exosomes are nano-sized vesicles (30-150 nm in diameter) that facilitate intercellular communication by delivering bioactive molecules, such as proteins, lipids, and various RNAs. These vesicles can modulate the behavior of recipient cells, promoting processes such as cell proliferation, differentiation, and angiogenesis, which are critical for tissue regeneration. This case report aimed to investigate the effect of exosomes derived from hUCMSCs on the regeneration of human pulpectomized teeth. This innovative approach was investigated to provide preliminary evidence supporting the feasibility and potential efficacy of exosome therapy in dental pulp regeneration. The findings of this study could pave the way for future research and clinical applications, ultimately contributing to the development of more effective treatments for endodontic diseases. This project was conceived and executed as a joint project in Bushehr (Iran), Shiraz (Iran), and Aktobe (Kazakhstan) from 2022 to 2024. This case report described the treatment of a 40-year-old man, who was diagnosed with irreversible pulpitis in a mandibular second premolar. The study was conducted in accordance with the Local Bioethical Committee of NJSC “West Kazakhstan Marat Ospanov Medical University” (registration No. 1,1/01). Once written informed consent was obtained, the patient, who met the inclusion criteria, received treatment. The inclusion criteria were as follows: participants aged 20–55 years with a clinical diagnosis of irreversible pulpitis in a single root canal, no tooth fractures, and periapical radiographs showing no radiolucency. The exclusion criteria included infections, severe diseases, pregnancy, mental illness, and allergies that would interfere with cone beam computed tomography (CBCT) imaging. Blood and urine tests confirmed the absence of infections or disqualifying conditions. The hUCMSCs were obtained from a commercial cell bank and cultured under optimal conditions. After achieving confluence, the cells were prepared for exosome isolation using a commercial Exovista kit (PerciaVista R&D Co., Iran). The exosomes were characterized with scanning electron microscopy (SEM, TESCAN MIRA3, TESCAN Co., Czech Republic) and transmission electron microscopy (TEM, Zeiss EM10C transmission electron microscope, Zeiss Co., Germany), which revealed sizes ranging from 20 to 180 nm, with a mean diameter of 101.06 nm . This characterization confirmed their suitability for use in dental pulp regeneration. The patient had a pulpectomy under local anesthetic during the first treatment session, followed by conventional root canal preparation . The canal was irrigated and filled with temporary material. Then, the temporary infill was removed, and the canal was thoroughly cleaned during the second session, which was two weeks later. Afterward, a mixture of chitosan powder and hUCMSC-derived exosomes was placed into the root canal and sealed with glass ionomer cement. The patient underwent follow-up assessments at 1, 2, 4, 12, 16, and 24 weeks post-treatment. Clinical examinations, such as percussion, palpation, and pulp vitality tests, revealed normal responses throughout the follow-up period, with no signs of infection or swelling. Radiographic assessments revealed initial signs of periapical radiolucency and periodontal ligament widening, which healed gradually over time. CBCT imaging indicated that the exosome-chitosan mixture was well integrated into the tooth structure, with no missing canals or adverse reactions . The patient reported no pain, indicating that the tissue had successfully regenerated and healed. This study investigated the therapeutic potential of exosomes derived from human umbilical cord mesenchymal stromal cells (hUCMSCs) for dental pulp regeneration. While previous research primarily focused on cell-based therapies using dental pulp stromal/stem cells (DPSCs) combined with granulocyte colony-stimulating factor (G-CSF), this study introduced exosome-based therapy as an innovative approach in regenerative endodontics. In contrast to earlier studies, which employed hUCMSCs directly, , this research used hUCMSCs-derived exosomes in combination with chitosan to treat pulpectomized teeth. The treatment resulted in successful pulp regeneration with no adverse reactions, as confirmed by clinical assessments and radiographic evaluations. Despite the presence of radiolucency and periodontal ligament (PDL) widening, which indicated an active healing process, no signs of inflammation or infection were found, implying effective tissue regeneration and integration. Several studies supported the potential of exosomes in dental applications. Zhang and others showed that dental pulp-derived exosomes facilitated neovascularization and collagen deposition. Zhuang and colleagues demonstrated that exosomes from stem cells promoted dentin-pulp complex regeneration. Swanson and others confirmed that exosomes induce dentinogenesis when combined with a biodegradable polymer. Exosomes have also been demonstrated to prevent apoptosis, promote cell proliferation, and enhance angiogenesis, making them promising candidates for regenerative therapies. Despite these encouraging findings, the limitations of the present study must be acknowledged. Firstly, it was based on a single case report, underscoring the need for further research with larger sample sizes and longer follow-up periods. The exact mechanisms through which exosomes promote tissue regeneration remain unclear, necessitating further investigation of the molecular pathways involved. Thus, it is recommended that future studies concentrate on improving treatment protocols to maximize clinical outcomes and gain a better understanding of the long-term effects of exosome-based therapies for pulp vitality. This study highlighted the potential of hUCMSCs-derived exosomes as a promising alternative to cell-based therapies for dental pulp regeneration. However, more extensive research is required to confirm their efficacy. |
SEOM-GEMCAD-TTD clinical guideline for the diagnosis and treatment of gastric cancer (2023) | 4bd6affc-9208-4e88-bf54-85c9710f1811 | 11467061 | Internal Medicine[mh] | Gastric cancer (GC) incidence has been declining somewhat worldwide in recent decades. Nonetheless, it is the fifth most common cancer and the fifth leading cause of cancer deaths . In Spain, the incidence is lower and GC represents the tenth most frequent tumor and the seventh cause of cancer mortality . In contrast, gastro-esophageal junction cancer (GEJC) incidence is on the rise, especially in Western countries, probably due to changes in risk factors and more precise anatomic location registers. The median age at the time of diagnosis is 70 years; nevertheless, 15% are early onset GC and this subset is trending upward. Incidence differs by sex (men are 2–3 times more susceptible than women) and geographical location (incidence rates are higher in SE Asia, South America, and Eastern European countries). The risk factors, background, and tumor biology differ based on tumor location: H pylori infection has been clearly correlated to distal or antral GC. These tumors follow a pattern of stepwise progression (known as the Correa Cascade , from normal mucosa to non-atrophic gastritis, atrophic gastritis with and without intestinal metaplasia (IM), dysplasia and, finally, to cancer. Smoking is a risk factor for which there is great evidence. Other risk factors with an intermediate level of evidence include diet (low in fruits and vegetables, excessive salt, high consumption of processed meats), atrophic gastritis, and autoimmune conditions. In GC involving the gastric fundus or body, alcohol and EBV infection have been correlated as risk factors. GEJC is associated with obesity and gastroesophageal reflux, the incidence of which is increasing in Western countries. Some 10% of subjects with GC have a positive family history and > 3% are related to hereditary syndromes, the most common one being hereditary diffuse gastric cancer (HDGC) syndrome, caused by cadherin 1 gene (CDH1) alterations. In terms of primary prevention, there is consistent evidence that eradication of H. pylori in healthy individuals and in patients with gastric atrophy, significantly lowers the future incidence of GC . In high-risk East Asian countries, endoscopy-based screening programs have been implemented and resulted in higher detection rates of early-stage GC, with a substantial reduction in mortality. Diagnosis should be made on the basis of a gastroscopic or surgical biopsy reviewed by an experienced pathologist, and histology should be reported according to the WHO criteria. Upper GI endoscopy is mandatory not only to describe the details of the anatomic location for surgical purposes, but also to be able to perform a direct biopsy. Diagnostic accuracy is 70% when made on the basis of a single biopsy and as high as 98% when several biopsies are taken. Multiple biopsies should be carried out to provide sufficient material for histological and molecular interpretation, particularly in the setting of ulcerated lesions. Approximately 90% of GC are adenocarcinomas (Ac), which are subdivided into diffuse and intestinal (Lauren classification) depending on histological appearance. Thoracic and abdomino-pelvic CT scan is the preferred examination for staging and is highly accurate in detecting metastasis; albeit less sensitive for evaluating T and N spread. Echo-guided ultrasound (EUS) is more consistently accurate than CT for the diagnosis of malignant lymph nodes and is also indicated in very early stages to identify patients in whom mucosectomy can be offered. The use of PET-CT scan in GEJC can improve staging by detecting the lymph nodes involved or metastatic disease. Furthermore, the diffuse Ac subtype, very common in GC, displays poor FDG uptake, and PET-CT has not been proven to improve N staging and reveals a 7% rate of false negative results . Peritoneal metastases are present in almost 20% of GC at diagnosis and laparoscopy ± peritoneal lavage for malignant cells is recommended in all stage IB-III GC and in cT3-4 Siewert III GEJC, all of which are considered potentially resectable, to exclude radiologically occult peritoneal disease . Surgery for GC entails high morbidity. Nutrition, body mass index (BMI), sarcopenia, hypercoagulation, comorbidities, age, and ECOG are predictive factors for mortality and outcomes in this setting. They are also predictive of overall survival (OS) in patients with advanced disease treated with systemic therapies. Therefore, they must be factored into the decision-making process prior to initiating any treatment strategy in all settings. A multidisciplinary evaluation, including a proper nutritional evaluation, inclusion of the patient into a prehabilitation programs, and geriatric assessment should be weighed based on each patient’s profile. Special efforts to assure nutritional intake must always be offered for all stages. The acquisition of high-quality samples and coordination with anatomical pathology laboratories for processing is critical for biomarker/molecular assessments acquiring. The biomarkers currently under consideration include HER2 overexpression or amplification, mismatch repair (MMR) or microsatellite instability (MSI) determination, and PD-L1 Combined Positive Score (CPS). Using both the histological grade and subtype as defined by the Lauren classification is useful, since they provide insights as to prognosis and predict treatment response and, consequently, their inclusion in clinical assessments is recommended. HER2 status is determined by immunohistochemistry (IHC) or in situ hybridization (ISH) techniques, applied to both biopsied material and surgical specimens . A result that is commonly considered as positive is 3 + on IHC or 2 + on IHC + fluorescent in situ hybridization (FISH) positive. For endoscopic samples, a minimum of five tumor fragments is required, ideally between 6 and 8 fragments, to control for potential false negatives. HER2 positivity ranges from 5 to 25%, depending on location and subtype. Approximately 8–25% of all patients exhibit high MSI or deficient mismatch repair (dMMR). To detect dMMR status, the first choice is IHC to determine the expression of the four genes involved in DNA MMR (MLH1, PMS2, MSH2, and MSH6) [I, A]. It is important to note that MLH1 and PMS2 are typically lost together, as are MSH2 and MSH6, although they can also be lost independently. In cases of uncertainty, confirmation via PCR to determine MSI-high status is recommended . The presence of MSI is a marker for dMMR and is characterized by a hypermutated state of neoplastic cells; likewise, it is also important to contemplate referral to a geneticist for assessment if Lynch syndrome is suspected. These determinations should be performed at diagnosis to plan the therapeutic approach. PD-L1 CPS by IHC is used to evaluate therapy with checkpoint inhibitors as first-line. CPS evaluates positivity in tumor cells, lymphocytes, and macrophages relative to the total number of viable tumor cells, and is expressed as a score. The cutoff points defined in clinical trials currently inform the clinical approach to be undertaken. In the immediate future, the biomarker repertoire will include claudin 18.2 determination by IHC. This biomarker is a member of the claudin protein family, which consists of membrane proteins expressed in tight epithelial junctions. This protein, in particular, plays a role in maintaining cell-to-cell barriers in epithelial tissues. Clinical trial data have often used a positivity criterion of ≥ 2 + in at least 75% of cells, a scenario observed in 38% of the cases, and even higher (42.3%) in HER2 positive tumors. Such positivity is more prevalent among younger patients with a diffuse subtype. In claudin 18.2 positive tumors, CPS is ≥ 5 in 17%. This biomarker and anti-claudin antibodies are still awaiting approval. Data from recent randomized controlled trials (RCT) suggest that other biomarkers, such as fibroblast growth factor receptor 2 (FGFR2) amplification (detected in 4–7% of cases) and FGFR2b overexpression (exhibited in 30% of the cases), might soon join the established biomarker panel. While EBV expression and tumor mutational burden (TMB) hold promise as indicators, more robust studies are warranted to confirm their clinical relevance. The usefulness of next-generation sequencing (NGS) in detecting neurotrophic tyrosine receptor kinase (NTRK) fusions remains dependent on the availability of effective therapies in a given region and is not yet deemed standard. ESCAT scores Endoscopic resection (ER) is considered for tumors at a very early stage that are very unlikely suitable for in bloc resection. This can be performed at high-volume medical centers with extensive experience in these techniques. Endoscopic resection could be performed through different techniques. The European Society of Gastrointestinal Endoscopy (ESGE) recommends endoscopic submucosal dissection (ESD) as the recommended therapy for most superficial gastric neoplastic lesions. ESD is indicated for well differentiated Ac, clinically staged as intramucosal (T1a), of any size if not ulcerated, and ≤ 30 mm if ulceration is present. Endoscopic mucosal resection (EMR) is an alternative for lesions ≤ 10 mm with low likelihood of malignancy. [III, B] Additionally, ESGE guidelines suggest that gastric Ac ≤ 30 mm, submucosal (sm1), and well-differentiated, or ≤ 20 mm, intramucosal, and poorly differentiated, both with no ulcerative findings, can be considered for ESD, albeit this decision should be made on a case-by-case basis. [III, C] A recent meta-analysis found that ER was correlated with fewer adverse events and shorter hospital stays compared to surgery, and, while ER was associated with lower complete resection rates and a higher risk of recurrence, the OS and 5-year cancer-specific survival were similar with both approaches . Post-ER, strict monitoring with an endoscopic follow-up program is mandatory to detect recurrences that can be treated early, either by endoscopy or surgery. Surgery Complete resection (ER) with negative margins is the treatment of choice for T1 tumors that do not meet criteria for endoscopic resection and for stage IB-III disease. [I, A] The type of resection, subtotal or total gastrectomy, and the extent of the margins depend on tumor location and histological subtype. The proper resection margin for ≥ T2 intestinal subtype tumors is considered to be 3 cm and 5 cm for those with diffuse histology. Routine splenectomy should not be performed in patients undergoing total gastrectomy for proximal gastric cancer, unless the spleen is involved or extensive hilar adenopathy is noted, as it increases operative morbidity without improving survival when compared to spleen-sparing surgery [I, A] . There is consensus concerning lymphadenectomy including a minimum of 15 nodes for reliable staging. Fifteen years follow-up data from the Dutch D1D2 trial demonstrated a significant decrease in locoregional recurrence and lower gastric cancer-related mortality after the D2 procedure, while the long-term survival analysis of the Italian randomized clinical trial exhibited improved disease-specific survival and GC-related mortality in those patients with advanced disease and lymph node metastases who underwent D2 lymphadenectomy . Based on these studies, D2 lymphadenectomy is accepted as the standard of care in Western countries in teams with expertise and low postoperative morbidity. [I, A] Multimodal rehabilitation programs (ERAS, Enhanced Recovery After Surgery) include all aspects of optimal peri-operative care for patients undergoing gastrectomy. These measures have been shown to accelerate postoperative recovery, reduce the stress caused by surgery, and shorten hospital stay. Recommendations ER is an appropriate approach to treat very early GC in patients without suspected lymph node involvement and that meet the criteria for ER. [III, B]. The recommended approach for resectable GC is gastrectomy with D2 lymph node dissection performed by experienced surgeons in high-volume centers. [I, A]. Complementary chemotherapy Perioperative CT has been shown to significantly improve R0 rates, disease-free survival (DFS), and OS as opposed to surgery alone in two European, landmark phase III studies: the MAGIC trial , using six cycles (three before and three after surgery) of epirubicin-cisplatin-5-fluorouracil (ECF) and the FNCLCC/FFCD 9703 study, that evaluated a perioperative cisplatin-5-fluorouracil regimen (Table ). These results led to the adoption of perioperative CT as a standard approach for potentially resectable, clinical stage ≥ T2 GC or GEJ Ac, mainly in Europe and other Western countries. [I, A]. The German phase II–III AIO-FLOT4 study subsequently reported four cycles of pre- and post-operative FLOT (docetaxel, oxaliplatin, 5-fluorouracil, and leucovorin) to significantly prolong OS compared to the MAGIC schema (median of 50 vs. 35 months; HR = 0.77; p = 0.012) . Based on these data, perioperative FLOT has been established as the standard of care for patients eligible for radical treatment. [I, A] For those unfit patients, other options considering a platinum- and fluoropyrimidine-based doublet or a modified dose of the FLOT schema can be considered. [II, B]. Moreover, two recent phase III trials have confirmed the superiority of neoadjuvant CT over D2 surgery followed by adjuvant CT in Asian patients with node positive or T4 gastric/GEJ cancer , thereby providing evidence to support the use of this perioperative approach in East Asia, where adjuvant CT had historically been the complementary therapy of choice (Table ). The benefit of the pre-operative approach is based on the potential chemotherapeutic effect in terms of downstaging and eradicating micrometastatic disease, along with better treatment tolerance (compared to the post-operative approach). Adjuvant CT can be considered for those patients with stage ≥ IB GC who have undergone upfront potentially curative surgery, without prior neoadjuvant therapy (i.e., cases of uncontrollable bleeding and/or stenosis not amenable to palliative solutions), and who have had an adequate D2 lymphadenectomy with node-negative disease. [I, A] A meta-analysis published in 2010 suggested a benefit in OS with this approach, which has been widely used in Asian countries. The ACTC-GC and CLASSIC phase III Asian studies have evidenced that 1 year of S1 or 6 months of CAPOX adjuvant CT, respectively, are associated with statistically significant DFS and OS benefit compared to observation following D2 gastrectomy of stage II-III GC. More recently, the addition of docetaxel to S1 has yielded better results than S1 alone in terms of 3-year relapse-free survival (RFS) (HR = 0.632; p < 0.001) for stage III disease (Table ). Based on these data, the recommended regimen when adjuvant CT is considered in Western populations is a doublet containing fluoropyrimidine + oxaliplatin or docetaxel for a total duration of 6 months. Microsatellite instability (MSI) GC: post hoc analyses of prospectively conducted randomized trials have revealed the positive prognostic effect of MSI in surgically resected GC and the potential lack of benefit of perioperative or adjuvant CT in this population . Although evidence is limited, treatment for MSI-high patients who have undergone curative surgery should be individualized. Dual immune checkpoint inhibition (ipilimumab + nivolumab in the NEONIPIGA and durvalumab + tremelimumab in the INFINITY study) has also proven to be promising as neoadjuvant treatment in cases with MSI-H tumors (pCR rate of 58.6–60%) and highlights the possibility of delaying or even avoiding surgery in highly selected subgroups (Table ). Complementary chemoradiotherapy Preoperative chemoradiotherapy (CRT) could be considered as an acceptable therapeutic approach for patients with GEJ adenocarcinoma over surgery alone. The phase III CROSS trial demonstrated a significant survival benefit with preoperative CRT (weekly carboplatin plus paclitaxel, and 41.4 Gy for 5 weeks) compared to surgery alone. Nevertheless, this study mixed different tumor types and the survival benefit was higher for squamous cell carcinoma of esophagus, GEJ adenocarcinoma patients also obtained recurrence-free and OS benefit [Level I, Grade A]. However, new data regarding the benefit of perioperative chemotherapy (CT) vs preoperative CRT for GEJ adenocarcinoma was recently presented at ASCO annual meeting 2024 according to findings from the phase 3 ESOPEC trial (NCT02509286) favoring the FLOT strategy . ESOPEC was a prospective multicenter study that enrolled patients with esophageal adenocarcinoma across 25 sites in Germany. Eligible patients needed to be at least 18 years old, have received no prior abdominal or thoracic radiotherapy, have an ECOG performance status of 2 or less, and have adequate organ function. Patients were also required to have pretreatment stage cT1N + , M0 or cT2-4a, N0/ + , M0 disease. Patients were randomly assigned 1:1 to receive either the FLOT or CROSS protocol. Those in the FLOT arm received repeated doses of 5-fluorouracil, leucovorin, oxaliplatin, and docetaxel every 2 weeks over 4 neoadjuvant cycles prior to surgery and 4 adjuvant cycles after surgery. In the CROSS arm, patients received neoadjuvant radiation therapy and concurrent chemotherapy with carboplatin and paclitaxel 5 weeks prior to surgery. The primary end point was OS. Secondary end points included progression-free survival, recurrence-free survival, and postsurgical quality of life. Both arms were well balanced. The results showed, at a median follow-up of 55 months, patients who received the FLOT protocol ( n = 221) achieved a median OS of 66 months compared with 37 months among those who were treated with the CROSS protocol ( n = 217). The 3-year OS rates were 57% vs 51%, respectively and a median PFS of 38 vs 16 months respectively. Pathological completed remission was seen in 16.8 and 10% respectively. It’s important to note that only 67.7% of the patients included in the CROSS arm were able to complete the neoadjuvant strategy while 87.3 and 52.5% completed the pre-operative and post-operative FLOT respectively, Previous evidence included the meta-analysis by Petrelli showed that preoperative CRT did not significantly benefit in the risk of distant metastasis nor OS, although may have a positive impact in the risk of relapse, the risk of locoregional failure and higher pathological response and the data from the phase III Neo-AEGIS trial compared the administration of preoperative CRT (CROSS) with perioperative CT (MAGIC/FLOT) in patients with adenocarcinoma of the esophagus or GEJ. OS was non-inferior for perioperative CT (3y-OS 55%) compared with preoperative CT (3y-OS 57% and HR 1.01). Nevertheless, the majority of patients were treated with the old MAGIC schema, which is currently obsolete [Level I, Grade C] The question that stills open is if perioperative FLOT is better than the multimodal approach for high risk patients who received CRT followed by surgery, and adjuvant immunotherapy based on the data of the Checkmate 577 discussed in the next section. Postoperative chemoradiotherapy (CRT) has demonstrated survival benefit in patients with gastric and GEJ cancer over surgery alone. The old phase III INT-0116 trial reported a statistically significant benefit for both RFS (HR 1.51; p < 0.001) an OS (HR 1.32; p = 0.0046) with postoperative 5FU-based CRT compared to the observation . Concerns about the credibility of this study are reasonable considering that about 90% of patients had underwent to insufficient lymphadenectomies. Data from the Nordic CRITICS trial confirms the lack of benefit on post-operative radiotherapy, and approach that should be individualized only in those patients that had infratherapeutic lymphadenectomies (III, B) and R1 resections (III,C) . Results of the TOP-GEAR, and CRITICS-II (two randomized trials that are currently exploring the role of preopertaive chemo-radiotherapy in GC) are still pending. Complementary immunotherapy Two large, randomized studies examining the potential role of adjuvant immunotherapy in resectable esophagogastric Ac have yielded negative results (Table ). The ATTRACTION-5 study, a phase III trial comparing adjuvant CT + nivolumab to placebo + nivolumab in Asian patients with pathological stage III gastric/GEJ cancer, failed to meet the primary endpoint of RFS . Similarly, the VESTIGE phase II study indicated that adjuvant nivolumab/ipilimumab did not improve DFS compared to adjuvant CT in European patients with gastroesophageal Ac at high risk for recurrence (N + or R1) following CT + surgery . Interestingly, the VESTIGE trial confirmed the benefit of performing post-operative CT in all patients irrespective of the pathological response to pre-operative CT. In contrast to these results, the phase III CheckMate 577 trial has reported improved DFS with the addition of one year of nivolumab vs. placebo in a mixed pool of patients that aggregated squamous cell carcinoma and esophageal and GEJ Ac who had evidence of residual pathological disease following neoadjuvant CRT and surgery (11 versus 22.4 months; HR = 0.69; p < 0.001) . This results allowed the approval from EMA for this indication and it´s funded by the Spanish health system. Consequently, adjuvant nivolumab should only be prescribed in GEJC patients who have been treated with CRT followed by surgery, regardless of PD-L1 expression status. [I, A] The overall survival results are still pending. In the peri-operative setting, the results from phase II and III trials incorporating immunotherapy to perioperative CT in resectable gastric/ GEJ cancer have exhibited statistically significant improvements in pathological complete response rates and pTNM compared with chemo alone . The Keynote 585 study evaluated the addition of pembrolizumab to CT in the perioperative setting. The results of event-free survival (EFS) data have recently been reported. EFS was longer in the experimental arm (44.4 versus 25.3 months; HR 0.81 [95% CI, 0.67–0.99]; p = 0.0198); regardless, these results failed to achieve statistical significance. In consequence, no studies have yet evinced improvement as far as survival parameters are concerned at the moment this guideline is published. Other complementary systemic treatments The introduction of targeted therapies with proven clinical activity in resectable metastatic GC has paved the way to the concept of personalizing treatments in this setting. Regarding HER2-targeted approaches in HER2-positive early GC, trastuzumab in combination with perioperative FLOT has been assessed in the phase II HER-FLOT study , which reached its primary endpoint (pCR > 20%), achieving a pCR rate of 21.4%. The phase II-III AIO-PETRARCA study explored the role of dual HER2 inhibition with trastuzumab and pertuzumab in combination with FLOT compared to CT alone. The primary endpoint of the phase II portion of the trial was the pCR rate, which was achieved in 35% of the patients in the experimental arm vs. 12% in the control group ( p = 0.019), albeit at the expense of increased toxicity. Unfortunately, and due to the sponsor's decision, this study could not continue to phase III, which precluded evaluation of this strategy in terms of survival. In contrast to these results, the INNOVATION phase II study tested the inclusion of trastuzumab alone or with pertuzumab into perioperative CT. The trial failed to reach the primary endpoint of demonstrating the superiority of the double blockade, probably attributable to unacceptable toxicity. Interestingly, the addition of trastuzumab alone to the perioperative CT schedule proved to be effective. Follow-up data including survival are necessary to define the clinical value of this regimen. In relation to VEGF inhibitors, phase II studies have explored the activity of bevacizumab or ramucirumab in addition to perioperative CT, failing to improve the outcomes. Other pathways involving claudin18 isoform 2 (CLDN18.2), FGFR, and DNA damage repair proteins may be of interest in certain patients, but no trials in early-stage disease are available yet during this guideline development. Follow-up, long-term implications, and survivorship There are no randomized trials that address the survival benefit of surveillance strategies for resected gastric cancer patients. In a recent meta-analysis of 15 observational studies, planned surveillance did not achieve significantly better detection of recurrence or OS benefit for gastric cancer patients. Despite this, a follow-up suggestion is shown in Table . [Level III, Grade C]. Finally, gastric cancer patients should be continually monitored for nutritional status, with a special emphasis on iron and vitamin B12. Complete resection (ER) with negative margins is the treatment of choice for T1 tumors that do not meet criteria for endoscopic resection and for stage IB-III disease. [I, A] The type of resection, subtotal or total gastrectomy, and the extent of the margins depend on tumor location and histological subtype. The proper resection margin for ≥ T2 intestinal subtype tumors is considered to be 3 cm and 5 cm for those with diffuse histology. Routine splenectomy should not be performed in patients undergoing total gastrectomy for proximal gastric cancer, unless the spleen is involved or extensive hilar adenopathy is noted, as it increases operative morbidity without improving survival when compared to spleen-sparing surgery [I, A] . There is consensus concerning lymphadenectomy including a minimum of 15 nodes for reliable staging. Fifteen years follow-up data from the Dutch D1D2 trial demonstrated a significant decrease in locoregional recurrence and lower gastric cancer-related mortality after the D2 procedure, while the long-term survival analysis of the Italian randomized clinical trial exhibited improved disease-specific survival and GC-related mortality in those patients with advanced disease and lymph node metastases who underwent D2 lymphadenectomy . Based on these studies, D2 lymphadenectomy is accepted as the standard of care in Western countries in teams with expertise and low postoperative morbidity. [I, A] Multimodal rehabilitation programs (ERAS, Enhanced Recovery After Surgery) include all aspects of optimal peri-operative care for patients undergoing gastrectomy. These measures have been shown to accelerate postoperative recovery, reduce the stress caused by surgery, and shorten hospital stay. Recommendations ER is an appropriate approach to treat very early GC in patients without suspected lymph node involvement and that meet the criteria for ER. [III, B]. The recommended approach for resectable GC is gastrectomy with D2 lymph node dissection performed by experienced surgeons in high-volume centers. [I, A]. ER is an appropriate approach to treat very early GC in patients without suspected lymph node involvement and that meet the criteria for ER. [III, B]. The recommended approach for resectable GC is gastrectomy with D2 lymph node dissection performed by experienced surgeons in high-volume centers. [I, A]. Perioperative CT has been shown to significantly improve R0 rates, disease-free survival (DFS), and OS as opposed to surgery alone in two European, landmark phase III studies: the MAGIC trial , using six cycles (three before and three after surgery) of epirubicin-cisplatin-5-fluorouracil (ECF) and the FNCLCC/FFCD 9703 study, that evaluated a perioperative cisplatin-5-fluorouracil regimen (Table ). These results led to the adoption of perioperative CT as a standard approach for potentially resectable, clinical stage ≥ T2 GC or GEJ Ac, mainly in Europe and other Western countries. [I, A]. The German phase II–III AIO-FLOT4 study subsequently reported four cycles of pre- and post-operative FLOT (docetaxel, oxaliplatin, 5-fluorouracil, and leucovorin) to significantly prolong OS compared to the MAGIC schema (median of 50 vs. 35 months; HR = 0.77; p = 0.012) . Based on these data, perioperative FLOT has been established as the standard of care for patients eligible for radical treatment. [I, A] For those unfit patients, other options considering a platinum- and fluoropyrimidine-based doublet or a modified dose of the FLOT schema can be considered. [II, B]. Moreover, two recent phase III trials have confirmed the superiority of neoadjuvant CT over D2 surgery followed by adjuvant CT in Asian patients with node positive or T4 gastric/GEJ cancer , thereby providing evidence to support the use of this perioperative approach in East Asia, where adjuvant CT had historically been the complementary therapy of choice (Table ). The benefit of the pre-operative approach is based on the potential chemotherapeutic effect in terms of downstaging and eradicating micrometastatic disease, along with better treatment tolerance (compared to the post-operative approach). Adjuvant CT can be considered for those patients with stage ≥ IB GC who have undergone upfront potentially curative surgery, without prior neoadjuvant therapy (i.e., cases of uncontrollable bleeding and/or stenosis not amenable to palliative solutions), and who have had an adequate D2 lymphadenectomy with node-negative disease. [I, A] A meta-analysis published in 2010 suggested a benefit in OS with this approach, which has been widely used in Asian countries. The ACTC-GC and CLASSIC phase III Asian studies have evidenced that 1 year of S1 or 6 months of CAPOX adjuvant CT, respectively, are associated with statistically significant DFS and OS benefit compared to observation following D2 gastrectomy of stage II-III GC. More recently, the addition of docetaxel to S1 has yielded better results than S1 alone in terms of 3-year relapse-free survival (RFS) (HR = 0.632; p < 0.001) for stage III disease (Table ). Based on these data, the recommended regimen when adjuvant CT is considered in Western populations is a doublet containing fluoropyrimidine + oxaliplatin or docetaxel for a total duration of 6 months. Microsatellite instability (MSI) GC: post hoc analyses of prospectively conducted randomized trials have revealed the positive prognostic effect of MSI in surgically resected GC and the potential lack of benefit of perioperative or adjuvant CT in this population . Although evidence is limited, treatment for MSI-high patients who have undergone curative surgery should be individualized. Dual immune checkpoint inhibition (ipilimumab + nivolumab in the NEONIPIGA and durvalumab + tremelimumab in the INFINITY study) has also proven to be promising as neoadjuvant treatment in cases with MSI-H tumors (pCR rate of 58.6–60%) and highlights the possibility of delaying or even avoiding surgery in highly selected subgroups (Table ). Preoperative chemoradiotherapy (CRT) could be considered as an acceptable therapeutic approach for patients with GEJ adenocarcinoma over surgery alone. The phase III CROSS trial demonstrated a significant survival benefit with preoperative CRT (weekly carboplatin plus paclitaxel, and 41.4 Gy for 5 weeks) compared to surgery alone. Nevertheless, this study mixed different tumor types and the survival benefit was higher for squamous cell carcinoma of esophagus, GEJ adenocarcinoma patients also obtained recurrence-free and OS benefit [Level I, Grade A]. However, new data regarding the benefit of perioperative chemotherapy (CT) vs preoperative CRT for GEJ adenocarcinoma was recently presented at ASCO annual meeting 2024 according to findings from the phase 3 ESOPEC trial (NCT02509286) favoring the FLOT strategy . ESOPEC was a prospective multicenter study that enrolled patients with esophageal adenocarcinoma across 25 sites in Germany. Eligible patients needed to be at least 18 years old, have received no prior abdominal or thoracic radiotherapy, have an ECOG performance status of 2 or less, and have adequate organ function. Patients were also required to have pretreatment stage cT1N + , M0 or cT2-4a, N0/ + , M0 disease. Patients were randomly assigned 1:1 to receive either the FLOT or CROSS protocol. Those in the FLOT arm received repeated doses of 5-fluorouracil, leucovorin, oxaliplatin, and docetaxel every 2 weeks over 4 neoadjuvant cycles prior to surgery and 4 adjuvant cycles after surgery. In the CROSS arm, patients received neoadjuvant radiation therapy and concurrent chemotherapy with carboplatin and paclitaxel 5 weeks prior to surgery. The primary end point was OS. Secondary end points included progression-free survival, recurrence-free survival, and postsurgical quality of life. Both arms were well balanced. The results showed, at a median follow-up of 55 months, patients who received the FLOT protocol ( n = 221) achieved a median OS of 66 months compared with 37 months among those who were treated with the CROSS protocol ( n = 217). The 3-year OS rates were 57% vs 51%, respectively and a median PFS of 38 vs 16 months respectively. Pathological completed remission was seen in 16.8 and 10% respectively. It’s important to note that only 67.7% of the patients included in the CROSS arm were able to complete the neoadjuvant strategy while 87.3 and 52.5% completed the pre-operative and post-operative FLOT respectively, Previous evidence included the meta-analysis by Petrelli showed that preoperative CRT did not significantly benefit in the risk of distant metastasis nor OS, although may have a positive impact in the risk of relapse, the risk of locoregional failure and higher pathological response and the data from the phase III Neo-AEGIS trial compared the administration of preoperative CRT (CROSS) with perioperative CT (MAGIC/FLOT) in patients with adenocarcinoma of the esophagus or GEJ. OS was non-inferior for perioperative CT (3y-OS 55%) compared with preoperative CT (3y-OS 57% and HR 1.01). Nevertheless, the majority of patients were treated with the old MAGIC schema, which is currently obsolete [Level I, Grade C] The question that stills open is if perioperative FLOT is better than the multimodal approach for high risk patients who received CRT followed by surgery, and adjuvant immunotherapy based on the data of the Checkmate 577 discussed in the next section. Postoperative chemoradiotherapy (CRT) has demonstrated survival benefit in patients with gastric and GEJ cancer over surgery alone. The old phase III INT-0116 trial reported a statistically significant benefit for both RFS (HR 1.51; p < 0.001) an OS (HR 1.32; p = 0.0046) with postoperative 5FU-based CRT compared to the observation . Concerns about the credibility of this study are reasonable considering that about 90% of patients had underwent to insufficient lymphadenectomies. Data from the Nordic CRITICS trial confirms the lack of benefit on post-operative radiotherapy, and approach that should be individualized only in those patients that had infratherapeutic lymphadenectomies (III, B) and R1 resections (III,C) . Results of the TOP-GEAR, and CRITICS-II (two randomized trials that are currently exploring the role of preopertaive chemo-radiotherapy in GC) are still pending. Two large, randomized studies examining the potential role of adjuvant immunotherapy in resectable esophagogastric Ac have yielded negative results (Table ). The ATTRACTION-5 study, a phase III trial comparing adjuvant CT + nivolumab to placebo + nivolumab in Asian patients with pathological stage III gastric/GEJ cancer, failed to meet the primary endpoint of RFS . Similarly, the VESTIGE phase II study indicated that adjuvant nivolumab/ipilimumab did not improve DFS compared to adjuvant CT in European patients with gastroesophageal Ac at high risk for recurrence (N + or R1) following CT + surgery . Interestingly, the VESTIGE trial confirmed the benefit of performing post-operative CT in all patients irrespective of the pathological response to pre-operative CT. In contrast to these results, the phase III CheckMate 577 trial has reported improved DFS with the addition of one year of nivolumab vs. placebo in a mixed pool of patients that aggregated squamous cell carcinoma and esophageal and GEJ Ac who had evidence of residual pathological disease following neoadjuvant CRT and surgery (11 versus 22.4 months; HR = 0.69; p < 0.001) . This results allowed the approval from EMA for this indication and it´s funded by the Spanish health system. Consequently, adjuvant nivolumab should only be prescribed in GEJC patients who have been treated with CRT followed by surgery, regardless of PD-L1 expression status. [I, A] The overall survival results are still pending. In the peri-operative setting, the results from phase II and III trials incorporating immunotherapy to perioperative CT in resectable gastric/ GEJ cancer have exhibited statistically significant improvements in pathological complete response rates and pTNM compared with chemo alone . The Keynote 585 study evaluated the addition of pembrolizumab to CT in the perioperative setting. The results of event-free survival (EFS) data have recently been reported. EFS was longer in the experimental arm (44.4 versus 25.3 months; HR 0.81 [95% CI, 0.67–0.99]; p = 0.0198); regardless, these results failed to achieve statistical significance. In consequence, no studies have yet evinced improvement as far as survival parameters are concerned at the moment this guideline is published. The introduction of targeted therapies with proven clinical activity in resectable metastatic GC has paved the way to the concept of personalizing treatments in this setting. Regarding HER2-targeted approaches in HER2-positive early GC, trastuzumab in combination with perioperative FLOT has been assessed in the phase II HER-FLOT study , which reached its primary endpoint (pCR > 20%), achieving a pCR rate of 21.4%. The phase II-III AIO-PETRARCA study explored the role of dual HER2 inhibition with trastuzumab and pertuzumab in combination with FLOT compared to CT alone. The primary endpoint of the phase II portion of the trial was the pCR rate, which was achieved in 35% of the patients in the experimental arm vs. 12% in the control group ( p = 0.019), albeit at the expense of increased toxicity. Unfortunately, and due to the sponsor's decision, this study could not continue to phase III, which precluded evaluation of this strategy in terms of survival. In contrast to these results, the INNOVATION phase II study tested the inclusion of trastuzumab alone or with pertuzumab into perioperative CT. The trial failed to reach the primary endpoint of demonstrating the superiority of the double blockade, probably attributable to unacceptable toxicity. Interestingly, the addition of trastuzumab alone to the perioperative CT schedule proved to be effective. Follow-up data including survival are necessary to define the clinical value of this regimen. In relation to VEGF inhibitors, phase II studies have explored the activity of bevacizumab or ramucirumab in addition to perioperative CT, failing to improve the outcomes. Other pathways involving claudin18 isoform 2 (CLDN18.2), FGFR, and DNA damage repair proteins may be of interest in certain patients, but no trials in early-stage disease are available yet during this guideline development. There are no randomized trials that address the survival benefit of surveillance strategies for resected gastric cancer patients. In a recent meta-analysis of 15 observational studies, planned surveillance did not achieve significantly better detection of recurrence or OS benefit for gastric cancer patients. Despite this, a follow-up suggestion is shown in Table . [Level III, Grade C]. Finally, gastric cancer patients should be continually monitored for nutritional status, with a special emphasis on iron and vitamin B12. Systemic treatment for advanced GC, including locally advanced unresectable and metastatic disease, includes CT, antiHER2 therapy, immunotherapy, and other targeted therapies. First-line chemotherapy Early CT trials in advanced GC demonstrated how CT improved survival compared to best supportive care and later proved the benefit of CT combinations over single agent treatment. . Table summarize the magnitude of benefit from the different strategies that supports our clinical practice in the perioperative and advanced setting according to ESMO magnitude of clinical benefit scale ( https://www.esmo.org/guidelines/esmo-mcbs ) The current standard first-line CT combinations for advanced GC include a platinum compound and fluoropyrimidine doublet. The first backbone to be established was the combination of cisplatin and 5-fluorouracil administered by continuous infusion (5-FU c.i.), while subsequent phase III clinical trials incorporated oxaliplatin and capecitabine as alternative compounds to cisplatin and 5-FU c.i., respectively . For older or frail patients, oxaliplatin choice as well as the use of reduced doses of oxaliplatin-based CT combinations have exhibited better safety profile and, at least, comparable survival outcomes versus cisplatin and standard dosing, respectively . Triplet combination CT in advanced GC includes the addition of docetaxel or anthracycline to the platinum and fluoropyrimidine doublet, although at the expense of increased toxicity . First-line HER2-targeted therapies HER2 overexpression or amplification has been identified as a potential prognostic factor and is considered a valid therapeutic target. In the ToGA trial, the addition of the HER2-targeted monoclonal antibody trastuzumab to cisplatin and fluoropyrimidine (capecitabine or 5-FU c.i.) doublet CT significantly improved OS with low and manageable additional toxicity in patients with previously untreated HER2 positive locally advanced or metastatic GC or GEJ cancer compared to CT alone, specifically with increased efficacy in patients with tumors that were immunohistochemistry (IHC) 3 + regardless of fluorescence in-situ hybridization (FISH) status, or that were IHC 2 + and FISH positive . Thenceforth, trastuzumab plus a platinum-based drug and fluoropyrimidine have been regarded as the first-line standard of care treatment in patients with HER2 positive (IHC 3 + or IHC 2 + and FISH positive) advanced GC. Further attempts to increase antiHER2 therapy efficacy by adding pertuzumab, a humanized monoclonal HER2-targeted antibody that binds to a different epitope on the HER2 receptor protein than trastuzumab, to trastuzumab plus a platinum-based drug and fluoropyrimidine failed to demonstrate significant improvement in OS in the phase III JACOB trial. . On the other hand, the recently reported results of the phase III KEYNOTE-811 study revealed that the addition of pembrolizumab to the TOGA schema improved survival outcomes for those patients with PDL1 CPS ≥ 1. [I, A] This therapy strategy was approved by EMA for patients Her2 positive with PDL1 CPS ≥ 1 but is not funded by the Spanish government yet. First-line immunotherapy for HER 2 negative The use of immunotherapy as first-line treatment in combination with chemotherapy for gastric and GEJ Ad is guided by the PD-L1 expression status assessed by the CPS score. The first positive study supporting this strategy is based on the Phase III Checkmate 649 study, which evaluated the addition of nivolumab to an oxaliplatin and fluoropyrimidine doublet (capecitabine or 5-fluorouracil). The primary endpoint was the OS and PFS for patients with PD-L1 expression CPS ≥ 5. The results showed after a 24-month follow up, substantiated OS benefit (HR 0.70, 95% CI 0.61–0.81, p < 0.0001) and PFS (HR 0.70, 95% CI 0.61–0.81, p < 0.0001) with respect to CT. This strategy was approved by EMA for this subgroup, and is a treatment strategy funded by the Spanish government. As a secondary endpoint, the OS and PFS was also tested for all randomized showing also significant improvement with the incorporation of immunotherapy to the strategy. (HR 0.70, 95% CI 0.71–0.88) and progression-free survival (PFS) (HR 0.79, 95% CI 0.70–0.89) respectively. The results from 48 months of follow-up continue to support these results. The study included also an arm testing the combination of nivolumab and ipilumumab vs. CT which in this context did not achieve statistical significance . A second important study that supports the use of immunotherapy added to CT in the first line strategy for GC is the Keynote-859 trial which compared the addition of pembrolizumab to a platin + fluoropyrimidine-based CT doublet vs CT. The primary endpoint was the OS benefit, and secondary endpoints included the PFS, DOR, ORR and safety and with a significant improvement for the OS (HR 0.78, 95% CI 0.70–0.8, p < 0.0001) and PFS (HR 0.76, 95% CI 0.67–0.85, p < 0.0001) benefit with respect to both variables in the overall study population; the magnitude was greater in the CPS ≥ 1 subgroup (HR 0.73, 95% CI 0.64–0.83) and for PFS (HR 0.72, 95% CI 0.63–0.82), as well as in the CPS ≥ 10 group (OS: HR 0.64, 95% CI 0.52–0.77; PFS: HR 0.62, 95% CI 0.51–0.75) . Based on this benefit, this strategy got the EMA approval for patients with PD-L1 CPS ≥ 1 subgroup. (Not funded by the Spanish government yet). On the other hand, the Keynote 590 study evaluated in patients with adenocarcinoma and squamous cancer of the esophagus and esophagogastric junction (EGJ) the role of adding pembrolizumab to cisplatin and fluorouracil, with the primary objectives being overall survival (OS) in patients with squamous cell carcinoma (SCC) PD-L1 CPS ≥ 10; OS and PFS in patients with SCC; in all patients with PD-L1 CPS ≥ 10 and in all patients of the study, showing a benefit in OS and PFS for all primary objectives of the study, with better benefits for the population with CPS ≥ 10, therefore achieving approval of the EMA for this subgroup and it is a treatment that is funded by the Spanish national health system . Similarly, positive phase III studies have been reported when other anti-PD-1 agents, such as sintilimab (ORIENT-16) or tislelizumab (RATIONAL 305), have been added, although they have yet to be approved by the EMA. The presence of MSI-H or dMMR is highly important to define a subgroup who deeply benefit from immunotherapy. The participants with MSI-H or dMMR, displayed an OS benefit in the Checkmate 649 trial when nivolumab was added in 44 patients (median OS 38.7 months vs 12.3 months, HR 0.38, 95% CI 0.17–0.84) . Likewise, a meta-analysis of first-line and successive lines of therapy evidenced benefit in response rate (RR) when pembrolizumab was used, regardless of the treatment line in the subgroup of patients with MSI-H or cMMR . Recommendations The addition of nivolumab to an oxaliplatin and fluorouracil-based CT doublet to treat gastric and GEJ Ac is recommended in patients with a CPS ≥ 5 [I, A]. Incorporating pembrolizumab to a platin and fluorouracil doublet is recommended in subjects with gastric and GEJ Ac and a CPS ≥ 1 [II, A], albeit the benefit is greatest in the CPS ≥ 10 subgroup [I, A]. The addition of pembrolizumab in monotherapy or the combination of CT and nivolumab can be contemplated in individuals with MSI-H or dMMR [II, B]. Other new therapeutics in first-line Anti-Claudin 18.2 therapy is expected to be integrated into the therapeutic armamentarium before long. [I, A] The phase III Spotlight RCT compared FOLFOX plus zolbetuximab or placebo in tumors with moderate-to-strong claudin expression (in > 75% of tumor cells) . The primary endpoint was PFS. The trial pointed toward improved PFS from 8.67 to 10.6 months (HR 0.75, 95% CI 0.58–0.94, p = 0.006). Moreover, median OS rose from 15.5 months to 18.2 months (HR 0.75, 95% CI 0.60–0.93, p = 0.0053). Meanwhile, the phase III GLOW RCT largely confirmed these findings. The eligibility criteria and design mirrored the Spotlight trial, except for the use of CAPOX as the CT backbone, in combination with either zolbetuximab or placebo . The experimental group exhibited improved PFS with 8.2 months as compared to 6.8 months in the control arm (HR 0.68, 95% CI 0.54–0.86, p = 0.0007). Median OS also rose from 12.1 to 14.3 months (HR 0.77, 95% CI 0.61–0.96, p = 0.01). Furthermore, the anti-claudin strategy is amenable to alternative approaches. Notably, humanized claudin18.2-redirected CAR-T cells or anti-claudin 18.2 antibody drug conjugates (ADC) are promising strategies currently in the early phases of research. [III, C] . Awaiting validation, targeted anti-FGFR therapy is another approach likely to be adopted in the coming years. [II, B] As of now, in the phase II FIGHT trial that included individuals exhibiting FGFR2b overexpression or FGFR2 amplification were randomized to receive FOLFOX ± bemarituzumab. In the experimental arm, median PFS improved from 7.4 to 9.5 months (HR 0.68, 95% CI 0.44–1.04). Meanwhile, OS increased from 12.9 months to not yet reached (HR 0.58, 95% CI 0.35–0.95) . Should the phase III FORTITUDE 101 and 102 trials confirm a favorable risk–benefit profile, this strategy could expand the therapeutic arsenal. Other targeted therapies contingent on availability, such as NTRK inhibitors for fusions (e.g., entrectinib, larotrectinib), may be of interest in certain cases, subject to individual assessment. [II, B]. Role of surgery in limited metastatic disease Recently the results of the phase III IKF-575/RENAISSANCE have been communicated, This trial randomized 139 pts with limited-metastatic adenocarcinoma of the stomach or esophagogastric junction to be treated with chemotherapy/targeted therapy alone vs. chemotherapy/targeted therapy followed by radical surgical resection. The primary endpoint was overall survival and was not met due to increased early mortality in the surgery arm Pts with RPLN metastases only seemed to benefit most from the surgical approach (mOS, 30 vs.17 months; 5y OS 38% vs. 19%; still having increased early mortality), while pts showing no response to chemo (mOS, 13 vs. 22 months) or pts with peritoneal disease (mOS, 12 vs. 19 months) derived a detrimental effect . Second and subsequent lines Up to 50% of GC and GEJC patients are fit enough to receive a second line of treatment, fewer in the case of subsequent lines. The benefit of a second CT line has been widely demonstrated in terms of survival and quality of life. Standard options are paclitaxel, docetaxel, and irinotecan with equivalent efficacy, yet different toxicity profiles. [I, B] The addition of the anti-vascular endothelial growth factor receptor 2 antibody ramucirumab to paclitaxel improves effectiveness, according to the phase III RAINBOW trial [I, A]. When considering immunotherapy, the phase II KEYNOTE-158 basket trial proved the efficacy of pembrolizumab in previously treated advanced patients with dMMR/MSI-H tumors. For GC patients, ORR was 45.8% and median PFS was 11 months, with median OS not yet reached . Considering that such results have never been previously reported, pembrolizumab is the preferred treatment option in this setting. [II, A]. Upon progression to trastuzumab-based line of therapy, HER2-positive patients should be considered separately. While earlier trials with the combination of trastuzumab + lapatinib and with trastuzumab emtastine had been negative, the paradigm of these patients has changed with the incorporation of the ADC trastuzumab deruxtecan (T-Dxd). The randomized phase II trial conducted in Asia evinced the advantage of T-Dxd over CT . These results were confirmed in a Western population, in a single arm, phase II trial that evinced that T-Dxd yielded an ORR of 42%, median PFS of 5.6 months, and median OS of 12.1 months in HER2-positive GC and GEJC patients that had progressed to first-line trastuzumab-based treatment. [II, B]. The results of this last study made it possible for EMA approval for trastuzumab-deruxtecan following post-trastuzumab progression. Results of the current global phase III DESTINY-Gastric04 trial will establish the actual benefit of this drug in this setting. In the third-line setting, oral treatment with trifluridine-tipiracil has the strongest evidence, with proven efficacy for survival and time to ECOG/PS deterioration based on the phase III TAGS trial. [I, A]. Nevertheless, this treatment is currently not reimbursed in Spain. Other options include a taxane or irinotecan, depending on the second line therapy [II, A]. An algorithm proposed for second and subsequent lines is presented in Fig. . Special consideration must be paid to the need for close symptom monitoring and nutritional assessment during all GC and GEJC treatment lines. Close follow-up radiological imaging will early detect tumor progression and prevent potential treatment-related toxicities. Early CT trials in advanced GC demonstrated how CT improved survival compared to best supportive care and later proved the benefit of CT combinations over single agent treatment. . Table summarize the magnitude of benefit from the different strategies that supports our clinical practice in the perioperative and advanced setting according to ESMO magnitude of clinical benefit scale ( https://www.esmo.org/guidelines/esmo-mcbs ) The current standard first-line CT combinations for advanced GC include a platinum compound and fluoropyrimidine doublet. The first backbone to be established was the combination of cisplatin and 5-fluorouracil administered by continuous infusion (5-FU c.i.), while subsequent phase III clinical trials incorporated oxaliplatin and capecitabine as alternative compounds to cisplatin and 5-FU c.i., respectively . For older or frail patients, oxaliplatin choice as well as the use of reduced doses of oxaliplatin-based CT combinations have exhibited better safety profile and, at least, comparable survival outcomes versus cisplatin and standard dosing, respectively . Triplet combination CT in advanced GC includes the addition of docetaxel or anthracycline to the platinum and fluoropyrimidine doublet, although at the expense of increased toxicity . HER2 overexpression or amplification has been identified as a potential prognostic factor and is considered a valid therapeutic target. In the ToGA trial, the addition of the HER2-targeted monoclonal antibody trastuzumab to cisplatin and fluoropyrimidine (capecitabine or 5-FU c.i.) doublet CT significantly improved OS with low and manageable additional toxicity in patients with previously untreated HER2 positive locally advanced or metastatic GC or GEJ cancer compared to CT alone, specifically with increased efficacy in patients with tumors that were immunohistochemistry (IHC) 3 + regardless of fluorescence in-situ hybridization (FISH) status, or that were IHC 2 + and FISH positive . Thenceforth, trastuzumab plus a platinum-based drug and fluoropyrimidine have been regarded as the first-line standard of care treatment in patients with HER2 positive (IHC 3 + or IHC 2 + and FISH positive) advanced GC. Further attempts to increase antiHER2 therapy efficacy by adding pertuzumab, a humanized monoclonal HER2-targeted antibody that binds to a different epitope on the HER2 receptor protein than trastuzumab, to trastuzumab plus a platinum-based drug and fluoropyrimidine failed to demonstrate significant improvement in OS in the phase III JACOB trial. . On the other hand, the recently reported results of the phase III KEYNOTE-811 study revealed that the addition of pembrolizumab to the TOGA schema improved survival outcomes for those patients with PDL1 CPS ≥ 1. [I, A] This therapy strategy was approved by EMA for patients Her2 positive with PDL1 CPS ≥ 1 but is not funded by the Spanish government yet. The use of immunotherapy as first-line treatment in combination with chemotherapy for gastric and GEJ Ad is guided by the PD-L1 expression status assessed by the CPS score. The first positive study supporting this strategy is based on the Phase III Checkmate 649 study, which evaluated the addition of nivolumab to an oxaliplatin and fluoropyrimidine doublet (capecitabine or 5-fluorouracil). The primary endpoint was the OS and PFS for patients with PD-L1 expression CPS ≥ 5. The results showed after a 24-month follow up, substantiated OS benefit (HR 0.70, 95% CI 0.61–0.81, p < 0.0001) and PFS (HR 0.70, 95% CI 0.61–0.81, p < 0.0001) with respect to CT. This strategy was approved by EMA for this subgroup, and is a treatment strategy funded by the Spanish government. As a secondary endpoint, the OS and PFS was also tested for all randomized showing also significant improvement with the incorporation of immunotherapy to the strategy. (HR 0.70, 95% CI 0.71–0.88) and progression-free survival (PFS) (HR 0.79, 95% CI 0.70–0.89) respectively. The results from 48 months of follow-up continue to support these results. The study included also an arm testing the combination of nivolumab and ipilumumab vs. CT which in this context did not achieve statistical significance . A second important study that supports the use of immunotherapy added to CT in the first line strategy for GC is the Keynote-859 trial which compared the addition of pembrolizumab to a platin + fluoropyrimidine-based CT doublet vs CT. The primary endpoint was the OS benefit, and secondary endpoints included the PFS, DOR, ORR and safety and with a significant improvement for the OS (HR 0.78, 95% CI 0.70–0.8, p < 0.0001) and PFS (HR 0.76, 95% CI 0.67–0.85, p < 0.0001) benefit with respect to both variables in the overall study population; the magnitude was greater in the CPS ≥ 1 subgroup (HR 0.73, 95% CI 0.64–0.83) and for PFS (HR 0.72, 95% CI 0.63–0.82), as well as in the CPS ≥ 10 group (OS: HR 0.64, 95% CI 0.52–0.77; PFS: HR 0.62, 95% CI 0.51–0.75) . Based on this benefit, this strategy got the EMA approval for patients with PD-L1 CPS ≥ 1 subgroup. (Not funded by the Spanish government yet). On the other hand, the Keynote 590 study evaluated in patients with adenocarcinoma and squamous cancer of the esophagus and esophagogastric junction (EGJ) the role of adding pembrolizumab to cisplatin and fluorouracil, with the primary objectives being overall survival (OS) in patients with squamous cell carcinoma (SCC) PD-L1 CPS ≥ 10; OS and PFS in patients with SCC; in all patients with PD-L1 CPS ≥ 10 and in all patients of the study, showing a benefit in OS and PFS for all primary objectives of the study, with better benefits for the population with CPS ≥ 10, therefore achieving approval of the EMA for this subgroup and it is a treatment that is funded by the Spanish national health system . Similarly, positive phase III studies have been reported when other anti-PD-1 agents, such as sintilimab (ORIENT-16) or tislelizumab (RATIONAL 305), have been added, although they have yet to be approved by the EMA. The presence of MSI-H or dMMR is highly important to define a subgroup who deeply benefit from immunotherapy. The participants with MSI-H or dMMR, displayed an OS benefit in the Checkmate 649 trial when nivolumab was added in 44 patients (median OS 38.7 months vs 12.3 months, HR 0.38, 95% CI 0.17–0.84) . Likewise, a meta-analysis of first-line and successive lines of therapy evidenced benefit in response rate (RR) when pembrolizumab was used, regardless of the treatment line in the subgroup of patients with MSI-H or cMMR . Recommendations The addition of nivolumab to an oxaliplatin and fluorouracil-based CT doublet to treat gastric and GEJ Ac is recommended in patients with a CPS ≥ 5 [I, A]. Incorporating pembrolizumab to a platin and fluorouracil doublet is recommended in subjects with gastric and GEJ Ac and a CPS ≥ 1 [II, A], albeit the benefit is greatest in the CPS ≥ 10 subgroup [I, A]. The addition of pembrolizumab in monotherapy or the combination of CT and nivolumab can be contemplated in individuals with MSI-H or dMMR [II, B]. The addition of nivolumab to an oxaliplatin and fluorouracil-based CT doublet to treat gastric and GEJ Ac is recommended in patients with a CPS ≥ 5 [I, A]. Incorporating pembrolizumab to a platin and fluorouracil doublet is recommended in subjects with gastric and GEJ Ac and a CPS ≥ 1 [II, A], albeit the benefit is greatest in the CPS ≥ 10 subgroup [I, A]. The addition of pembrolizumab in monotherapy or the combination of CT and nivolumab can be contemplated in individuals with MSI-H or dMMR [II, B]. Anti-Claudin 18.2 therapy is expected to be integrated into the therapeutic armamentarium before long. [I, A] The phase III Spotlight RCT compared FOLFOX plus zolbetuximab or placebo in tumors with moderate-to-strong claudin expression (in > 75% of tumor cells) . The primary endpoint was PFS. The trial pointed toward improved PFS from 8.67 to 10.6 months (HR 0.75, 95% CI 0.58–0.94, p = 0.006). Moreover, median OS rose from 15.5 months to 18.2 months (HR 0.75, 95% CI 0.60–0.93, p = 0.0053). Meanwhile, the phase III GLOW RCT largely confirmed these findings. The eligibility criteria and design mirrored the Spotlight trial, except for the use of CAPOX as the CT backbone, in combination with either zolbetuximab or placebo . The experimental group exhibited improved PFS with 8.2 months as compared to 6.8 months in the control arm (HR 0.68, 95% CI 0.54–0.86, p = 0.0007). Median OS also rose from 12.1 to 14.3 months (HR 0.77, 95% CI 0.61–0.96, p = 0.01). Furthermore, the anti-claudin strategy is amenable to alternative approaches. Notably, humanized claudin18.2-redirected CAR-T cells or anti-claudin 18.2 antibody drug conjugates (ADC) are promising strategies currently in the early phases of research. [III, C] . Awaiting validation, targeted anti-FGFR therapy is another approach likely to be adopted in the coming years. [II, B] As of now, in the phase II FIGHT trial that included individuals exhibiting FGFR2b overexpression or FGFR2 amplification were randomized to receive FOLFOX ± bemarituzumab. In the experimental arm, median PFS improved from 7.4 to 9.5 months (HR 0.68, 95% CI 0.44–1.04). Meanwhile, OS increased from 12.9 months to not yet reached (HR 0.58, 95% CI 0.35–0.95) . Should the phase III FORTITUDE 101 and 102 trials confirm a favorable risk–benefit profile, this strategy could expand the therapeutic arsenal. Other targeted therapies contingent on availability, such as NTRK inhibitors for fusions (e.g., entrectinib, larotrectinib), may be of interest in certain cases, subject to individual assessment. [II, B]. Role of surgery in limited metastatic disease Recently the results of the phase III IKF-575/RENAISSANCE have been communicated, This trial randomized 139 pts with limited-metastatic adenocarcinoma of the stomach or esophagogastric junction to be treated with chemotherapy/targeted therapy alone vs. chemotherapy/targeted therapy followed by radical surgical resection. The primary endpoint was overall survival and was not met due to increased early mortality in the surgery arm Pts with RPLN metastases only seemed to benefit most from the surgical approach (mOS, 30 vs.17 months; 5y OS 38% vs. 19%; still having increased early mortality), while pts showing no response to chemo (mOS, 13 vs. 22 months) or pts with peritoneal disease (mOS, 12 vs. 19 months) derived a detrimental effect . Second and subsequent lines Up to 50% of GC and GEJC patients are fit enough to receive a second line of treatment, fewer in the case of subsequent lines. The benefit of a second CT line has been widely demonstrated in terms of survival and quality of life. Standard options are paclitaxel, docetaxel, and irinotecan with equivalent efficacy, yet different toxicity profiles. [I, B] The addition of the anti-vascular endothelial growth factor receptor 2 antibody ramucirumab to paclitaxel improves effectiveness, according to the phase III RAINBOW trial [I, A]. When considering immunotherapy, the phase II KEYNOTE-158 basket trial proved the efficacy of pembrolizumab in previously treated advanced patients with dMMR/MSI-H tumors. For GC patients, ORR was 45.8% and median PFS was 11 months, with median OS not yet reached . Considering that such results have never been previously reported, pembrolizumab is the preferred treatment option in this setting. [II, A]. Upon progression to trastuzumab-based line of therapy, HER2-positive patients should be considered separately. While earlier trials with the combination of trastuzumab + lapatinib and with trastuzumab emtastine had been negative, the paradigm of these patients has changed with the incorporation of the ADC trastuzumab deruxtecan (T-Dxd). The randomized phase II trial conducted in Asia evinced the advantage of T-Dxd over CT . These results were confirmed in a Western population, in a single arm, phase II trial that evinced that T-Dxd yielded an ORR of 42%, median PFS of 5.6 months, and median OS of 12.1 months in HER2-positive GC and GEJC patients that had progressed to first-line trastuzumab-based treatment. [II, B]. The results of this last study made it possible for EMA approval for trastuzumab-deruxtecan following post-trastuzumab progression. Results of the current global phase III DESTINY-Gastric04 trial will establish the actual benefit of this drug in this setting. In the third-line setting, oral treatment with trifluridine-tipiracil has the strongest evidence, with proven efficacy for survival and time to ECOG/PS deterioration based on the phase III TAGS trial. [I, A]. Nevertheless, this treatment is currently not reimbursed in Spain. Other options include a taxane or irinotecan, depending on the second line therapy [II, A]. An algorithm proposed for second and subsequent lines is presented in Fig. . Special consideration must be paid to the need for close symptom monitoring and nutritional assessment during all GC and GEJC treatment lines. Close follow-up radiological imaging will early detect tumor progression and prevent potential treatment-related toxicities. Recently the results of the phase III IKF-575/RENAISSANCE have been communicated, This trial randomized 139 pts with limited-metastatic adenocarcinoma of the stomach or esophagogastric junction to be treated with chemotherapy/targeted therapy alone vs. chemotherapy/targeted therapy followed by radical surgical resection. The primary endpoint was overall survival and was not met due to increased early mortality in the surgery arm Pts with RPLN metastases only seemed to benefit most from the surgical approach (mOS, 30 vs.17 months; 5y OS 38% vs. 19%; still having increased early mortality), while pts showing no response to chemo (mOS, 13 vs. 22 months) or pts with peritoneal disease (mOS, 12 vs. 19 months) derived a detrimental effect . Up to 50% of GC and GEJC patients are fit enough to receive a second line of treatment, fewer in the case of subsequent lines. The benefit of a second CT line has been widely demonstrated in terms of survival and quality of life. Standard options are paclitaxel, docetaxel, and irinotecan with equivalent efficacy, yet different toxicity profiles. [I, B] The addition of the anti-vascular endothelial growth factor receptor 2 antibody ramucirumab to paclitaxel improves effectiveness, according to the phase III RAINBOW trial [I, A]. When considering immunotherapy, the phase II KEYNOTE-158 basket trial proved the efficacy of pembrolizumab in previously treated advanced patients with dMMR/MSI-H tumors. For GC patients, ORR was 45.8% and median PFS was 11 months, with median OS not yet reached . Considering that such results have never been previously reported, pembrolizumab is the preferred treatment option in this setting. [II, A]. Upon progression to trastuzumab-based line of therapy, HER2-positive patients should be considered separately. While earlier trials with the combination of trastuzumab + lapatinib and with trastuzumab emtastine had been negative, the paradigm of these patients has changed with the incorporation of the ADC trastuzumab deruxtecan (T-Dxd). The randomized phase II trial conducted in Asia evinced the advantage of T-Dxd over CT . These results were confirmed in a Western population, in a single arm, phase II trial that evinced that T-Dxd yielded an ORR of 42%, median PFS of 5.6 months, and median OS of 12.1 months in HER2-positive GC and GEJC patients that had progressed to first-line trastuzumab-based treatment. [II, B]. The results of this last study made it possible for EMA approval for trastuzumab-deruxtecan following post-trastuzumab progression. Results of the current global phase III DESTINY-Gastric04 trial will establish the actual benefit of this drug in this setting. In the third-line setting, oral treatment with trifluridine-tipiracil has the strongest evidence, with proven efficacy for survival and time to ECOG/PS deterioration based on the phase III TAGS trial. [I, A]. Nevertheless, this treatment is currently not reimbursed in Spain. Other options include a taxane or irinotecan, depending on the second line therapy [II, A]. An algorithm proposed for second and subsequent lines is presented in Fig. . Special consideration must be paid to the need for close symptom monitoring and nutritional assessment during all GC and GEJC treatment lines. Close follow-up radiological imaging will early detect tumor progression and prevent potential treatment-related toxicities. |
Knowledge, attitude, and practice of cardiac rehabilitation among patients after coronary artery stenting | 7f9a156e-fe88-413f-bc35-9ad959a057c2 | 11821840 | Surgical Procedures, Operative[mh] | Coronary artery stenting represents a pivotal intervention in the management of coronary artery disease (CAD). This procedure involves the insertion of a small, expandable metal mesh tube into a narrowed or blocked coronary artery, thereby promoting the vessel’s patency and preventing recurrent ischemic events . The utilization of coronary artery stenting has revolutionized the field of interventional cardiology, and offers a minimally invasive alternative to conventional coronary artery bypass grafting (CABG) surgery. In the clinical practice, percutaneous coronary intervention (PCI) has emerged as a standard therapeutic modality, owing to its procedural simplicity, reduced invasiveness, and favorable short-term outcomes . According to data from the American Heart Association, around 600,000 coronary stent procedures are performed annually in the United States . In 2021, the total number of registered coronary intervention cases in mainland China reached 1,164,117, representing a 20.18% increase from 2020 . Concurrently, mortality rates associated with CAD have shown a significant decline of 36.4% from 2006 to 2016 . While stent implantation effectively relieves myocardial ischemia and prevents acute coronary events, it does not address the underlying risk factors and pathophysiological processes contributing to CAD development. Consequently, patients undergoing coronary artery stenting remain susceptible to recurrent ischemic events, disease progression, and cardiovascular morbidity and mortality in the absence of comprehensive secondary prevention strategies. Considering the wide range of population undergoing stenting, it is warranted to optimize patients’ prognosis by incorporating cardiac rehabilitation. Preventive measures in the cardiac rehabilitation encompass lifestyle modifications such as smoking cessation, adoption of a heart-healthy diet, and weight management . Additionally, adherence to prescribed medications, including antiplatelet agents, statins, beta-blockers, and angiotensin-converting enzyme inhibitors, can reduce cardiovascular risk factors . Involvement in electrocardiography, echocardiography, and laboratory investigations can help identify cardiac functions . Moreover, supervised exercise training and psychosocial interventions are integrated into cardiac rehabilitation to improve cardiovascular fitness and address emotional barriers , . For the better management of prognosis, understanding the knowledge, attitudes, and practices (KAP) of patients toward cardiac rehabilitation is therefore needed. The KAP study seeks to assess understanding, attitudes, and behaviors pertaining to a specific health issue . It commonly utilizes both quantitative and qualitative methods, including surveys, interviews, or observations. Previous KAP research has primarily focused on cardiac rehabilitation among healthcare providers, with limited exploration among patients with cardiac conditions in China. One study in China identified suboptimal mental well-being and inadequate adherence to medical recommendations among myocardial infarction patients . However, KAP studies of patients undergoing coronary artery stenting and their engagement with cardiac rehabilitation remain scarce. The study aimed to bridge such gaps by exploring the KAP towards cardiac rehabilitation among patients after coronary artery stenting. Besides, the factors influencing the KAP were determined. The hypotheses were posited as follows: patient’s knowledge positively influences their attitudes toward cardiac rehabilitation, which further enhances their engagement in rehabilitation practices. The findings can contribute to the advancement of evidence-based practice and the promotion of optimal cardiovascular health outcomes among patients after coronary artery stenting. Study design and participants This cross-sectional study was conducted from December 1, 2023, to January 31, 2024, at the Fourth People’s Hospital of Jinan City, Dongying People’s Hospital, and the Second Hospital of Shandong University. The study participants were patients who had undergone coronary artery stent implantation. This study has obtained ethical approval from the Ethics Committee of the Fourth People’s Hospital of Jinan City and informed consent from the research participants. Inclusion Criteria: (1) Diagnosed with coronary heart disease and undergoing coronary artery stent implantation treatment; (2) Signed informed consent. Exclusion Criteria: (1) Patients with cognitive impairment or inability to perform activities of daily living; (2) Patients with severe postoperative complications; (3) Patients undergoing additional surgeries. This cross-sectional study was conducted from December 1, 2023, to January 31, 2024, at the Fourth People’s Hospital of Jinan City, Dongying People’s Hospital, and the Second Hospital of Shandong University. The study participants were patients who had undergone coronary artery stent implantation. This study has obtained ethical approval from the Ethics Committee of the Fourth People’s Hospital of Jinan City and informed consent from the research participants. Inclusion Criteria: (1) Diagnosed with coronary heart disease and undergoing coronary artery stent implantation treatment; (2) Signed informed consent. Exclusion Criteria: (1) Patients with cognitive impairment or inability to perform activities of daily living; (2) Patients with severe postoperative complications; (3) Patients undergoing additional surgeries. The design of the questionnaire was based on the Expert Consensus on Cardiac Rehabilitation After Coronary Artery Bypass Grafting , Best evidence summary of cardiac rehabilitation health education for patients after PCl and relevant literature , . After the initial design, modifications were made based on feedback from six experts, including four experts in the field of cardiac rehabilitation and two experts in general medicine. Regarding the total sample size of 452, the reliability of the questionnaire items for the knowledge and attitude sections was assessed, yielding a Cronbach’s α coefficient value of 0.908, indicating good internal consistency. For the subset of 262 individuals who had already initiated cardiac rehabilitation, the Cronbach’s α coefficient values for the knowledge, attitude, and practice sections were 0.854, 0.761 and 0.851, respectively. The overall Cronbach’s α coefficient value of the questionnaire was 0.898, suggesting good internal consistency. The final questionnaire was in Chinese and consisted of five sections: demographic information (age, gender, residence, education level, monthly family income, medical insurance status, occupation relevance to medicine, family history of cardiovascular disease, overweight or obesity, smoking status after coronary artery stent implantation, alcohol consumption after coronary artery stent implantation, type of coronary heart disease, other underlying diseases, health education on cardiac rehabilitation), knowledge dimension, attitude dimension, psychological health dimension, and practice dimension. The knowledge dimension comprised two aspects with a total of 10 questions, with 2 points awarded for “Very Familiar”, 1 point for “Heard of”, and 0 points for “Not sure”, with a score range of 0–20 points. The attitude dimension included 9 questions using a five-point Likert scale, ranging from “Strongly Agree” (5 points) to “Strongly Disagree” (1 point), with the first question being reverse-coded, yielding a score range of 9–45 points. The psychological health dimension comprised 4 questions and underwent descriptive analysis. The practice dimension included 9 questions, with items 2–8 ranging from “Always " (5 points) to “Never” (1 point), and items 1 and 9 undergoing descriptive analysis, with a score range of 7–35 points. A total score exceeding 80% in each dimension was considered indicative of adequate knowledge, positive attitude, and proactive practice . Questionnaires were distributed to the subjects by questionnaire survey of outpatient and inpatient patients, telephone follow-up of previous patients, and electronic questionnaire collection through WeChat communication group of patients. The online questionnaire was distributed via Questionnaire Star ( https://www.wjx.cn ) to healthcare professionals. Participants could scan the QR code using WeChat or follow the provided link to access and complete the questionnaire. Upon scanning the QR code, participants are first presented with the informed consent form. Only those who select “Agree” are granted access to the subsequent survey pages. To maintain data quality and ensure comprehensive responses, a one-submission-per-IP address restriction was enforced, and all questionnaire items were mandatory. Participants were assured of anonymity during the survey process. The research team, comprising three doctors trained as research assistants responsible for questionnaire promotion and distribution, meticulously reviewed all submissions for completeness, internal consistency, and logical coherence. Investigators were trained to grasp the problem’s meaning and the investigation process, enhancing data accuracy and consistency. Questionnaires containing logical errors, incomplete answers, or uniform responses across all items were categorized as invalid. The sample size was computed according to the published literature as follows : \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:n={\left(\frac{{Z}_{1-\frac{\alpha\:}{2}}}{\delta\:}\right)}^{2}\times\:p\times\:(1-p)$$\end{document} where n denoted the sample size, and p was assumed to be 0.5 to ensure the maximum sample size. α, also known as the type I error, was set to 0.05. In this case, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{Z}_{1-\frac{\alpha\:}{2}}=1.96.$$\end{document} Assuming a questionnaire recovery rate of 80%, the final target is to collect at least 480 completed questionnaires. Descriptive analysis was conducted on the demographic data and the KAP scores. Continuous variables were presented as Mean ± SD, while categorical indicators and responses to each question were expressed as frequencies (percentages). The Kolmogorov-Smirnov test was used to assess the distribution of continuous data. Due to their skewed distribution, Mann-Whitney U-test (for two groups) or the Kruskal-Wallis H test (for more than two groups) were applied for comparison analysis. Spearman correlation analysis was employed to examine the relationships between KAP scores. To identify the influential factors of KAP scores, univariate and multivariate logistic regression were used. Classification of KAP scores was based on the top 80% threshold in each dimension, with the group of lower score as reference. Variables included in the multivariate logistic regression were determined by variables with P < 0.05 in the univariate logistic regression. A P value < 0.05 was considered statistically significant for all statistical results. Statistical analysis was conducted using SPSS 26.0 software (IBM Corp., Armonk, NY, USA). A total of 527 responses were collected at the initial stage. Exclusions were made based on the following criteria: (1) 3 responses disagreed with participation; (2) 54 responses were completed in less than 60 s; (3) 1 respondent was under 18 years old; (4) 10 responses failed the trap questions; (5) 6 participants had started cardiac rehabilitation but did not complete the final section of the questionnaire. After exclusions, 452 valid responses remained, of which 262 participants had initiated cardiac rehabilitation. The participants had an average age of 56.76 ± 12.97 years, with a majority being male (57.52%), urban residents (72.35%), and with a family history of cardiovascular disease (59.96%). More than half of participants abstained from smoking (59.51%) and drinking (56.86%) after post-coronary artery stent implantation. Angina was the most common form of coronary heart disease (41.81%), followed by myocardial infarction (30.97%) and other conditions (27.21%) (Table ). The mean knowledge and attitude scores of all respondents ( N = 452) were 14.63 ± 4.70 (possible range: 0–20) and 37.50 ± 4.15 (possible range: 9–45), respectively. Significant variations in knowledge were observed across demographic factors, such as residence ( P = 0.001), education ( P = 0.004), and receipt of health education on cardiac rehabilitation ( P < 0.001). Differences in attitude were also noted among groups, such as monthly family income ( P = 0.048), and commencement of cardiac rehabilitation postoperatively ( P < 0.001). Among participants enrolled in cardiac rehabilitation ( N = 262), the mean practice score was 30.96 ± 4.49 (possible range: 7–35). Variations in practice were observed across variables, such as smoking status ( P = 0.010) and drinking status after post-coronary artery stent implantation ( P < 0.001) (Table ). Accuracy rates in the knowledge dimension ranged from 38.94 to 68.81%. The highest percentage (68.81%) correctly identified the significant risks posed by smoking to heart health and emphasized the importance of actively quitting smoking and avoiding secondhand smoke exposure (K3). Conversely, the lowest percentage (38.94%) demonstrated clarity regarding the appropriate level of exercise post-coronary artery stent implantation (K2). Additionally, a limited percentage (46.68%) accurately acknowledged the necessity for psychological assessment and potential psychological intervention following coronary artery stent implantation (K6) (Table ). In the attitudes dimension, positivity rates varied from 30.53 to 95.35%. The majority (95.35%) expressed high confidence in maintaining normal social activities post-surgery (A9). Conversely, only 30.53% perceived coronary artery stent implantation as a minor surgery and expressed lack of concern regarding postoperative rehabilitation outcomes (A1) (Table ). As regards psychological health dimension, the vast majority (from 96.02 to 97.79%) never or sometimes reported concerns (Supplementary Fig. 1). Practice adherence rates ranged from 46.46 to 53.32%. The highest percentage (53.32%) involved avoiding excessive fatigue and mental stress post-surgery while maintaining a positive mood (P6). A limited proportion (46.46%) actively advocated postoperative cardiac rehabilitation for individuals who had undergone coronary artery stent implantation or other cardiac surgeries (P8). Additionally, reasons for not undergoing postoperative rehabilitation included being busy with work and having insufficient time (14.16%), distance-related challenges or transportation inconvenience, and lack of family support (13.94%), as well as insufficient information provided by healthcare professionals (11.06%) (Table ). Spearman correlation analysis demonstrated positive correlations between knowledge and attitude among both the entire population ( r = 0.532, P < 0.001) and participants involved in cardiac rehabilitation ( r = 0.474, P < 0.001). Among participants engaged in cardiac rehabilitation, positive correlations were observed between knowledge and practice ( r = 0.446, P < 0.001), and between attitude and practice ( r = 0.378, P < 0.001) (Table ). In both univariate and multivariate analyses, the threshold for KAP scores was set at 80%. For knowledge, the threshold was 16 points, and 192 individuals scored above and 260 below this mark. Attitude scores had a threshold of 36 points, with 218 individuals scoring above and 234 below. Practice scores were only considered for subgroups with a threshold of 28 points. This resulted in 193 individuals scoring above and 69 below the threshold. In the multivariate logistic regression analysis, urban residence (OR = 2.136, 95% CI: 1.165–3.915; P = 0.014), diagnosis of angina (OR = 2.355, 95% CI: 1.349–4.112; P = 0.003), and myocardial infarction (OR = 1.881, 95% CI: 1.012–3.494; P = 0.046), as well as initiating cardiac rehabilitation postoperatively (OR = 3.192, 95% CI: 1.925–5.294; P < 0.001), were positively associated with knowledge. Conversely, having medical insurance (OR = 0.193, 95% CI: 0.091–0.405; P < 0.001), continued drinking after coronary artery stent implantation (OR = 0.325, 95% CI: 0.138–0.765; P = 0.010), and the presence of other underlying diseases (OR = 0.592, 95% CI: 0.380–0.923; P = 0.021) exhibited negative associations with knowledge. Knowledge (OR = 1.265, 95% CI: 1.194–1.340; P < 0.001), diagnosis of myocardial infarction (OR = 2.059, 95% CI: 1.092–3.882; P = 0.026), and receipt of health education on cardiac rehabilitation (OR = 1.830, 95% CI: 1.110–3.017; P = 0.018) were positively associated with attitude (Table S4). Moreover, knowledge (OR = 1.172, 95% CI: 1.072–1.282; P = 0.001), attitude (OR = 1.150, 95% CI: 1.050–1.259; P = 0.003), and having medical insurance (OR = 4.567, 95% CI: 1.182–17.648; P = 0.028) were positively associated with practice among patients undergoing cardiac rehabilitation (Table S5). Patients after coronary artery stenting had moderate knowledge, positive attitude and proactive practice towards cardiac rehabilitation. Besides, positive relationships between KAP scores were robustly identified. Several influential factors of KAP were further determined, such as residence, medical insurance, type of coronary heart disease and health education on cardiac rehabilitation. The moderate to high KAP scores in our study are consistent with the growing recognition of the importance of cardiac rehabilitation in improving post-procedure quality of life. Similarly, the post-CABG patients from India had average knowledge towards cardiac rehabilitation, and the majority held positive attitude and good practices . However, another study from China revealed suboptimal attitude and inadequate adherence to medical recommendations among myocardial infarction patients . The above discrepancy may be attributed to differences in study populations, questionnaire items, and healthcare settings. Healthcare providers should consider integrating mental health support and counseling services into cardiac rehabilitation programs to address these issues effectively. In the knowledge dimension, 68.81% correctly recognized the significant risks associated with smoking to heart health. Smoking cessation has been established to associated with reduced cardiovascular events and its related mortality . Conversely, 38.94% lacked clarity regarding the appropriate level of exercise post-procedure. This finding was concerning as exercise plays a crucial role in cardiac rehabilitation and overall cardiovascular health . Lack of understanding regarding appropriate exercise levels may lead to either insufficient physical activity, which could compromise rehabilitation outcomes, or excessive exertion, potentially leading to adverse events. Besides, the limited proportion (46.68%) acknowledged the necessity for psychological assessment and potential psychological intervention following the surgery. Reportedly, depression was associated with increased risk of adverse cardiovascular events among patients with coronary artery disease . Our finding suggests a need for improved education regarding the importance of addressing psychological stress of cardiac rehabilitation. Healthcare providers should integrate psychological assessment and support into cardiac rehabilitation to ensure care for patients after coronary artery stenting. The majority (95.35%) expressed high confidence in maintaining normal social activities post-surgery. Engaging in social activities can provide emotional support, reduce feelings of isolation, and promote a sense of belongingness . Therefore, the high level of confidence in our study regarding their ability to resume social activities is indicative of positive psychosocial adaptation and resilience post-surgery. However, only 30.53% perceived the procedure as a minor surgery and expressed a lack of concern regarding postoperative rehabilitation outcomes. In other words, patients recognized the invasive nature of the intervention, and thus showed concerns about the potential risks. Simeone, Vellone reported the post-surgery pain and lifestyle changes among patients undergoing emergency PCI, which contributed to anxiety and uncertainty. Healthcare providers should engage in open communication with patients, addressing their concerns, and providing realistic expectations regarding postoperative rehabilitation. After surgery, 53.32% engaged in practices aimed at avoiding excessive fatigue and mental stress post-surgery, while simultaneously maintaining a positive mood. Excessive fatigue can exacerbate postoperative symptoms, impair physical functioning, and delay recovery . Meanwhile, mental stress can contribute to psychological distress, which may hinder overall rehabilitation progress. The relatively low rate of advocacy (46.46%) for postoperative cardiac rehabilitation was also observed in our study. The positive associations of cardiac rehabilitation with reductions in cardiovascular mortality and hospital readmissions have been well established . Our result underscores the need for targeted interventions to enhance patient awareness and engagement in rehabilitation. Innovative strategies, including tele-rehabilitation programs, community-based initiatives, and peer support networks, may help increase patient engagement in rehabilitation activities. Positive correlations between KAP scores indicate that individuals with higher knowledge regarding cardiac rehabilitation are inclined to possess more positive attitudes and engage actively in rehabilitation practices. These findings underscore the educational interventions aimed at shaping patient attitudes and behaviors, thereby enhancing recovery and subsequent life quality. Furthermore, our study identified factors influencing KAP scores. First, urban residents have better access to healthcare resources and health promotion campaigns, which facilitate the dissemination of knowledge about post-stenting care . Second, individuals diagnosed with cardiac disorders have higher levels of knowledge and more positive attitudes towards their condition. One possible explanation can be the enhanced counseling with healthcare provides. Third, continued drinking after coronary artery stent acted as a barrier of knowledge. Alcohol consumption may be associated with reduced motivation to seek out health-promoting behaviors . Fourth, positive association between receipt of health education on cardiac rehabilitation and attitudes was observed. By providing accurate information and addressing concerns, education programs can help patients develop positive attitudes towards rehabilitation. Fifth, individuals with medical insurance may be more likely to participate in rehabilitation due to financial coverage for medical expenses. The study had several limitations. The study adopted a cross-sectional design, thereby limiting the determination of causality. The design of questionnaire can be impacted by regional practices and institutional policies, thereby constraining the broader extrapolation of findings. Furthermore, reliance on self-reported data introduced the possibility of social desirability bias, potentially leading to inflated scores . Patients following coronary artery stenting exhibited moderate knowledge, positive attitudes, and proactive practice in cardiac rehabilitation. Positive correlations among KAP scores were observed. Our findings suggest the necessity for targeted educational interventions in cardiac rehabilitation, especially for rural residents, individuals who did not initiate rehabilitation postoperatively, and those who continued drinking after implantation. Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 |
National-wide survey of ophthalmic human resources in China in 2021 | c530451d-d12b-46e5-8e32-7c31b9382bc3 | 11603949 | Ophthalmology[mh] | Ophthalmologists, ophthalmic nurses, and optometrists are the major forces involved in the provision of ophthalmic outpatient services. The World Health Organization (WHO), in its “Global Initiative for the Elimination of Avoidable Blindness: action plan 2006–2011” recommended that the ratio of ophthalmologists to the local population in Asia should reach 1:50,000 by 2021 . Data show that in 2015, China’s figure reached 1.32 . After a gap of 5 years, whether China still maintains this level and whether it can still reach the recommended level of WHO is of great interest to people. China is currently facing the challenge of a large aging population. According to the 7th census in 2021, 18.7% of the population is aged 60 years or older, and 13.5% is aged 65 years or older . This aging demographic is increasing the burden of non-communicable diseases , including cataracts , age-related macular degeneration, glaucoma , and diabetic retinopathy . It is estimated that by 2050, the number of age-related cataract patients in China will reach 187.26 million, significantly challenging the capacity of ophthalmic outpatient services. To answer the above questions, we conducted a comprehensive nationwide survey in 2021, encompassing all secondary and tertiary hospitals in the 31 provinces of mainland China, to gather vital information on ophthalmic human resources. Study design This was a nationwide survey with 2996 registered tertiary and 10,404 registered secondary hospitals in mainland China at the end of 2020. All hospitals were sent questionnaires.This study does not involve specific patients or participants, but is a survey of ophthalmic human resources data. Data collection Data were collected using a self-designed web-based questionnaire ( https://www.wjx.cn/ ). Each participating hospital arranged for a doctor/nurse to complete the questionnaire. To minimize the time burden on medical staff, the questionnaire contains only 10 questions, the key information included: the number of ophthalmologists, the number of ophthalmic nurses, the number of optometrists, and the number of ophthalmic outpatient visits per year. Data cleaning (quality control) Before statistical analysis, we performed a data cleaning procedure, and any outliers were returned to the corresponding hospitals for verification. Outliers were identified in two steps: firstly, using the outlier test function from the “car” package of the open source R Program to locate the outliers and corresponding hospitals. For each outlier, the corresponding hospital was contacted for verification. For example, the number of ophthalmologists in secondary hospitals being higher than in a tertiary hospital indicates a mistake, as the tertiary hospitals in China usually represent the highest level of medical care and have the most medical resources. Estimation of total ophthalmic medical resources The vast majority of ophthalmic medical resources in China are in tertiary and secondary hospitals. Because some hospitals did not respond to the survey, we used the following formula to estimate the total number of ophthalmologists, ophthalmic nurses, and optometrists in mainland China: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm N=\sum\nolimits_1^{31}(\mathrm N1_{\mathrm i}/\mathrm R1_{\mathrm i}+\mathrm N2_{\mathrm i}/\mathrm R2_{\mathrm i})$$\end{document} N = ∑ 1 31 ( N 1 i / R 1 i + N 2 i / R 2 i ) where N represents the total number of ophthalmologists (ophthalmic nurses or optometrists). i represents the province. N1 represents the total number of ophthalmologists in responding secondary hospitals. R1 represents the response rate of secondary hospitals. N2 represents the total number of ophthalmologists in responding tertiary hospitals. R2 represents the response rate of tertiary hospitals. Definition of ophthalmologist density The number of ophthalmologists per 50,000 people (similarly for ophthalmic nurses and optometrists). The 2021 population data of the 31 provinces came from the National Bureau of Statistics ( https://data.stats.gov.cn ) . Doctor-patient ratio This ratio reflects the pressure of ophthalmic outpatient services in each province, defined as the total number of ophthalmic outpatient services divided by the number of ophthalmologists. Statistical analysis The open-source R program ( https://www.r-project.org/ , version 4.3.3) was used for data analysis. Frequencies and percentages were used to describe categorical variables. Lorenz curves and Gini coefficients were used to estimate the equity of each indicator among the 31 provinces of mainland China. The Gini coefficient, ranging from 0 to 1, was calculated from the Lorenz curve; a coefficient < 0.2 indicates that human resources are well allocated among the 31 provinces while a Gini coefficient > 0.6 indicates significant disparity . This was a nationwide survey with 2996 registered tertiary and 10,404 registered secondary hospitals in mainland China at the end of 2020. All hospitals were sent questionnaires.This study does not involve specific patients or participants, but is a survey of ophthalmic human resources data. Data were collected using a self-designed web-based questionnaire ( https://www.wjx.cn/ ). Each participating hospital arranged for a doctor/nurse to complete the questionnaire. To minimize the time burden on medical staff, the questionnaire contains only 10 questions, the key information included: the number of ophthalmologists, the number of ophthalmic nurses, the number of optometrists, and the number of ophthalmic outpatient visits per year. Before statistical analysis, we performed a data cleaning procedure, and any outliers were returned to the corresponding hospitals for verification. Outliers were identified in two steps: firstly, using the outlier test function from the “car” package of the open source R Program to locate the outliers and corresponding hospitals. For each outlier, the corresponding hospital was contacted for verification. For example, the number of ophthalmologists in secondary hospitals being higher than in a tertiary hospital indicates a mistake, as the tertiary hospitals in China usually represent the highest level of medical care and have the most medical resources. The vast majority of ophthalmic medical resources in China are in tertiary and secondary hospitals. Because some hospitals did not respond to the survey, we used the following formula to estimate the total number of ophthalmologists, ophthalmic nurses, and optometrists in mainland China: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm N=\sum\nolimits_1^{31}(\mathrm N1_{\mathrm i}/\mathrm R1_{\mathrm i}+\mathrm N2_{\mathrm i}/\mathrm R2_{\mathrm i})$$\end{document} N = ∑ 1 31 ( N 1 i / R 1 i + N 2 i / R 2 i ) where N represents the total number of ophthalmologists (ophthalmic nurses or optometrists). i represents the province. N1 represents the total number of ophthalmologists in responding secondary hospitals. R1 represents the response rate of secondary hospitals. N2 represents the total number of ophthalmologists in responding tertiary hospitals. R2 represents the response rate of tertiary hospitals. The number of ophthalmologists per 50,000 people (similarly for ophthalmic nurses and optometrists). The 2021 population data of the 31 provinces came from the National Bureau of Statistics ( https://data.stats.gov.cn ) . This ratio reflects the pressure of ophthalmic outpatient services in each province, defined as the total number of ophthalmic outpatient services divided by the number of ophthalmologists. The open-source R program ( https://www.r-project.org/ , version 4.3.3) was used for data analysis. Frequencies and percentages were used to describe categorical variables. Lorenz curves and Gini coefficients were used to estimate the equity of each indicator among the 31 provinces of mainland China. The Gini coefficient, ranging from 0 to 1, was calculated from the Lorenz curve; a coefficient < 0.2 indicates that human resources are well allocated among the 31 provinces while a Gini coefficient > 0.6 indicates significant disparity . Characteristics of investigated hospitals The names of the 31 provinces in mainland China are shown in Fig. . These provinces were classified into four regions: east, central, west, and northeast. The eastern region includes ten provinces: Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan. The central region includes six provinces: Shānxi, Anhui, Jiangxi, Henan, Hubei, and Hunan. The western region contains 12 provinces: Neimeng, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Xizang, Shǎnxi, Gansu, Qinghai, Ningxia, and Xinjiang. Finally, there are three provinces in the northeastern region: Liaoning, Jilin, and Heilongjiang. In this survey, 9856 hospitals responded, yielding an overall response rate of 73.6% (9856/13400). The response rates for secondary and tertiary hospitals were 77.9% (8100/10404) and 58.6% (1756/2996), respectively. Among the responding hospitals, 78.7% (7759/9856) had ophthalmic human resources and were able to provide ophthalmic services. Ophthalmic human resources in mainland China in 2021 The total numbers of ophthalmologists, ophthalmic nurses, and optometrists in mainland China in 2021 were 48,652, 64,495, and 14,320, respectively. On average, for every 50,000 people in mainland China, there were 1.70 ophthalmologists, 2.25 ophthalmic nurses, and 0.47 optometrists. The ophthalmologist density for each province is shown in Fig. A. The density of ophthalmologists was < 1 only in the Xizang Autonomous Region. Ophthalmic nurse density ranged from 0.38 (Tibet Autonomous Region) to 4.32 (Ningxia province) (Fig. B). The density of optometrists ranged from 0.19 (Xizang Autonomous Region) to 1.04 (Tianjin) (Fig. C). Doctor-patient ratio The total number of ophthalmic outpatient visits in mainland China was 125,760,140 person-times in year 2021, and the average doctor-patient ratio was 2584.8. 12 provinces had a doctor-patient ratio higher than the average (Fig. ). Among the 12 provinces, 8 were in the eastern region, accounting for 66.7% (8/12). Zhejiang Province, Shanghai Municipality, and Guangdong Province were the top three in terms of this ratio. Meanwhile, Xizang region had the lowest doctor-patient ratio, which was 1074.67. Equity of ophthalmic human resources across the whole mainland China The Lorenz curves showed the equity of distribution of ophthalmologists, ophthalmic nurses, optometrists among 31 provinces of China mainland. The best equity across provinces is the nurse distribution, then the ophthalmologists, the equity of the optometrist was the worst (Fig. ). The Gini coefficients were 0.411, 0.399, and 0.423, respectively. The names of the 31 provinces in mainland China are shown in Fig. . These provinces were classified into four regions: east, central, west, and northeast. The eastern region includes ten provinces: Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan. The central region includes six provinces: Shānxi, Anhui, Jiangxi, Henan, Hubei, and Hunan. The western region contains 12 provinces: Neimeng, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Xizang, Shǎnxi, Gansu, Qinghai, Ningxia, and Xinjiang. Finally, there are three provinces in the northeastern region: Liaoning, Jilin, and Heilongjiang. In this survey, 9856 hospitals responded, yielding an overall response rate of 73.6% (9856/13400). The response rates for secondary and tertiary hospitals were 77.9% (8100/10404) and 58.6% (1756/2996), respectively. Among the responding hospitals, 78.7% (7759/9856) had ophthalmic human resources and were able to provide ophthalmic services. The total numbers of ophthalmologists, ophthalmic nurses, and optometrists in mainland China in 2021 were 48,652, 64,495, and 14,320, respectively. On average, for every 50,000 people in mainland China, there were 1.70 ophthalmologists, 2.25 ophthalmic nurses, and 0.47 optometrists. The ophthalmologist density for each province is shown in Fig. A. The density of ophthalmologists was < 1 only in the Xizang Autonomous Region. Ophthalmic nurse density ranged from 0.38 (Tibet Autonomous Region) to 4.32 (Ningxia province) (Fig. B). The density of optometrists ranged from 0.19 (Xizang Autonomous Region) to 1.04 (Tianjin) (Fig. C). The total number of ophthalmic outpatient visits in mainland China was 125,760,140 person-times in year 2021, and the average doctor-patient ratio was 2584.8. 12 provinces had a doctor-patient ratio higher than the average (Fig. ). Among the 12 provinces, 8 were in the eastern region, accounting for 66.7% (8/12). Zhejiang Province, Shanghai Municipality, and Guangdong Province were the top three in terms of this ratio. Meanwhile, Xizang region had the lowest doctor-patient ratio, which was 1074.67. The Lorenz curves showed the equity of distribution of ophthalmologists, ophthalmic nurses, optometrists among 31 provinces of China mainland. The best equity across provinces is the nurse distribution, then the ophthalmologists, the equity of the optometrist was the worst (Fig. ). The Gini coefficients were 0.411, 0.399, and 0.423, respectively. China, a rapidly developing low- and middle-income country, accounts for approximately 18.5% of the world’s population and about 17% of the world’s gross domestic product . For a country with a large population and a large economy, the situation of human resources in ophthalmology is worthy of attention, but in this regard we are very lack of data, so we conducted this survey. In 2015, the global ophthalmologist density was 1.58 per 50,000 people while in China it was 1.32. Around the same period in 2014, India, with a population size close to China’s, had an ophthalmologist density of 0.672 . By 2021, according to our survey, the ophthalmologist density in mainland China reached 1.70 per 50,000 people. It is estimated that the number of global ophthalmologists has been increasing at a rate of approximately 2.6% per year , suggesting that the global ophthalmologist density in 2021 should be approximately 1.79 per 50,000 people. Although the ophthalmologist density in China is still below the global average, the gap has been narrowed, and China has made remarkable improvements in the development of ophthalmologist resources in recent years. Since 2009, China has implemented reforms in the medical and health system to fulfill residents’ equal access to medical services , and these reforms have been somewhat successful, according to our survey. However, the Xizang Autonomous Region remains the weak link in China; the density of ophthalmologists in all other provinces in mainland China has reached the WHO’s recommended level, except for Xizang Autonomous Region. Indeed, the development of ophthalmic nurse and optometrist resources in the Xizang Autonomous Region lags significantly behind other provinces. We must acknowledge that the ophthalmic service pressure was significantly higher in the eastern coastal provinces of China, including Zhejiang province, Shanghai, Guangdong province, Tianjin, Jiangsu province, Beijing and Fujian province. Especially in Shanghai and Zhejiang province, the doctor-patient ratio exceeded 4000 in the year 2021. Meanwhile for Shānxi province, Hunan province, Jilin province, Hubei province, Heilongjiang province, Guizhou province and Xizang Autonomous Region, the doctor-patient ratio was below 2000. Thus, the Chinese government needs to consider optimizing medical resources allocation and strengthening the capacity building of ophthalmic services in underdeveloped provinces to enable residents to access treatment in local areas. Overall, mainland China has made remarkable improvements in the construction of ophthalmic human resources compared with that in year 2015 , the gap in ophthalmologist density between China and the global average has narrowed. However, several challenges remain. There is an imbalance in ophthalmic service pressure, with the Shanghai and Zhejiang province facing significantly greater pressure than other provinces. Additionally, there is also an imbalance in the development of ophthalmic human resources, with the Xizang Autonomous Region lagging behind other provinces in terms of ophthalmologists, nurses, and optometrists. The limitations of our study include a relatively low response rate. Besides, as this is a questionnaire study, there might be bias like non-response bias and selection bias. Additionally, due to limited research on ophthalmic outpatient service evaluation, it is challenging to make cross-sectional comparisons between China and other countries, or to conduct longitudinal comparisons within China itself. Such comparisons would help us better understand the development of China’s ophthalmic human resources construction. Supplementary Material 1. |
Assessment of a Light-Curable Hydrogel to Be Used for Root Canal Obturation | 0d6dbee3-80d5-430b-ad23-97971c208255 | 11843793 | Dentistry[mh] | Solid cones composed of gutta-percha and root canal sealers are used to fill the root canal space with the aim of preventing reinfection. The root canal space has a complex morphology ; therefore, preparation with standardized instruments leaves most of the root canal space unprepared . Root canal irrigation and the use of sodium hypochlorite aids the elimination of bacterial biofilms and pulpal remnants, whereas calcium chelators remove the smear layer created by mechanical debridement . Following chemomechanical preparation, the root canal space has an irregular shape and is encased with modified dentine. The micro structure of the dentine depends on the irrigation regime used , with different adjuncts resulting in deeper penetration than others . The use of standardized solid cones has evolved from the standardized preparation of the root canal space with the sealer aiming at filling the remaining intricacies . The use of gutta-percha and sealer creates an interface between the gutta-percha and sealer and between the sealer and the root dentine. Methods of filling the root canal without the standardized gutta-percha cones will enable more conservative root canal preparation techniques whereby the root canal is not prepared to fit a standardized cone but can retain its own natural anatomy ( ; Lussi et al. 2002). The use of adequate disinfection remains of utmost importance, but the reduced use of instrumentation will also reduce the smear layer created, leading to a reduction in use of chelators as well. The aim of the current study was the assessment of the physical properties and antimicrobial characteristics of OdneFill (Odne), a photocurable filled hydrogel used to obturate root canals. The proposed method is an easy obturation technique composed of a single phase in which, after irrigation, the material will be injected into the root canal and light cured with an appropriate light source. The investigated materials and their composition are shown in . The OdneFill material was dispensed through a syringe and light cured using Odne AG curing light for 120 s as instructed by the manufacturer. AH Plus and BioRoot were prepared according to the manufacturers’ instructions and allowed to set at 37 °C and 100% relative humidity for 48 and 24 h, respectively. Characterization of Materials The microstructure and elemental analysis were assessed by scanning electron microscopy (SEM; EVO MA10, Carl Zeiss Ltd.) and energy-dispersive spectroscopy (EDS; INCA, Zeiss Oxford Labs) performed on material specimens (10-mm diameter, 2-mm thickness) that were allowed to set at 37 °C and 100% relative humidity. The specimens were then embedded in resin followed by grinding and polishing. Characterization of the Light Source and Assessment of Degree of Conversion of Hydrogel The methodology is given in the Appendix . Effect of Irrigating Solution on the Material Chemistry and Microstructure The effect of 0.2% chlorhexidine, 5.25% sodium hypochlorite solutions compared with water on OdneFill, BioRoot, and AH Plus was assessed using a split-tooth model using human single rooted teeth described in previous research . The decoronated teeth with 1 root canal (ethical approval REC Ref 14/EM/1128), free of caries, were split and reassembled then root treated. The materials were dispensed into the root canal, and the AH Plus and BioRoot were allowed to set while the OdneFill was light cured. After 1 wk, each root was unwrapped, and the root fragments were gently detached with the use of a scalpel to expose the sealers, which were retrieved intact from the dentin walls. The material that was in contact with the irrigated dentine was characterized by SEM and EDS. Determination of Physical Properties of the Materials The radiopacity, flow, film thickness, and solubility were measured according to ISO 6876; . The material hydrophilicity was assessed by the advancing contact angle method whereby a 5-µL drop of water was dispensed on the sample surface using a micro syringe, and for each specimen, 3 angle measurements were taken every 10 s. The drop volume was then slowly increased, and steps were repeated until constant contact angle measurements were achieved. The angles were measured by a specific image analysis software (OPTIMAS 6) interfaced with the camera. The microstructure and elemental analysis were assessed by scanning electron microscopy (SEM; EVO MA10, Carl Zeiss Ltd.) and energy-dispersive spectroscopy (EDS; INCA, Zeiss Oxford Labs) performed on material specimens (10-mm diameter, 2-mm thickness) that were allowed to set at 37 °C and 100% relative humidity. The specimens were then embedded in resin followed by grinding and polishing. The methodology is given in the Appendix . The effect of 0.2% chlorhexidine, 5.25% sodium hypochlorite solutions compared with water on OdneFill, BioRoot, and AH Plus was assessed using a split-tooth model using human single rooted teeth described in previous research . The decoronated teeth with 1 root canal (ethical approval REC Ref 14/EM/1128), free of caries, were split and reassembled then root treated. The materials were dispensed into the root canal, and the AH Plus and BioRoot were allowed to set while the OdneFill was light cured. After 1 wk, each root was unwrapped, and the root fragments were gently detached with the use of a scalpel to expose the sealers, which were retrieved intact from the dentin walls. The material that was in contact with the irrigated dentine was characterized by SEM and EDS. The radiopacity, flow, film thickness, and solubility were measured according to ISO 6876; . The material hydrophilicity was assessed by the advancing contact angle method whereby a 5-µL drop of water was dispensed on the sample surface using a micro syringe, and for each specimen, 3 angle measurements were taken every 10 s. The drop volume was then slowly increased, and steps were repeated until constant contact angle measurements were achieved. The angles were measured by a specific image analysis software (OPTIMAS 6) interfaced with the camera. For the microbiology evaluations, phosphate-buffered saline (PBS; Sigma Aldrich), lipase PS enzyme (Sigma Aldrich), and cholesterol esterase enzyme (EMD Millipore Corp) were used. The enzyme powders were dissolved in 0.1 M PBS containing 0.1% sodium azide (Sigma Aldrich) to inhibit bacterial growth. Degradation of Endodontic Materials by Bacterial Enzymes Weight loss analysis after contact with 3 mL of enzyme solutions, 0.15 wt/v% lipase (45 U/mL) and cholesterol esterase (40 U/mL) with pH 7 was undertaken in line with Tay et al. (2005). PBS was used as a control solution and discs not immersed in solution as a general control. The weight loss was measured to the accuracy of 0.001 g. The weight of the PBS was subtracted to make up for the liquid uptake of materials in solution. Topography was analyzed by SEM in secondary electron mode. Assessment of Material Changes in Contact with Bacterial Enzymes and Biofilm Material discs were tested with bacterial species Streptococcus mutans (ATCC 3209), Enterococcus faecalis (ATCC 29212), Fusobacterium nucleatum (FNN-25; ATCC 25586), and Veillonella dispar (NCTC 11831) grown at 37 °C in an anaerobic chamber (Whitley DG250 workstation, Don Whitley Scientific Limited) for 3 d to form biofilms. As negative controls, discs (of OdneFill, AH Plus, and BioRoot) without bacteria and an empty coverslip were used. Positive controls were a coverslip with bacteria and, dependent on the experiment, discs of AH Plus and BioRoot with bacteria. For SEM analysis, a disc with an enzyme solution, cholesterol esterase in PBS, was used as an additional control. The test samples were placed in a 24-well culture plate for the biofilm assay. Overnight cultures of the bacteria were diluted to 10 3 bacteria/mL in artificial saliva, which was then used to inoculate the discs initially (500 µL of each bacterial o/n suspension). The biofilm was allowed to form over the samples for 3 d, during which the media (artificial saliva or enzyme solutions) were changed every 24 h. After 3 d, the discs were carefully lifted into a fresh 24-well plate and stained with live/dead staining (Baclight, Invitrogen; ). Biofilms were visualized and quantified using a laser scanning confocal microscope (Leica sp8 Microsystems GmbH, Mannheim, Germany). Images were taken for biofilm viability analysis using our published method . Furthermore, a separate set of discs, incubated in the same way, were used to determine viable counts (as described below for the direct contact test). A final set of discs was prepared for and observed using SEM to analyze changes to the material after biofilm growth ( n = 3). Direct Contact Test Overnight bacterial cultures of S. mutans and E. faecalis in brain heart infusion (BHI) were diluted until 0.1 optical density using a Jen 6400 spectrometer. Twenty microliters of this suspension (containing 0.1 × 10 8 bacteria/mL) was then pipetted onto the disc surface. An autoclaved glass coverslip (2 cm × 2 cm) was carefully positioned on top to spread the inoculum over the entire disc surface. The Petri dishes were closed, wrapped in cling film, labeled, and incubated at 37 °C in a CO 2 5% incubator for S. mutans and in a standard 37 °C incubator for E. faecalis for 24 h. A dry disc without bacteria, 1 disc with cholesterol esterase change, and AH Plus/BioRoot discs were used as controls. Following the incubation period, the whole disc with the coverslip assembly was immersed in 10 mL of BHI in a Falcon tube and vortexed for 1 min at full power to dislodge the surface-adhered bacteria. This was followed by the Miles and Misra method , with serial dilutions of the bacterial suspension. Of each dilution, 3 × 20 µL was pipetted into a quadrant of a BHI agar plate, and the number of colony-forming units (CFUs) at each dilution was recorded after overnight incubation at 37 °C. The number of colonies per milliliter was then calculated. The assay was run in triplicate. Statistical Analysis The statistical analysis was performed using GraphPad Prism version 9.5.0. The data were tested to ensure they were a normally distributed analysis of variance with P = 0.05. One-way analysis of variance was then used to determine whether there were significant differences between data sets, and Tukey post hoc tests were used to determine the difference between the groups. Weight loss analysis after contact with 3 mL of enzyme solutions, 0.15 wt/v% lipase (45 U/mL) and cholesterol esterase (40 U/mL) with pH 7 was undertaken in line with Tay et al. (2005). PBS was used as a control solution and discs not immersed in solution as a general control. The weight loss was measured to the accuracy of 0.001 g. The weight of the PBS was subtracted to make up for the liquid uptake of materials in solution. Topography was analyzed by SEM in secondary electron mode. Material discs were tested with bacterial species Streptococcus mutans (ATCC 3209), Enterococcus faecalis (ATCC 29212), Fusobacterium nucleatum (FNN-25; ATCC 25586), and Veillonella dispar (NCTC 11831) grown at 37 °C in an anaerobic chamber (Whitley DG250 workstation, Don Whitley Scientific Limited) for 3 d to form biofilms. As negative controls, discs (of OdneFill, AH Plus, and BioRoot) without bacteria and an empty coverslip were used. Positive controls were a coverslip with bacteria and, dependent on the experiment, discs of AH Plus and BioRoot with bacteria. For SEM analysis, a disc with an enzyme solution, cholesterol esterase in PBS, was used as an additional control. The test samples were placed in a 24-well culture plate for the biofilm assay. Overnight cultures of the bacteria were diluted to 10 3 bacteria/mL in artificial saliva, which was then used to inoculate the discs initially (500 µL of each bacterial o/n suspension). The biofilm was allowed to form over the samples for 3 d, during which the media (artificial saliva or enzyme solutions) were changed every 24 h. After 3 d, the discs were carefully lifted into a fresh 24-well plate and stained with live/dead staining (Baclight, Invitrogen; ). Biofilms were visualized and quantified using a laser scanning confocal microscope (Leica sp8 Microsystems GmbH, Mannheim, Germany). Images were taken for biofilm viability analysis using our published method . Furthermore, a separate set of discs, incubated in the same way, were used to determine viable counts (as described below for the direct contact test). A final set of discs was prepared for and observed using SEM to analyze changes to the material after biofilm growth ( n = 3). Overnight bacterial cultures of S. mutans and E. faecalis in brain heart infusion (BHI) were diluted until 0.1 optical density using a Jen 6400 spectrometer. Twenty microliters of this suspension (containing 0.1 × 10 8 bacteria/mL) was then pipetted onto the disc surface. An autoclaved glass coverslip (2 cm × 2 cm) was carefully positioned on top to spread the inoculum over the entire disc surface. The Petri dishes were closed, wrapped in cling film, labeled, and incubated at 37 °C in a CO 2 5% incubator for S. mutans and in a standard 37 °C incubator for E. faecalis for 24 h. A dry disc without bacteria, 1 disc with cholesterol esterase change, and AH Plus/BioRoot discs were used as controls. Following the incubation period, the whole disc with the coverslip assembly was immersed in 10 mL of BHI in a Falcon tube and vortexed for 1 min at full power to dislodge the surface-adhered bacteria. This was followed by the Miles and Misra method , with serial dilutions of the bacterial suspension. Of each dilution, 3 × 20 µL was pipetted into a quadrant of a BHI agar plate, and the number of colony-forming units (CFUs) at each dilution was recorded after overnight incubation at 37 °C. The number of colonies per milliliter was then calculated. The assay was run in triplicate. The statistical analysis was performed using GraphPad Prism version 9.5.0. The data were tested to ensure they were a normally distributed analysis of variance with P = 0.05. One-way analysis of variance was then used to determine whether there were significant differences between data sets, and Tukey post hoc tests were used to determine the difference between the groups. Characterization of Materials The SEM images of OdneFill exhibited a smooth microstructure with deposits over the surface, which included radiopaque circular structures that were rich in zirconium ( , red arrows). The higher-magnification micrograph showed some porosity. The material elemental composition is shown in . The amount of zirconium was higher in OdneFill compared with AH Plus and BioRoot RCS . BioRoot RCS exhibited a microstructure composed of dark gray particles, some of which had a halo of lighter gray around them (cement particle in red circle and halo with red arrow). These particles were composed of calcium, silicon, and oxygen. Lighter and shinier particles (blue arrows) rich in zirconium were also present. The material also exhibited porosity and some cracks. The elemental analysis showed the presence of calcium, silicon, zirconium, and oxygen . AH Plus was composed of a rounder particle morphology of different sizes and opacity. The larger particles were rich in calcium and tungsten, while the smaller ones were composed of zirconium and oxygen . Characterization of the Light Source and Assessment of Degree of Conversion of Hydrogel The results for the work undertaken are given in the Appendix and shown in Appendix Figures 1 and 2 . Effect of Irrigating Solution on the Material Chemistry and Microstructure The material microstructure and elemental analysis of OdneFill and AH Plus did not change in contact with different irrigating solutions . When compared with the sealer not exposed to solution, in , the AH Plus exhibited filler plucking when in contact with all 3 solutions. The BioRoot RCS exhibited microstructural changes in contact with sodium hypochlorite and chlorhexidine. There was a reduction in the zirconium content in contact with sodium hypochlorite and a reduction in both calcium and zirconium in contact with chlorhexidine. Determination of Physical Properties The BioRoot RCS and AH Plus exceeded the minimum value specified in the ISO norm for flow, while the OdneFill flow could not be measured as it was wider than the size of the plates used . A second trial was undertaken with 80-mm × 80-mm plates, which is double the minimum recommended by the ISO 6876; , but the material was still out of range. Both OdneFill and AH Plus had a film thickness <50 µm, which complied with the ISO norm, while BioRoot RCS had a film thickness of 51.3 µm, which was slightly higher than the 50 µm specified in the standard. All materials had a radiopacity greater than 3 mm aluminum thickness. All materials had a contact angle less than 90° with AH Plus being the least wettable surface. Low-contact angles signify hydrophilicity. The BioRoot RCS was the most hydrophilic and AH Plus the most hydrophobic material. Degradation of Endodontic Materials by Bacterial Enzymes There was an initial loss in weight for OdneFill when incubated with lipase and a sustained weight loss for AH Plus in both enzyme solutions . OdneFill and AH Plus did not show any microstructural changes in contact with the enzyme solutions . OdneFill exhibited some surface changes in contact with PBS. The BioRoot control had an aggregation of circular crystals on its surface, which is in keeping with the surface carbonation reported previously over the surfaces of hydraulic cements. This is caused by a reaction of the calcium hydroxide produced as a by-product of hydration to the atmospheric carbon dioxide . Immersion in solution produced varying levels of surface carbonation with the least shown in PBS and the most extensive in the cholesterol esterase. Assessment of Material Changes in Contact with Bacterial Enzymes and Biofilms The control biofilm showed mostly green fluorescence , indicating that the bacteria were viable throughout the experimental period. Both OdneFill and BioRoot RCS exhibited reduced bacterial growth/biofilm formation, with BioRoot RCS showing higher reduction in bacterial load . AH Plus was similar to the control ( P > 0.05). The results of CFUs showed a similar trend . The scanning electron micrographs show changes to OdneFill and AH Plus surfaces in contact with cholesterol esterase and the biofilm with the changes more marked in the latter . The cholesterol esterase led to the formation of ridges, resulting in a rough surface of the OdneFill. This was also observed in AH Plus, where the radiopacifier particles were more evident after contact with cholesterol esterase, indicating some degradation. Biofilm was observed on both OdneFill and AH Plus covering the whole material surface. BioRoot RCS exhibited carrying degrees of carbonation on its surface for all specimens. Direct Contact Test shows the results of the direct contact test. BioRoot RCS did not exhibit any bacterial growth with all groups ( P < 0.001 compared with control). Both OdneFill and AH Plus reduced the microbial load for both species but to a lesser extent than BioRoot RCS did. AH Plus was least antimicrobial among all materials tested. The SEM images of OdneFill exhibited a smooth microstructure with deposits over the surface, which included radiopaque circular structures that were rich in zirconium ( , red arrows). The higher-magnification micrograph showed some porosity. The material elemental composition is shown in . The amount of zirconium was higher in OdneFill compared with AH Plus and BioRoot RCS . BioRoot RCS exhibited a microstructure composed of dark gray particles, some of which had a halo of lighter gray around them (cement particle in red circle and halo with red arrow). These particles were composed of calcium, silicon, and oxygen. Lighter and shinier particles (blue arrows) rich in zirconium were also present. The material also exhibited porosity and some cracks. The elemental analysis showed the presence of calcium, silicon, zirconium, and oxygen . AH Plus was composed of a rounder particle morphology of different sizes and opacity. The larger particles were rich in calcium and tungsten, while the smaller ones were composed of zirconium and oxygen . The results for the work undertaken are given in the Appendix and shown in Appendix Figures 1 and 2 . The material microstructure and elemental analysis of OdneFill and AH Plus did not change in contact with different irrigating solutions . When compared with the sealer not exposed to solution, in , the AH Plus exhibited filler plucking when in contact with all 3 solutions. The BioRoot RCS exhibited microstructural changes in contact with sodium hypochlorite and chlorhexidine. There was a reduction in the zirconium content in contact with sodium hypochlorite and a reduction in both calcium and zirconium in contact with chlorhexidine. The BioRoot RCS and AH Plus exceeded the minimum value specified in the ISO norm for flow, while the OdneFill flow could not be measured as it was wider than the size of the plates used . A second trial was undertaken with 80-mm × 80-mm plates, which is double the minimum recommended by the ISO 6876; , but the material was still out of range. Both OdneFill and AH Plus had a film thickness <50 µm, which complied with the ISO norm, while BioRoot RCS had a film thickness of 51.3 µm, which was slightly higher than the 50 µm specified in the standard. All materials had a radiopacity greater than 3 mm aluminum thickness. All materials had a contact angle less than 90° with AH Plus being the least wettable surface. Low-contact angles signify hydrophilicity. The BioRoot RCS was the most hydrophilic and AH Plus the most hydrophobic material. There was an initial loss in weight for OdneFill when incubated with lipase and a sustained weight loss for AH Plus in both enzyme solutions . OdneFill and AH Plus did not show any microstructural changes in contact with the enzyme solutions . OdneFill exhibited some surface changes in contact with PBS. The BioRoot control had an aggregation of circular crystals on its surface, which is in keeping with the surface carbonation reported previously over the surfaces of hydraulic cements. This is caused by a reaction of the calcium hydroxide produced as a by-product of hydration to the atmospheric carbon dioxide . Immersion in solution produced varying levels of surface carbonation with the least shown in PBS and the most extensive in the cholesterol esterase. The control biofilm showed mostly green fluorescence , indicating that the bacteria were viable throughout the experimental period. Both OdneFill and BioRoot RCS exhibited reduced bacterial growth/biofilm formation, with BioRoot RCS showing higher reduction in bacterial load . AH Plus was similar to the control ( P > 0.05). The results of CFUs showed a similar trend . The scanning electron micrographs show changes to OdneFill and AH Plus surfaces in contact with cholesterol esterase and the biofilm with the changes more marked in the latter . The cholesterol esterase led to the formation of ridges, resulting in a rough surface of the OdneFill. This was also observed in AH Plus, where the radiopacifier particles were more evident after contact with cholesterol esterase, indicating some degradation. Biofilm was observed on both OdneFill and AH Plus covering the whole material surface. BioRoot RCS exhibited carrying degrees of carbonation on its surface for all specimens. shows the results of the direct contact test. BioRoot RCS did not exhibit any bacterial growth with all groups ( P < 0.001 compared with control). Both OdneFill and AH Plus reduced the microbial load for both species but to a lesser extent than BioRoot RCS did. AH Plus was least antimicrobial among all materials tested. The current study investigates a novel light-curing injectable hydrogel proposed to be used as an obturating material to replace gutta-percha and sealer combinations. This technique allows for a more conservative root canal preparation. A noninstrumentation technique has been proposed in the past , exhibiting comparable obturations clinically to root canals filled with gutta-percha different sealers . Noninstrumentation may be a solution to avoid overcutting and overpreparation of the root canal, which together with irrigating solutions destroys the root dentine microstructure . However, obturation of the root canal will be challenging. OdneFill is an injectable hydrogel that will enable the obturation of all root canal anatomy regardless of the preparation method. The light source was optimized to have an irradiance matching the photoinitiator in OdneFill, thus enabling an adequate cure. The polymerization at different thicknesses was assessed to optimize the clinical method. Measurement of the double bond conversion of the monomer provided inconclusive results due to the presence of water in the hydrogel, impeding precise measurement of double bond absorptions. Nevertheless, the current results demonstrate the highest degree of conversion at 2- to 5-mm sample thickness. Further studies are needed to determine precisely the degree of conversion. Notably, there is currently no comparable obturation system in clinical use; thus, 2 root canal sealers have been used as comparators. OdneFill was characterized using various methods including SEM and EDS for microstructure and elemental evaluation. These techniques enable the visualization of the material at high magnifications and the assessment of interactions with dentine modified by irrigating solutions. Electron microscopy is also a useful tool to assess the presence of bacteria over a material surface. The main limitation in the latter is the high magnification, thus limiting the field of view and also the inability to detect whether the bacteria are viable. OdneFill exhibited physical characteristics comparable to both BioRoot RCS and AH Plus. Any nonconformity is acceptable as the material is not a sealer. Furthermore, it is also command cured; thus, extrusion past the apex is not high risk, as it can be cured immediately. The contact angle was in the middle range between AH Plus and BioRoot RCS, thus allowing the material to retain itself within the confines of the root canal. The radiopacity was within range, indicating that OdneFill is a suitable root canal–filling material possessing all the prerequisite properties. The solubility was higher than the ISO specifications. The solubility testing of root canal sealers was undertaken using a gravimetric method with sealer discs placed in 50 mL of water. It has been shown in various studies that changes to the testing methodology result in variation in the material solubility. The final irrigating solution affects the antimicrobial properties of the obturating material . For hydraulic cement sealers, matching the irrigation to the obturating technique and sealer characteristics is crucial with chelation necessary . OdneFill was not affected by water, sodium hypochlorite, or chlorhexidine, which are used regularly in root canal therapy. The current method used to assess the interaction of irrigated dentine with the test materials/sealers also included dentine in the test. Placing materials in irrigating solutions for a length of time is not clinically relevant. The replacement of gutta-percha with a novel system has already been attempted with Resilon, which promised a monobloc technique with polycaprolactone-based points and an associated sealer . The main problem with Resilon was the degradation , thus resulting in microbial recolonization and treatment failure . The microbial degradation was assessed here for OdneFill because it is a replacement of the gutta-percha obturation. The OdneFill did not exhibit any degradation in contact bacterial enzymes. This difference in results could be attributed to the possibility that any deterioration of the material could have been masked by the increase in weight of the material through water uptake. Microbes that are not affected by chemomechanical preparation may be deterred by the antimicrobial activity of root canal–filling materials, contributing to a higher treatment success rate . BioRoot RCS has shown strong antimicrobial potential due to release of calcium ions elevating the pH of the surrounding environment as compared with set AH Plus, which has been proven to be nonantibacterial . The viability of a multispecies, endodontic biofilm was greatly reduced when grown on OdneFill and BioRoot RCS. This was confirmed by a reduced colony count. The antimicrobial characteristics of the OdneFill could potentially be associated with the presence of methacrylates in the composition . The degree of conversion was low; thus, the methacrylate diffuses in solution. This phenomenon needs further investigation. AH Plus did not exhibit antimicrobial activity indicated in the direct contact test. SEM was performed to double-check the findings from the direct contact test, which supported the results obtained in the previous test. A range of microbial species was used throughout the experimentation, as this allowed for better clinical translation since the endodontic biofilm is multispecies in nature. OdneFill, a new light-curable injectable hydrogel used for root canal obturation, exhibited comparable physical and antimicrobial properties to currently used root canal sealers. It was stable in contact with irrigating solutions and bacterial enzymes. S. Bhandari, contributed to data analysis, drafted the manuscript; S. Kuehne, contributed to conception and design, critically revised the manuscript; J. Camilleri, contributed to conception and design, drafted and critically revised the manuscript. All authors gave their final approval and agree to be accountable for all aspects of the work sj-docx-1-jdr-10.1177_00220345241287504 – Supplemental material for Assessment of a Light-Curable Hydrogel to Be Used for Root Canal Obturation Supplemental material, sj-docx-1-jdr-10.1177_00220345241287504 for Assessment of a Light-Curable Hydrogel to Be Used for Root Canal Obturation by S. Bhandari, S. Kuehne and J. Camilleri in Journal of Dental Research |
The Silva Pattern-based Classification for HPV-associated Invasive Endocervical Adenocarcinoma and the Distinction Between In Situ and Invasive Adenocarcinoma: Relevant Issues and Recommendations From the International Society of Gynecological Pathologists | 4c779181-a7c0-4429-a347-05c2f86fd9a3 | 7969170 | Gynaecology[mh] | Introduction Cervical cancer is staged according to the Fédération Internationale de Gynécologie et d´Óbstétrique (FIGO) system , using a combination of clinical, imaging, and pathology findings. The experience with this staging system, however, is based primarily on studies of squamous cell carcinoma, which is by far more common, and has been extrapolated to adenocarcinoma. As a result, both adenocarcinomas and squamous carcinomas are staged and treated similarly, although there is increasing evidence to suggest that adenocarcinomas show different epidemiology, prognostic factors, patterns of spread and failure after treatment compared with squamous cell carcinomas , . Staging of FIGO IA1, IA2, and IB1 invasive endocervical adenocarcinomas (EACs) is currently based on the depth of invasion , . However, an accurate assessment of this parameter can be challenging in a variety of scenarios such as: (1) well-differentiated invasive adenocarcinomas without architectural complexity and no stromal reaction that are difficult to distinguish from in-situ adenocarcinoma, (2) tumors where it is not possible to separate the invasive from the in situ component, (3) polypoid lesions, and (4) specimens lacking proper orientation or integrity of the mucosal surface. In spite of these potential challenges, depth of invasion is a major determinant of treatment. According to the current National Comprehensive Cancer Network (NCCN) guidelines, patients with FIGO stage IA1 tumors that lack lymphovascular invasion (LVI) could undergo conservative treatment with conization and follow-up (if margins are negative) or simple hysterectomy when preservation of fertility is not required. Patients with FIGO stage IA2 tumors and those with IA1 tumors associated with LVI or with positive margins undergo radical surgery (radical hysterectomy, or alternatively large conization or radical trachelectomy as fertility preservation approaches); sentinel lymph node (SLN) mapping and/or pelvic lymph node (LN) dissection are also considered in this group of patients . Patients that undergo simple/radical hysterectomy or radical trachelectomy may experience surgical complications such as bladder dysfunction, vascular or ureteral injuries, and blood loss among others . In addition, 10% to 41% of patients treated with LN dissection can experience lower extremity lymphedema as postoperative morbidity , . Importantly, the literature indicates that few patients with early FIGO stage tumors have evidence of LN metastasis, seen in <1% of patients with stage IA1 tumors and in ~2% of patients with stage IA2 tumors . In an attempt to improve the current risk stratification system for patients affected by HPV-associated invasive cervical adenocarcinoma, a group of pathologists led by Dr Elvio Silva have proposed the use of a system based on the following histologic features: tumor-stromal interface, presence or absence of LVI, architecture and grade of cytologic atypia , . The Silva Pattern-based Classification: Definitions The Silva classification stratifies HPV-associated invasive endocervical adenocarcinoma into 3 patterns (A, B, C) based on the presence or absence of destructive stromal invasion, the degree of destructive stromal invasion (if present), the presence or absence of LVI, and grade of cytologic atypia. This classification system does not take into account the depth of invasion or the relationship of the tumor to large vessels in the cervical stroma (Table ). The definitions and specific cut-offs presented herein have been established by consensus of the original group that defined the Silva classification , . Pattern A This pattern is characterized by the absence of destructive stromal invasion (i.e. there is no desmoplasia or associated inflammatory infiltrate with single cells or detached clusters of tumor cells within the stroma). It consists of well-demarcated glands with rounded contours, commonly forming groups that sometimes have a relatively well-preserved lobular architecture (Figs. , ). This pattern should not be diagnosed in the presence of high-grade cytologic features or solid architecture. Although cribriform or papillary growth may be seen, these should not fill a 4× field (5 mm in diameter). As stated above, this cut-off is not evidence based and was obtained by consensus of the investigators who developed this system. LVI is absent in pattern A, thus when considering assigning this pattern, close scrutiny should be carried out in order to exclude this finding, this may require levels or immunohistochemical studies. As destructive invasion needs to be excluded, a pattern A designation requires examination of the entire tumor on excisional material [eg, loop electrosurgical excision procedure (LEEP) or cone] with negative resection margins. Pattern B This pattern’s hallmark is the presence of localized (early, limited) destructive stromal invasion. It consists of tumor nests, ragged glands or individual cells budding off well-demarcated glands (pattern A) and usually associated with an inflamed or desmoplastic stroma. Foci of localized destructive invasion may be single, multiple, or linear at the base of tumor, but they should not fill a 4× field (5 mm in diameter). No solid growth is seen while LVI may be present or absent (Figs. , ). Pattern C The presence of diffuse destructive stromal invasion is the cardinal feature of this pattern. Tumors with pattern C can show any of the following histological appearances: A growth of haphazardly distributed, variably sized and shaped, often angulated glands in a desmoplastic stroma; the glands can be interconnected (canalicular pattern), and sometimes they are interspersed with dilated, elongated, and fragmented glands that resemble those seen in the microcystic, elongated and fragmented pattern of invasion of endometrial endometrioid carcinomas (Figs. A–C). A confluent glandular or papillary growth with minimal intervening stroma—endophytic growth only, or mucin lakes with tumor cells filling a 4× field (5 mm in diameter) (Fig. D). A micropapillary growth with small papillae composed only of tumor cells, lacking fibrovascular cores, and surrounded by clear spaces. A linear proliferation of irregular glands and individual cells in a desmoplastic stroma at the base of a partially exophytic tumor and filling a 4× field (5 mm in diameter). A proliferation of irregular glands or tumor cell aggregates surrounded by an extensive and dense band-like inflammatory infiltrate at the base of a tumor and filling a 4× field (5 mm in diameter). A solid growth of tumor cells with small or abortive glands. LVI may be present or absent and it is not crucial for the diagnosis of pattern C. Interestingly, the micropapillary variant of pattern C has been found to be associated with large tumors, and a high propensity to have lymph node metastasis, recurrences, and an adverse outcome – . In addition, some of the investigators involved with the development of the Silva classification have recently published data that appear to indicate differences in the biologic behavior of tumors within the pattern C category; for example, tumors with a diffuse, destructive growth pattern have a tendency to recur while tumors with a band-like lymphocytic infiltrate or extensive linear destructive invasion do not. Also, patients with tumors showing a mixed diffuse and confluent destructive invasion had a worse 6-yr overall survival than patients with other subtypes of pattern C tumors . However, additional studies are needed to confirm these findings. Pathologists using the Silva classification must be aware of the following: The worst pattern seen in a given tumor is the one to be reported (i.e. tumors with pattern B and focal pattern C, should be classified as pattern C). In exophytic tumors the Silva pattern is evaluated at the tumor base within the cervical wall and not within the exophytic portion of the neoplasm. For example, an exophytic tumor with a villoglandular pattern should not be classified as a pattern C, even if complex, if the invasion at the interface with the underlying cervical wall is nondestructive (therefore, a pattern A tumor). In contrast, if the invasion at the interface with the underlying stroma shows a confluent pattern filling a 4× field (5 mm), the tumor is classified as a pattern C tumor (Fig. ). It is worth noting that exophytic lesions are challenging because their gross size by itself might determine the stage according to the current FIGO system , . Silva Pattern-based Classification: Clinical Impact We performed an exhaustive literature search using PubMed (US Library of Medicine, Bethesda, MD) and EMBASE. The search included studies published on or before February 2020, using the keywords “adenocarcinoma,” “cervix” or “endocervical,” and “pattern.” Since the first description of the Silva classification, 9 studies reporting on the subgrouping of HPV-associated EACs by pattern of invasion and outcome have been published in the English literature , – . These 9 studies amount to a total of 1319 patients with invasive endocervical adenocarcinoma. A summary of their distribution by pattern is shown in Table while rates of lymph node metastases, stage distribution, recurrence, and cancer-related deaths are summarized in Table . A total of 253 (19%) patients had tumors with pattern A. None had lymph node metastases. Stage information was available in 224 patients of which 222 (99%) had stage I tumors at presentation. A total of 201 patients had available follow-up (range, 3–352 mo; median, 62 mo) and none had documented recurrences or cancer-related deaths. A total of 262 (20%) patients had tumors with pattern B. Fifty-three (20%) had LVI, and 14 (5%) had lymph node metastases. Of the 239 patients with stage information, 233 (97%) had FIGO stage I, and 6 (2.5%) stage II tumors. Among 216 patients with follow-up (range, 5–392 mo; median, 69 mo) 7 (3%) developed recurrences: 2 patients exhibited locoregional recurrences, while 1 each developed ovarian and vaginal recurrence; information was not available in the remaining 3 patients. Three (1%) patients died of disease. A total of 804 (61%) patients had pattern C tumors. LVI was present in 490 (61%) tumors, and lymph node metastases in 177 (22%). Compared with patients with pattern A and pattern B adenocarcinomas, the proportion of patients with stage I disease in this group was lower (526 of 789 cases with staging information available, 65%). Among the 359 patients with follow-up (range, 3–258 mo; median, 55 mo), 70 developed recurrences (19%), 11 in the vagina, 6 were locoregional, and approximately half had distant metastases. Cancer-related death occurred in 39 (11%) patients. Current evidence, while retrospective, supports the use of the Silva classification for the clinical management of patients with HPV-associated invasive adenocarcinomas. The differences in outcome suggest that patients with pattern A adenocarcinomas can be treated conservatively with conization with negative margins and no lymph node dissection, similar to patients with adenocarcinoma in situ (AIS). Follow-up of these patients is still required as rare examples of cervical tumors with an in situ adenocarcinoma appearance have been associated with ovarian metastasis . In contrast, patients with pattern B tumors with LVI may benefit from SLN mapping or a limited LN sampling. This recommendation differs from an initial recommendation where all patients with pattern B tumors were thought to benefit from SLN mapping. Currently, it is felt that patients with pattern B tumors with no LVI should be treated as those with pattern A tumors. Patients with pattern C tumors have the highest prevalence of adverse outcomes, and therefore are more likely to benefit from standard surgical treatment including SLN sampling or LN dissection. Substratification of pattern C into variants with less (extensive linear, band-like lymphocytic) versus more aggressive (diffusely destructive or confluent, micropapillary) behavior may help in the future to choose specific management strategies – . The role of systemic therapy (chemotherapy and/or radiation) in patients stratified by pattern of invasion is, to date, unclear. Reproducibility The interobserver reproducibility of the Silva pattern-based classification has been addressed by 3 independent studies to date. The first study included 2 institutions and 49 cases of usual type invasive adenocarcinoma . The investigators found consensus diagnosis in 50% of cases, with kappa values ranging from fair to almost perfect agreement (range, 0.24–0.84); kappa agreement improved when using a 2-tier system (pattern A vs. pattern B or C). The second study was multi-institutional, included 96 cases and found a good overall reproducibility (κ=0.65). While perfect agreement (9/9 reviewers) was seen in only 11 cases (11%), consensus (≥5/9 reviewer) concordance was achieved in 82/96 cases (85%). Interobserver agreement was the highest when distinguishing in situ adenocarcinoma and pattern A from pattern B and C tumors. Poor agreement was seen in the distinction between in situ adenocarcinoma and pattern A adenocarcinoma . The third study was also muti-institutional, encompassed 84 cases, and found an overall concordance of 74% with kappa values of 0.54, 0.32 and 0.59 for patterns A, B, and C, respectively . We conclude that the Silva pattern-based classification has overall an acceptable reproducibility, especially when distinguishing pattern A from pattern B or C tumors. Pathologists are encouraged to become proficient in using this classification by completing the ISGyP training module on the Silva classification ( http://www.gpecimage.ubc.ca/aperio/images/eac/ ). This resource offers training and test sets of cervical adenocarcinomas classified by pattern of invasion. Lastly, routine intradepartmental consultation and consensus opinion with colleagues, at least in difficult cases, can be helpful. Current Issues and Recommendations Reporting of Pattern of Invasion The Silva classification is not part of the current FIGO or American Joint Commission on Cancer (AJCC) staging systems . Nonetheless, it is now mentioned in synoptic reporting guidelines such as the College of American Pathologists (as a fillable field under “Stromal Invasion”) and the International Collaboration on Cancer Reporting (as an explanatory note under “grading”) , . Moreover, the latest National Comprehensive Cancer Network (NCCN) guidelines introduce the Silva classification as an “emerging concept” . We recommend including these patterns of invasion in the pathology reports with a diagnosis of invasive HPV-associated endocervical adenocarcinoma. The pattern of invasion can be included as a subheading of the main diagnosis line, or in the comment section. The former is preferred by this group. Including an explanatory note can also be considered. Specimen Type and Silva Pattern-based Classification A prerequisite for the application of the Silva classification is the histologic examination of the entire tumor. Thus, pattern assignment is best done in a cone or LEEP with negative margins, or in a hysterectomy or trachelectomy specimen. Biopsy material is not suitable for pattern assignment given its limited size and superficial nature , . Conversely, it has been shown that the Silva pattern of invasion in LEEP and cone material is highly predictive of the overall pattern of residual tumor in hysterectomy) , . LVI and the Silva Pattern-based Classification LVI is an important parameter in the management of cervical cancer. However, not all studies support its independent prognostic significance, especially in multivariate analyses. Creasman and Kohler reviewed the published literature encompassing 25 studies with data on 6500 patients with early cervical cancer and LVI; only 3 (12%) studies found LVI as an independent risk factor. In a study focused on 127 patients with pattern C EACs. Roma et al. found that LVI was not an independent predictor of survival. Despite this evidence, it is still important to report the LVI status as it currently affects patient management. In terms of the value of quantifying LVI in cervical adenocarcinoma, a study of 189 pattern C tumors showed that the extent of lymphatic vascular invasion may have prognostic significance, as those with extensive LVI (≥20 individual spaces containing tumor) had significantly higher rates of lymph node metastases and recurrence compared to those with low volume LVI (0–4 spaces) . This evidence suggests a potential role for quantifying the extent of LVI similar to endometrial carcinoma. However, further studies are needed to confirm this finding. HPV-independent Adenocarcinoma and the Silva Pattern-based Classification The Silva classification was conceived using cohorts of usual-type adenocarcinomas, and it is applicable to this tumor type as outlined in the seminal study by Diaz de Vivar et al. . It is also applicable to other types of HPV-related adenocarcinoma as recently demonstrated by Stolnicu et al. . Conversely, patients with HPV-independent adenocarcinomas, gastric-type being most common, do not benefit from pattern-based stratification as most show pattern C invasion even when well-differentiated (namely gastric-type adenocarcinoma, minimal deviation type) and are associated with poor prognosis. Cervical cancer is staged according to the Fédération Internationale de Gynécologie et d´Óbstétrique (FIGO) system , using a combination of clinical, imaging, and pathology findings. The experience with this staging system, however, is based primarily on studies of squamous cell carcinoma, which is by far more common, and has been extrapolated to adenocarcinoma. As a result, both adenocarcinomas and squamous carcinomas are staged and treated similarly, although there is increasing evidence to suggest that adenocarcinomas show different epidemiology, prognostic factors, patterns of spread and failure after treatment compared with squamous cell carcinomas , . Staging of FIGO IA1, IA2, and IB1 invasive endocervical adenocarcinomas (EACs) is currently based on the depth of invasion , . However, an accurate assessment of this parameter can be challenging in a variety of scenarios such as: (1) well-differentiated invasive adenocarcinomas without architectural complexity and no stromal reaction that are difficult to distinguish from in-situ adenocarcinoma, (2) tumors where it is not possible to separate the invasive from the in situ component, (3) polypoid lesions, and (4) specimens lacking proper orientation or integrity of the mucosal surface. In spite of these potential challenges, depth of invasion is a major determinant of treatment. According to the current National Comprehensive Cancer Network (NCCN) guidelines, patients with FIGO stage IA1 tumors that lack lymphovascular invasion (LVI) could undergo conservative treatment with conization and follow-up (if margins are negative) or simple hysterectomy when preservation of fertility is not required. Patients with FIGO stage IA2 tumors and those with IA1 tumors associated with LVI or with positive margins undergo radical surgery (radical hysterectomy, or alternatively large conization or radical trachelectomy as fertility preservation approaches); sentinel lymph node (SLN) mapping and/or pelvic lymph node (LN) dissection are also considered in this group of patients . Patients that undergo simple/radical hysterectomy or radical trachelectomy may experience surgical complications such as bladder dysfunction, vascular or ureteral injuries, and blood loss among others . In addition, 10% to 41% of patients treated with LN dissection can experience lower extremity lymphedema as postoperative morbidity , . Importantly, the literature indicates that few patients with early FIGO stage tumors have evidence of LN metastasis, seen in <1% of patients with stage IA1 tumors and in ~2% of patients with stage IA2 tumors . In an attempt to improve the current risk stratification system for patients affected by HPV-associated invasive cervical adenocarcinoma, a group of pathologists led by Dr Elvio Silva have proposed the use of a system based on the following histologic features: tumor-stromal interface, presence or absence of LVI, architecture and grade of cytologic atypia , . The Silva classification stratifies HPV-associated invasive endocervical adenocarcinoma into 3 patterns (A, B, C) based on the presence or absence of destructive stromal invasion, the degree of destructive stromal invasion (if present), the presence or absence of LVI, and grade of cytologic atypia. This classification system does not take into account the depth of invasion or the relationship of the tumor to large vessels in the cervical stroma (Table ). The definitions and specific cut-offs presented herein have been established by consensus of the original group that defined the Silva classification , . Pattern A This pattern is characterized by the absence of destructive stromal invasion (i.e. there is no desmoplasia or associated inflammatory infiltrate with single cells or detached clusters of tumor cells within the stroma). It consists of well-demarcated glands with rounded contours, commonly forming groups that sometimes have a relatively well-preserved lobular architecture (Figs. , ). This pattern should not be diagnosed in the presence of high-grade cytologic features or solid architecture. Although cribriform or papillary growth may be seen, these should not fill a 4× field (5 mm in diameter). As stated above, this cut-off is not evidence based and was obtained by consensus of the investigators who developed this system. LVI is absent in pattern A, thus when considering assigning this pattern, close scrutiny should be carried out in order to exclude this finding, this may require levels or immunohistochemical studies. As destructive invasion needs to be excluded, a pattern A designation requires examination of the entire tumor on excisional material [eg, loop electrosurgical excision procedure (LEEP) or cone] with negative resection margins. Pattern B This pattern’s hallmark is the presence of localized (early, limited) destructive stromal invasion. It consists of tumor nests, ragged glands or individual cells budding off well-demarcated glands (pattern A) and usually associated with an inflamed or desmoplastic stroma. Foci of localized destructive invasion may be single, multiple, or linear at the base of tumor, but they should not fill a 4× field (5 mm in diameter). No solid growth is seen while LVI may be present or absent (Figs. , ). Pattern C The presence of diffuse destructive stromal invasion is the cardinal feature of this pattern. Tumors with pattern C can show any of the following histological appearances: A growth of haphazardly distributed, variably sized and shaped, often angulated glands in a desmoplastic stroma; the glands can be interconnected (canalicular pattern), and sometimes they are interspersed with dilated, elongated, and fragmented glands that resemble those seen in the microcystic, elongated and fragmented pattern of invasion of endometrial endometrioid carcinomas (Figs. A–C). A confluent glandular or papillary growth with minimal intervening stroma—endophytic growth only, or mucin lakes with tumor cells filling a 4× field (5 mm in diameter) (Fig. D). A micropapillary growth with small papillae composed only of tumor cells, lacking fibrovascular cores, and surrounded by clear spaces. A linear proliferation of irregular glands and individual cells in a desmoplastic stroma at the base of a partially exophytic tumor and filling a 4× field (5 mm in diameter). A proliferation of irregular glands or tumor cell aggregates surrounded by an extensive and dense band-like inflammatory infiltrate at the base of a tumor and filling a 4× field (5 mm in diameter). A solid growth of tumor cells with small or abortive glands. LVI may be present or absent and it is not crucial for the diagnosis of pattern C. Interestingly, the micropapillary variant of pattern C has been found to be associated with large tumors, and a high propensity to have lymph node metastasis, recurrences, and an adverse outcome – . In addition, some of the investigators involved with the development of the Silva classification have recently published data that appear to indicate differences in the biologic behavior of tumors within the pattern C category; for example, tumors with a diffuse, destructive growth pattern have a tendency to recur while tumors with a band-like lymphocytic infiltrate or extensive linear destructive invasion do not. Also, patients with tumors showing a mixed diffuse and confluent destructive invasion had a worse 6-yr overall survival than patients with other subtypes of pattern C tumors . However, additional studies are needed to confirm these findings. Pathologists using the Silva classification must be aware of the following: The worst pattern seen in a given tumor is the one to be reported (i.e. tumors with pattern B and focal pattern C, should be classified as pattern C). In exophytic tumors the Silva pattern is evaluated at the tumor base within the cervical wall and not within the exophytic portion of the neoplasm. For example, an exophytic tumor with a villoglandular pattern should not be classified as a pattern C, even if complex, if the invasion at the interface with the underlying cervical wall is nondestructive (therefore, a pattern A tumor). In contrast, if the invasion at the interface with the underlying stroma shows a confluent pattern filling a 4× field (5 mm), the tumor is classified as a pattern C tumor (Fig. ). It is worth noting that exophytic lesions are challenging because their gross size by itself might determine the stage according to the current FIGO system , . This pattern is characterized by the absence of destructive stromal invasion (i.e. there is no desmoplasia or associated inflammatory infiltrate with single cells or detached clusters of tumor cells within the stroma). It consists of well-demarcated glands with rounded contours, commonly forming groups that sometimes have a relatively well-preserved lobular architecture (Figs. , ). This pattern should not be diagnosed in the presence of high-grade cytologic features or solid architecture. Although cribriform or papillary growth may be seen, these should not fill a 4× field (5 mm in diameter). As stated above, this cut-off is not evidence based and was obtained by consensus of the investigators who developed this system. LVI is absent in pattern A, thus when considering assigning this pattern, close scrutiny should be carried out in order to exclude this finding, this may require levels or immunohistochemical studies. As destructive invasion needs to be excluded, a pattern A designation requires examination of the entire tumor on excisional material [eg, loop electrosurgical excision procedure (LEEP) or cone] with negative resection margins. This pattern’s hallmark is the presence of localized (early, limited) destructive stromal invasion. It consists of tumor nests, ragged glands or individual cells budding off well-demarcated glands (pattern A) and usually associated with an inflamed or desmoplastic stroma. Foci of localized destructive invasion may be single, multiple, or linear at the base of tumor, but they should not fill a 4× field (5 mm in diameter). No solid growth is seen while LVI may be present or absent (Figs. , ). The presence of diffuse destructive stromal invasion is the cardinal feature of this pattern. Tumors with pattern C can show any of the following histological appearances: A growth of haphazardly distributed, variably sized and shaped, often angulated glands in a desmoplastic stroma; the glands can be interconnected (canalicular pattern), and sometimes they are interspersed with dilated, elongated, and fragmented glands that resemble those seen in the microcystic, elongated and fragmented pattern of invasion of endometrial endometrioid carcinomas (Figs. A–C). A confluent glandular or papillary growth with minimal intervening stroma—endophytic growth only, or mucin lakes with tumor cells filling a 4× field (5 mm in diameter) (Fig. D). A micropapillary growth with small papillae composed only of tumor cells, lacking fibrovascular cores, and surrounded by clear spaces. A linear proliferation of irregular glands and individual cells in a desmoplastic stroma at the base of a partially exophytic tumor and filling a 4× field (5 mm in diameter). A proliferation of irregular glands or tumor cell aggregates surrounded by an extensive and dense band-like inflammatory infiltrate at the base of a tumor and filling a 4× field (5 mm in diameter). A solid growth of tumor cells with small or abortive glands. LVI may be present or absent and it is not crucial for the diagnosis of pattern C. Interestingly, the micropapillary variant of pattern C has been found to be associated with large tumors, and a high propensity to have lymph node metastasis, recurrences, and an adverse outcome – . In addition, some of the investigators involved with the development of the Silva classification have recently published data that appear to indicate differences in the biologic behavior of tumors within the pattern C category; for example, tumors with a diffuse, destructive growth pattern have a tendency to recur while tumors with a band-like lymphocytic infiltrate or extensive linear destructive invasion do not. Also, patients with tumors showing a mixed diffuse and confluent destructive invasion had a worse 6-yr overall survival than patients with other subtypes of pattern C tumors . However, additional studies are needed to confirm these findings. Pathologists using the Silva classification must be aware of the following: The worst pattern seen in a given tumor is the one to be reported (i.e. tumors with pattern B and focal pattern C, should be classified as pattern C). In exophytic tumors the Silva pattern is evaluated at the tumor base within the cervical wall and not within the exophytic portion of the neoplasm. For example, an exophytic tumor with a villoglandular pattern should not be classified as a pattern C, even if complex, if the invasion at the interface with the underlying cervical wall is nondestructive (therefore, a pattern A tumor). In contrast, if the invasion at the interface with the underlying stroma shows a confluent pattern filling a 4× field (5 mm), the tumor is classified as a pattern C tumor (Fig. ). It is worth noting that exophytic lesions are challenging because their gross size by itself might determine the stage according to the current FIGO system , . We performed an exhaustive literature search using PubMed (US Library of Medicine, Bethesda, MD) and EMBASE. The search included studies published on or before February 2020, using the keywords “adenocarcinoma,” “cervix” or “endocervical,” and “pattern.” Since the first description of the Silva classification, 9 studies reporting on the subgrouping of HPV-associated EACs by pattern of invasion and outcome have been published in the English literature , – . These 9 studies amount to a total of 1319 patients with invasive endocervical adenocarcinoma. A summary of their distribution by pattern is shown in Table while rates of lymph node metastases, stage distribution, recurrence, and cancer-related deaths are summarized in Table . A total of 253 (19%) patients had tumors with pattern A. None had lymph node metastases. Stage information was available in 224 patients of which 222 (99%) had stage I tumors at presentation. A total of 201 patients had available follow-up (range, 3–352 mo; median, 62 mo) and none had documented recurrences or cancer-related deaths. A total of 262 (20%) patients had tumors with pattern B. Fifty-three (20%) had LVI, and 14 (5%) had lymph node metastases. Of the 239 patients with stage information, 233 (97%) had FIGO stage I, and 6 (2.5%) stage II tumors. Among 216 patients with follow-up (range, 5–392 mo; median, 69 mo) 7 (3%) developed recurrences: 2 patients exhibited locoregional recurrences, while 1 each developed ovarian and vaginal recurrence; information was not available in the remaining 3 patients. Three (1%) patients died of disease. A total of 804 (61%) patients had pattern C tumors. LVI was present in 490 (61%) tumors, and lymph node metastases in 177 (22%). Compared with patients with pattern A and pattern B adenocarcinomas, the proportion of patients with stage I disease in this group was lower (526 of 789 cases with staging information available, 65%). Among the 359 patients with follow-up (range, 3–258 mo; median, 55 mo), 70 developed recurrences (19%), 11 in the vagina, 6 were locoregional, and approximately half had distant metastases. Cancer-related death occurred in 39 (11%) patients. Current evidence, while retrospective, supports the use of the Silva classification for the clinical management of patients with HPV-associated invasive adenocarcinomas. The differences in outcome suggest that patients with pattern A adenocarcinomas can be treated conservatively with conization with negative margins and no lymph node dissection, similar to patients with adenocarcinoma in situ (AIS). Follow-up of these patients is still required as rare examples of cervical tumors with an in situ adenocarcinoma appearance have been associated with ovarian metastasis . In contrast, patients with pattern B tumors with LVI may benefit from SLN mapping or a limited LN sampling. This recommendation differs from an initial recommendation where all patients with pattern B tumors were thought to benefit from SLN mapping. Currently, it is felt that patients with pattern B tumors with no LVI should be treated as those with pattern A tumors. Patients with pattern C tumors have the highest prevalence of adverse outcomes, and therefore are more likely to benefit from standard surgical treatment including SLN sampling or LN dissection. Substratification of pattern C into variants with less (extensive linear, band-like lymphocytic) versus more aggressive (diffusely destructive or confluent, micropapillary) behavior may help in the future to choose specific management strategies – . The role of systemic therapy (chemotherapy and/or radiation) in patients stratified by pattern of invasion is, to date, unclear. The interobserver reproducibility of the Silva pattern-based classification has been addressed by 3 independent studies to date. The first study included 2 institutions and 49 cases of usual type invasive adenocarcinoma . The investigators found consensus diagnosis in 50% of cases, with kappa values ranging from fair to almost perfect agreement (range, 0.24–0.84); kappa agreement improved when using a 2-tier system (pattern A vs. pattern B or C). The second study was multi-institutional, included 96 cases and found a good overall reproducibility (κ=0.65). While perfect agreement (9/9 reviewers) was seen in only 11 cases (11%), consensus (≥5/9 reviewer) concordance was achieved in 82/96 cases (85%). Interobserver agreement was the highest when distinguishing in situ adenocarcinoma and pattern A from pattern B and C tumors. Poor agreement was seen in the distinction between in situ adenocarcinoma and pattern A adenocarcinoma . The third study was also muti-institutional, encompassed 84 cases, and found an overall concordance of 74% with kappa values of 0.54, 0.32 and 0.59 for patterns A, B, and C, respectively . We conclude that the Silva pattern-based classification has overall an acceptable reproducibility, especially when distinguishing pattern A from pattern B or C tumors. Pathologists are encouraged to become proficient in using this classification by completing the ISGyP training module on the Silva classification ( http://www.gpecimage.ubc.ca/aperio/images/eac/ ). This resource offers training and test sets of cervical adenocarcinomas classified by pattern of invasion. Lastly, routine intradepartmental consultation and consensus opinion with colleagues, at least in difficult cases, can be helpful. Reporting of Pattern of Invasion The Silva classification is not part of the current FIGO or American Joint Commission on Cancer (AJCC) staging systems . Nonetheless, it is now mentioned in synoptic reporting guidelines such as the College of American Pathologists (as a fillable field under “Stromal Invasion”) and the International Collaboration on Cancer Reporting (as an explanatory note under “grading”) , . Moreover, the latest National Comprehensive Cancer Network (NCCN) guidelines introduce the Silva classification as an “emerging concept” . We recommend including these patterns of invasion in the pathology reports with a diagnosis of invasive HPV-associated endocervical adenocarcinoma. The pattern of invasion can be included as a subheading of the main diagnosis line, or in the comment section. The former is preferred by this group. Including an explanatory note can also be considered. Specimen Type and Silva Pattern-based Classification A prerequisite for the application of the Silva classification is the histologic examination of the entire tumor. Thus, pattern assignment is best done in a cone or LEEP with negative margins, or in a hysterectomy or trachelectomy specimen. Biopsy material is not suitable for pattern assignment given its limited size and superficial nature , . Conversely, it has been shown that the Silva pattern of invasion in LEEP and cone material is highly predictive of the overall pattern of residual tumor in hysterectomy) , . LVI and the Silva Pattern-based Classification LVI is an important parameter in the management of cervical cancer. However, not all studies support its independent prognostic significance, especially in multivariate analyses. Creasman and Kohler reviewed the published literature encompassing 25 studies with data on 6500 patients with early cervical cancer and LVI; only 3 (12%) studies found LVI as an independent risk factor. In a study focused on 127 patients with pattern C EACs. Roma et al. found that LVI was not an independent predictor of survival. Despite this evidence, it is still important to report the LVI status as it currently affects patient management. In terms of the value of quantifying LVI in cervical adenocarcinoma, a study of 189 pattern C tumors showed that the extent of lymphatic vascular invasion may have prognostic significance, as those with extensive LVI (≥20 individual spaces containing tumor) had significantly higher rates of lymph node metastases and recurrence compared to those with low volume LVI (0–4 spaces) . This evidence suggests a potential role for quantifying the extent of LVI similar to endometrial carcinoma. However, further studies are needed to confirm this finding. HPV-independent Adenocarcinoma and the Silva Pattern-based Classification The Silva classification was conceived using cohorts of usual-type adenocarcinomas, and it is applicable to this tumor type as outlined in the seminal study by Diaz de Vivar et al. . It is also applicable to other types of HPV-related adenocarcinoma as recently demonstrated by Stolnicu et al. . Conversely, patients with HPV-independent adenocarcinomas, gastric-type being most common, do not benefit from pattern-based stratification as most show pattern C invasion even when well-differentiated (namely gastric-type adenocarcinoma, minimal deviation type) and are associated with poor prognosis. The Silva classification is not part of the current FIGO or American Joint Commission on Cancer (AJCC) staging systems . Nonetheless, it is now mentioned in synoptic reporting guidelines such as the College of American Pathologists (as a fillable field under “Stromal Invasion”) and the International Collaboration on Cancer Reporting (as an explanatory note under “grading”) , . Moreover, the latest National Comprehensive Cancer Network (NCCN) guidelines introduce the Silva classification as an “emerging concept” . We recommend including these patterns of invasion in the pathology reports with a diagnosis of invasive HPV-associated endocervical adenocarcinoma. The pattern of invasion can be included as a subheading of the main diagnosis line, or in the comment section. The former is preferred by this group. Including an explanatory note can also be considered. A prerequisite for the application of the Silva classification is the histologic examination of the entire tumor. Thus, pattern assignment is best done in a cone or LEEP with negative margins, or in a hysterectomy or trachelectomy specimen. Biopsy material is not suitable for pattern assignment given its limited size and superficial nature , . Conversely, it has been shown that the Silva pattern of invasion in LEEP and cone material is highly predictive of the overall pattern of residual tumor in hysterectomy) , . LVI is an important parameter in the management of cervical cancer. However, not all studies support its independent prognostic significance, especially in multivariate analyses. Creasman and Kohler reviewed the published literature encompassing 25 studies with data on 6500 patients with early cervical cancer and LVI; only 3 (12%) studies found LVI as an independent risk factor. In a study focused on 127 patients with pattern C EACs. Roma et al. found that LVI was not an independent predictor of survival. Despite this evidence, it is still important to report the LVI status as it currently affects patient management. In terms of the value of quantifying LVI in cervical adenocarcinoma, a study of 189 pattern C tumors showed that the extent of lymphatic vascular invasion may have prognostic significance, as those with extensive LVI (≥20 individual spaces containing tumor) had significantly higher rates of lymph node metastases and recurrence compared to those with low volume LVI (0–4 spaces) . This evidence suggests a potential role for quantifying the extent of LVI similar to endometrial carcinoma. However, further studies are needed to confirm this finding. The Silva classification was conceived using cohorts of usual-type adenocarcinomas, and it is applicable to this tumor type as outlined in the seminal study by Diaz de Vivar et al. . It is also applicable to other types of HPV-related adenocarcinoma as recently demonstrated by Stolnicu et al. . Conversely, patients with HPV-independent adenocarcinomas, gastric-type being most common, do not benefit from pattern-based stratification as most show pattern C invasion even when well-differentiated (namely gastric-type adenocarcinoma, minimal deviation type) and are associated with poor prognosis. HPV-associated Endocervical Adenocarcinoma Definitions HPV-associated AIS is defined as a proliferation of neoplastic glandular cells confined to the epithelial endocervical compartment and related to infection by high-risk HPV. From a histopathologic perspective, HPV-associated AIS is defined by the following criteria – . Architecture: as the neoplastic cells are replacing the preexisting normal endocervical cells, there is preservation of the normal glandular architecture. Intraglandular and/or surface architectural complexity is allowed (papillary, micropapillary or cribriform growth), but should be limited (Fig. ). Cytology: columnar cells with enlarged, elongated or plump, hyperchromatic nuclei, mucin-depleted (more often) or mucin rich epithelium, and easily identifiable apical mitotic figures and apoptotic bodies (at least one in each gland). Nuclear stratification is common (Fig. ). Histologic Variants . The most common HPV-related AIS is the usual type, which is similar to its invasive counterpart. Less commonly, HPV-related AIS is of intestinal type featuring goblet cell differentiation. The stratified variant is also known as stratified mucin-producing intraepithelial lesion—SMILE . Other variants described in literature before IECC include endometrioid and tubal. The term endometrioid no longer applies to the spectrum of HPV-associated endocervical adenocarcinoma, and its use is discouraged as the vast majority are thought to represent mucin depleted HPV related in situ adenocarcinomas – . Similarly, tubal AIS is poorly characterized in the literature although it may arise from tubal metaplasia within the cervix (Fig. ). HPV-associated invasive endocervical adenocarcinoma is defined as a proliferation of neoplastic glandular cells, related to infection by high-risk HPV, and showing cervical stromal invasion. Invasion of the cervical stroma is characterized by , : Infiltrative/destructive growth: glands with irregular or angulated contours; desmoplastic stromal reaction; non–gland-forming elements (individual cells, cell clusters, buds or nests). Complex confluent growth: anastomosed, fused or interconnected glandular elements with scant to no stroma in between; complex cribriform, labyrinth-like or solid patterns occupying a 4× field (5 mm in diameter). Under the Silva classification, the features described above define “destructive” types of invasion, namely patterns B and C (see The Silva pattern-based classification: definitions section) (Figs. – , ). A “nondestructive” or “AIS-like” pattern of growth has also been historically classified as a form of invasive carcinoma. This type of invasion is characterized by: Increased glandular density: gland crowding that deviates from the normal endocervical crypt distribution; tight clustering of small glands, sometimes with a lobulated appearance, and lacking high-grade nuclear features. Deep glandular proliferation: glands with a haphazard distribution present in deep cervical stroma without stromal reaction often in close proximity to thick-walled vessels. This growth is analogous to pattern A (Figs. , ). The cytologic features of HPV-related invasive adenocarcinoma are the same as described previously for HPV-related AIS . Current Issues Distinction Between In Situ and Invasive Adenocarcinoma . The reproducibility in distinguishing in situ and invasive endocervical adenocarcinoma is fair to poor . In fact, it has been estimated that such distinction cannot be made in as much as 20% of cases . The lowest degree of interobserver agreement is observed between AIS and pattern A adenocarcinoma . The architectural overlap between AIS and nondestructive invasive adenocarcinomas that lacks stromal reaction may explain the inconsistent interobserver agreement , although evaluating the pre-existing adjacent endocervical glands may be helpful in deciding how much complexity can be allowed to establish a diagnosis of AIS. To this point, it is important to remember that the endocervical mucosa as a complex system of mucosal infoldings, first described by Fluhmann , . The basic structural unit of the endocervix is an array of haphazardly distributed epithelial infoldings (clefts and grooves) rather than a vertical tubular gland as occurs in the endometrium. The haphazard orientation of these infoldings results in a heterogeneous, or “pattern-less,” appearance which contributes to our established limitation in distinguishing in situ adenocarcinoma (occupying pre-existing endocervix) from invasive adenocarcinomas. The biologic behavior of nondestructive (pattern A) adenocarcinoma is indolent. As discussed previously, of a total of 253 patients with pattern A adenocarcinoma reported in the literature to date, none have associated lymph node metastases , – . Moreover, no recurrences or cancer-related deaths were documented among the 201 patients reported with available follow-up (mean follow-up period 62 mo, range, 3–262 mo) , – . Given the excellent outcome of patients with tumors showing pattern A, mirroring the behavior as AIS, it has been suggested to lump pattern A tumors as part of the AIS category. However, ovarian spread has been documented in adenocarcinomas with reported AIS-like growth pattern , . In a series of 29 patients with endocervical adenocarcinoma and synchronous or metachronous ovarian metastases reported by Ronnett et al. , 11 had AIS-like appearance. The study included tumors with superficial or subtle invasion, comprised of “haphazardly distributed smaller glands in a pattern more extensive than typical AIS,” or foci suspicious but not unequivocal for invasion. Of note, a subset of cases in this study underwent review of only representative slides, not the entire histologic material. While there are no reports of pattern A tumors with ovarian metastases in the Silva classification literature, ovarian status has not usually been specified, thus a definitive statement about pattern A tumors and their potential risk of ovarian metastasis cannot be provided. It can be inferred that tumors with ovarian spread and AIS-like growth reported represent pattern A lesions; however, this needs to be confirmed by further studies that describe the true prevalence of ovarian metastases in lesions defined as per the Silva system. Therefore, it is advisable to follow-up patients with pattern A cervical adenocarcinomas. Recommendations On the basis of the above cumulative evidence, and while more data on the risk of ovarian spread by pattern A adenocarcinomas becomes available, we advise against categorizing pattern A adenocarcinomas as AIS. Instead, we recommend an approach that emphasizes the tumor growth pattern, as follows (Fig. ): Look for destructive stromal invasion. If present, diagnosis of “invasive endocervical adenocarcinoma” is appropriate. If destructive invasion is absent, determine if the lesion is within the volume and distribution expected for an in situ lesion; if so, the diagnosis of “AIS” is appropriate. If the lesion exceeds the volume and distribution expected for AIS, or the distinction between AIS and invasive is difficult, the diagnosis of “ pattern-A (non-destructive) adenocarcinoma ” is appropriate. In the context of marked inflammation, mucosal erosion or ulceration and previous biopsy site reaction, the architecture of the lesion may be distorted. Evaluation of growth and stromal invasion should be made in other areas. In all the above considerations, consensus review with colleagues and outside consultation are helpful steps to reach a final diagnosis. HPV-independent Endocervical Adenocarcinoma According to the new classification of endocervical adenocarcinoma, the HPV-independent category includes gastric, clear cell, mesonephric, and endometrioid types. Among these subtypes, there is emerging evidence on the spectrum of in situ gastric type endocervical neoplasia. In situ counterparts for clear cell, mesonephric and endometrioid carcinomas of the cervix have not been described in the literature. Definitions Gastric-type AIS . This lesion is defined by architectural criteria identical to HPV-related AIS. From a cytomorphology perspective, gastric type AIS is composed of mucinous cells with abundant foamy vacuolated cytoplasm, distinct cell borders and nuclear atypia , (Figs. A, B). Intraglandular complexity, in the form of cribriform, papillary or micropapillary growth, can be seen. Atypical Lobular Endocervical Glandular Hyperplasia (LEGH) . LEGH is a benign glandular proliferation composed of cells with a gastric mucinous phenotype. As the name implies, it has an acinar (lobular) configuration, comprised of a central gland/duct surrounded by smaller round glands arranged in a floret-like pattern , . In contrast to the conventional LEGH, atypical LEGH shows a spectrum of cytologic features , . It has been proposed that this diagnosis requires the presence of at least 4 of the following, in a lesion architecturally consistent with LEGH: nuclear enlargement; irregular nuclear contours; distinct nucleoli and coarse chromatin; loss of polarity; mitoses; apoptotic bodies or luminal nuclear debris; intraluminal papillary projections. The distinction between A-LEGH and the more recently characterized gastric-type AIS is expected to be subjective, as there is overlap in the established definitions. In order to harmonize nomenclature, we recommend the term gastric-type AIS if the lesion displays significant nuclear atypia or proliferation regardless of the preexisting architecture (Figs. C, D). Current Issues Distinction Between In Situ and Invasive Gastric Type Adenocarcinoma . HPV-related adenocarcinoma can display nondestructive and destructive growth patterns, as discussed above. In contrast, the vast majority of HPV-independent adenocarcinomas, including gastric type, have a pattern C of invasion , . Thus, the distinction between in situ and invasive is typically not problematic. However, an important exception is the well-differentiated end of the spectrum of invasive gastric type adenocarcinoma, namely minimal deviation adenocarcinoma. This variant is remarkable for the highly differentiated glands, minimal to absent cytologic atypia, and absence of desmoplastic reaction , . Unlike AIS, LEGH and atypical LEGH, minimal deviation adenocarcinoma features a haphazard distribution of glands which vary greatly in size and shape, lack lobular organization and typically extend to the outer half of the cervical wall. It is important to note that the reproducibility of these criteria has not been thoroughly assessed. Recommendations In the distinction between in situ and invasive gastric-type adenocarcinoma, we provide the following recommendations: If the glandular proliferation is well-differentiated (eg, composed of well-formed glands with smooth round outlines), consider the following scenarios: Gastric type AIS: limited to the surface, similar in density and distribution to the normal endocervical glands, overt nuclear atypia. Atypical LEGH: floret-like arrangement with small, round glands surrounding a larger, duct-like structure, typically with a superficial location; nuclear atypia is variable but often present. Invasive gastric type adenocarcinoma, minimal deviation type: haphazard gland distribution with variation of gland size and shape, as well as lack of lobular architecture; extension into the deep cervical stroma. Definitions HPV-associated AIS is defined as a proliferation of neoplastic glandular cells confined to the epithelial endocervical compartment and related to infection by high-risk HPV. From a histopathologic perspective, HPV-associated AIS is defined by the following criteria – . Architecture: as the neoplastic cells are replacing the preexisting normal endocervical cells, there is preservation of the normal glandular architecture. Intraglandular and/or surface architectural complexity is allowed (papillary, micropapillary or cribriform growth), but should be limited (Fig. ). Cytology: columnar cells with enlarged, elongated or plump, hyperchromatic nuclei, mucin-depleted (more often) or mucin rich epithelium, and easily identifiable apical mitotic figures and apoptotic bodies (at least one in each gland). Nuclear stratification is common (Fig. ). Histologic Variants . The most common HPV-related AIS is the usual type, which is similar to its invasive counterpart. Less commonly, HPV-related AIS is of intestinal type featuring goblet cell differentiation. The stratified variant is also known as stratified mucin-producing intraepithelial lesion—SMILE . Other variants described in literature before IECC include endometrioid and tubal. The term endometrioid no longer applies to the spectrum of HPV-associated endocervical adenocarcinoma, and its use is discouraged as the vast majority are thought to represent mucin depleted HPV related in situ adenocarcinomas – . Similarly, tubal AIS is poorly characterized in the literature although it may arise from tubal metaplasia within the cervix (Fig. ). HPV-associated invasive endocervical adenocarcinoma is defined as a proliferation of neoplastic glandular cells, related to infection by high-risk HPV, and showing cervical stromal invasion. Invasion of the cervical stroma is characterized by , : Infiltrative/destructive growth: glands with irregular or angulated contours; desmoplastic stromal reaction; non–gland-forming elements (individual cells, cell clusters, buds or nests). Complex confluent growth: anastomosed, fused or interconnected glandular elements with scant to no stroma in between; complex cribriform, labyrinth-like or solid patterns occupying a 4× field (5 mm in diameter). Under the Silva classification, the features described above define “destructive” types of invasion, namely patterns B and C (see The Silva pattern-based classification: definitions section) (Figs. – , ). A “nondestructive” or “AIS-like” pattern of growth has also been historically classified as a form of invasive carcinoma. This type of invasion is characterized by: Increased glandular density: gland crowding that deviates from the normal endocervical crypt distribution; tight clustering of small glands, sometimes with a lobulated appearance, and lacking high-grade nuclear features. Deep glandular proliferation: glands with a haphazard distribution present in deep cervical stroma without stromal reaction often in close proximity to thick-walled vessels. This growth is analogous to pattern A (Figs. , ). The cytologic features of HPV-related invasive adenocarcinoma are the same as described previously for HPV-related AIS . Current Issues Distinction Between In Situ and Invasive Adenocarcinoma . The reproducibility in distinguishing in situ and invasive endocervical adenocarcinoma is fair to poor . In fact, it has been estimated that such distinction cannot be made in as much as 20% of cases . The lowest degree of interobserver agreement is observed between AIS and pattern A adenocarcinoma . The architectural overlap between AIS and nondestructive invasive adenocarcinomas that lacks stromal reaction may explain the inconsistent interobserver agreement , although evaluating the pre-existing adjacent endocervical glands may be helpful in deciding how much complexity can be allowed to establish a diagnosis of AIS. To this point, it is important to remember that the endocervical mucosa as a complex system of mucosal infoldings, first described by Fluhmann , . The basic structural unit of the endocervix is an array of haphazardly distributed epithelial infoldings (clefts and grooves) rather than a vertical tubular gland as occurs in the endometrium. The haphazard orientation of these infoldings results in a heterogeneous, or “pattern-less,” appearance which contributes to our established limitation in distinguishing in situ adenocarcinoma (occupying pre-existing endocervix) from invasive adenocarcinomas. The biologic behavior of nondestructive (pattern A) adenocarcinoma is indolent. As discussed previously, of a total of 253 patients with pattern A adenocarcinoma reported in the literature to date, none have associated lymph node metastases , – . Moreover, no recurrences or cancer-related deaths were documented among the 201 patients reported with available follow-up (mean follow-up period 62 mo, range, 3–262 mo) , – . Given the excellent outcome of patients with tumors showing pattern A, mirroring the behavior as AIS, it has been suggested to lump pattern A tumors as part of the AIS category. However, ovarian spread has been documented in adenocarcinomas with reported AIS-like growth pattern , . In a series of 29 patients with endocervical adenocarcinoma and synchronous or metachronous ovarian metastases reported by Ronnett et al. , 11 had AIS-like appearance. The study included tumors with superficial or subtle invasion, comprised of “haphazardly distributed smaller glands in a pattern more extensive than typical AIS,” or foci suspicious but not unequivocal for invasion. Of note, a subset of cases in this study underwent review of only representative slides, not the entire histologic material. While there are no reports of pattern A tumors with ovarian metastases in the Silva classification literature, ovarian status has not usually been specified, thus a definitive statement about pattern A tumors and their potential risk of ovarian metastasis cannot be provided. It can be inferred that tumors with ovarian spread and AIS-like growth reported represent pattern A lesions; however, this needs to be confirmed by further studies that describe the true prevalence of ovarian metastases in lesions defined as per the Silva system. Therefore, it is advisable to follow-up patients with pattern A cervical adenocarcinomas. Recommendations On the basis of the above cumulative evidence, and while more data on the risk of ovarian spread by pattern A adenocarcinomas becomes available, we advise against categorizing pattern A adenocarcinomas as AIS. Instead, we recommend an approach that emphasizes the tumor growth pattern, as follows (Fig. ): Look for destructive stromal invasion. If present, diagnosis of “invasive endocervical adenocarcinoma” is appropriate. If destructive invasion is absent, determine if the lesion is within the volume and distribution expected for an in situ lesion; if so, the diagnosis of “AIS” is appropriate. If the lesion exceeds the volume and distribution expected for AIS, or the distinction between AIS and invasive is difficult, the diagnosis of “ pattern-A (non-destructive) adenocarcinoma ” is appropriate. In the context of marked inflammation, mucosal erosion or ulceration and previous biopsy site reaction, the architecture of the lesion may be distorted. Evaluation of growth and stromal invasion should be made in other areas. In all the above considerations, consensus review with colleagues and outside consultation are helpful steps to reach a final diagnosis. HPV-associated AIS is defined as a proliferation of neoplastic glandular cells confined to the epithelial endocervical compartment and related to infection by high-risk HPV. From a histopathologic perspective, HPV-associated AIS is defined by the following criteria – . Architecture: as the neoplastic cells are replacing the preexisting normal endocervical cells, there is preservation of the normal glandular architecture. Intraglandular and/or surface architectural complexity is allowed (papillary, micropapillary or cribriform growth), but should be limited (Fig. ). Cytology: columnar cells with enlarged, elongated or plump, hyperchromatic nuclei, mucin-depleted (more often) or mucin rich epithelium, and easily identifiable apical mitotic figures and apoptotic bodies (at least one in each gland). Nuclear stratification is common (Fig. ). Histologic Variants . The most common HPV-related AIS is the usual type, which is similar to its invasive counterpart. Less commonly, HPV-related AIS is of intestinal type featuring goblet cell differentiation. The stratified variant is also known as stratified mucin-producing intraepithelial lesion—SMILE . Other variants described in literature before IECC include endometrioid and tubal. The term endometrioid no longer applies to the spectrum of HPV-associated endocervical adenocarcinoma, and its use is discouraged as the vast majority are thought to represent mucin depleted HPV related in situ adenocarcinomas – . Similarly, tubal AIS is poorly characterized in the literature although it may arise from tubal metaplasia within the cervix (Fig. ). HPV-associated invasive endocervical adenocarcinoma is defined as a proliferation of neoplastic glandular cells, related to infection by high-risk HPV, and showing cervical stromal invasion. Invasion of the cervical stroma is characterized by , : Infiltrative/destructive growth: glands with irregular or angulated contours; desmoplastic stromal reaction; non–gland-forming elements (individual cells, cell clusters, buds or nests). Complex confluent growth: anastomosed, fused or interconnected glandular elements with scant to no stroma in between; complex cribriform, labyrinth-like or solid patterns occupying a 4× field (5 mm in diameter). Under the Silva classification, the features described above define “destructive” types of invasion, namely patterns B and C (see The Silva pattern-based classification: definitions section) (Figs. – , ). A “nondestructive” or “AIS-like” pattern of growth has also been historically classified as a form of invasive carcinoma. This type of invasion is characterized by: Increased glandular density: gland crowding that deviates from the normal endocervical crypt distribution; tight clustering of small glands, sometimes with a lobulated appearance, and lacking high-grade nuclear features. Deep glandular proliferation: glands with a haphazard distribution present in deep cervical stroma without stromal reaction often in close proximity to thick-walled vessels. This growth is analogous to pattern A (Figs. , ). The cytologic features of HPV-related invasive adenocarcinoma are the same as described previously for HPV-related AIS . Distinction Between In Situ and Invasive Adenocarcinoma . The reproducibility in distinguishing in situ and invasive endocervical adenocarcinoma is fair to poor . In fact, it has been estimated that such distinction cannot be made in as much as 20% of cases . The lowest degree of interobserver agreement is observed between AIS and pattern A adenocarcinoma . The architectural overlap between AIS and nondestructive invasive adenocarcinomas that lacks stromal reaction may explain the inconsistent interobserver agreement , although evaluating the pre-existing adjacent endocervical glands may be helpful in deciding how much complexity can be allowed to establish a diagnosis of AIS. To this point, it is important to remember that the endocervical mucosa as a complex system of mucosal infoldings, first described by Fluhmann , . The basic structural unit of the endocervix is an array of haphazardly distributed epithelial infoldings (clefts and grooves) rather than a vertical tubular gland as occurs in the endometrium. The haphazard orientation of these infoldings results in a heterogeneous, or “pattern-less,” appearance which contributes to our established limitation in distinguishing in situ adenocarcinoma (occupying pre-existing endocervix) from invasive adenocarcinomas. The biologic behavior of nondestructive (pattern A) adenocarcinoma is indolent. As discussed previously, of a total of 253 patients with pattern A adenocarcinoma reported in the literature to date, none have associated lymph node metastases , – . Moreover, no recurrences or cancer-related deaths were documented among the 201 patients reported with available follow-up (mean follow-up period 62 mo, range, 3–262 mo) , – . Given the excellent outcome of patients with tumors showing pattern A, mirroring the behavior as AIS, it has been suggested to lump pattern A tumors as part of the AIS category. However, ovarian spread has been documented in adenocarcinomas with reported AIS-like growth pattern , . In a series of 29 patients with endocervical adenocarcinoma and synchronous or metachronous ovarian metastases reported by Ronnett et al. , 11 had AIS-like appearance. The study included tumors with superficial or subtle invasion, comprised of “haphazardly distributed smaller glands in a pattern more extensive than typical AIS,” or foci suspicious but not unequivocal for invasion. Of note, a subset of cases in this study underwent review of only representative slides, not the entire histologic material. While there are no reports of pattern A tumors with ovarian metastases in the Silva classification literature, ovarian status has not usually been specified, thus a definitive statement about pattern A tumors and their potential risk of ovarian metastasis cannot be provided. It can be inferred that tumors with ovarian spread and AIS-like growth reported represent pattern A lesions; however, this needs to be confirmed by further studies that describe the true prevalence of ovarian metastases in lesions defined as per the Silva system. Therefore, it is advisable to follow-up patients with pattern A cervical adenocarcinomas. On the basis of the above cumulative evidence, and while more data on the risk of ovarian spread by pattern A adenocarcinomas becomes available, we advise against categorizing pattern A adenocarcinomas as AIS. Instead, we recommend an approach that emphasizes the tumor growth pattern, as follows (Fig. ): Look for destructive stromal invasion. If present, diagnosis of “invasive endocervical adenocarcinoma” is appropriate. If destructive invasion is absent, determine if the lesion is within the volume and distribution expected for an in situ lesion; if so, the diagnosis of “AIS” is appropriate. If the lesion exceeds the volume and distribution expected for AIS, or the distinction between AIS and invasive is difficult, the diagnosis of “ pattern-A (non-destructive) adenocarcinoma ” is appropriate. In the context of marked inflammation, mucosal erosion or ulceration and previous biopsy site reaction, the architecture of the lesion may be distorted. Evaluation of growth and stromal invasion should be made in other areas. In all the above considerations, consensus review with colleagues and outside consultation are helpful steps to reach a final diagnosis. According to the new classification of endocervical adenocarcinoma, the HPV-independent category includes gastric, clear cell, mesonephric, and endometrioid types. Among these subtypes, there is emerging evidence on the spectrum of in situ gastric type endocervical neoplasia. In situ counterparts for clear cell, mesonephric and endometrioid carcinomas of the cervix have not been described in the literature. Definitions Gastric-type AIS . This lesion is defined by architectural criteria identical to HPV-related AIS. From a cytomorphology perspective, gastric type AIS is composed of mucinous cells with abundant foamy vacuolated cytoplasm, distinct cell borders and nuclear atypia , (Figs. A, B). Intraglandular complexity, in the form of cribriform, papillary or micropapillary growth, can be seen. Atypical Lobular Endocervical Glandular Hyperplasia (LEGH) . LEGH is a benign glandular proliferation composed of cells with a gastric mucinous phenotype. As the name implies, it has an acinar (lobular) configuration, comprised of a central gland/duct surrounded by smaller round glands arranged in a floret-like pattern , . In contrast to the conventional LEGH, atypical LEGH shows a spectrum of cytologic features , . It has been proposed that this diagnosis requires the presence of at least 4 of the following, in a lesion architecturally consistent with LEGH: nuclear enlargement; irregular nuclear contours; distinct nucleoli and coarse chromatin; loss of polarity; mitoses; apoptotic bodies or luminal nuclear debris; intraluminal papillary projections. The distinction between A-LEGH and the more recently characterized gastric-type AIS is expected to be subjective, as there is overlap in the established definitions. In order to harmonize nomenclature, we recommend the term gastric-type AIS if the lesion displays significant nuclear atypia or proliferation regardless of the preexisting architecture (Figs. C, D). Current Issues Distinction Between In Situ and Invasive Gastric Type Adenocarcinoma . HPV-related adenocarcinoma can display nondestructive and destructive growth patterns, as discussed above. In contrast, the vast majority of HPV-independent adenocarcinomas, including gastric type, have a pattern C of invasion , . Thus, the distinction between in situ and invasive is typically not problematic. However, an important exception is the well-differentiated end of the spectrum of invasive gastric type adenocarcinoma, namely minimal deviation adenocarcinoma. This variant is remarkable for the highly differentiated glands, minimal to absent cytologic atypia, and absence of desmoplastic reaction , . Unlike AIS, LEGH and atypical LEGH, minimal deviation adenocarcinoma features a haphazard distribution of glands which vary greatly in size and shape, lack lobular organization and typically extend to the outer half of the cervical wall. It is important to note that the reproducibility of these criteria has not been thoroughly assessed. Recommendations In the distinction between in situ and invasive gastric-type adenocarcinoma, we provide the following recommendations: If the glandular proliferation is well-differentiated (eg, composed of well-formed glands with smooth round outlines), consider the following scenarios: Gastric type AIS: limited to the surface, similar in density and distribution to the normal endocervical glands, overt nuclear atypia. Atypical LEGH: floret-like arrangement with small, round glands surrounding a larger, duct-like structure, typically with a superficial location; nuclear atypia is variable but often present. Invasive gastric type adenocarcinoma, minimal deviation type: haphazard gland distribution with variation of gland size and shape, as well as lack of lobular architecture; extension into the deep cervical stroma. Gastric-type AIS . This lesion is defined by architectural criteria identical to HPV-related AIS. From a cytomorphology perspective, gastric type AIS is composed of mucinous cells with abundant foamy vacuolated cytoplasm, distinct cell borders and nuclear atypia , (Figs. A, B). Intraglandular complexity, in the form of cribriform, papillary or micropapillary growth, can be seen. Atypical Lobular Endocervical Glandular Hyperplasia (LEGH) . LEGH is a benign glandular proliferation composed of cells with a gastric mucinous phenotype. As the name implies, it has an acinar (lobular) configuration, comprised of a central gland/duct surrounded by smaller round glands arranged in a floret-like pattern , . In contrast to the conventional LEGH, atypical LEGH shows a spectrum of cytologic features , . It has been proposed that this diagnosis requires the presence of at least 4 of the following, in a lesion architecturally consistent with LEGH: nuclear enlargement; irregular nuclear contours; distinct nucleoli and coarse chromatin; loss of polarity; mitoses; apoptotic bodies or luminal nuclear debris; intraluminal papillary projections. The distinction between A-LEGH and the more recently characterized gastric-type AIS is expected to be subjective, as there is overlap in the established definitions. In order to harmonize nomenclature, we recommend the term gastric-type AIS if the lesion displays significant nuclear atypia or proliferation regardless of the preexisting architecture (Figs. C, D). Distinction Between In Situ and Invasive Gastric Type Adenocarcinoma . HPV-related adenocarcinoma can display nondestructive and destructive growth patterns, as discussed above. In contrast, the vast majority of HPV-independent adenocarcinomas, including gastric type, have a pattern C of invasion , . Thus, the distinction between in situ and invasive is typically not problematic. However, an important exception is the well-differentiated end of the spectrum of invasive gastric type adenocarcinoma, namely minimal deviation adenocarcinoma. This variant is remarkable for the highly differentiated glands, minimal to absent cytologic atypia, and absence of desmoplastic reaction , . Unlike AIS, LEGH and atypical LEGH, minimal deviation adenocarcinoma features a haphazard distribution of glands which vary greatly in size and shape, lack lobular organization and typically extend to the outer half of the cervical wall. It is important to note that the reproducibility of these criteria has not been thoroughly assessed. In the distinction between in situ and invasive gastric-type adenocarcinoma, we provide the following recommendations: If the glandular proliferation is well-differentiated (eg, composed of well-formed glands with smooth round outlines), consider the following scenarios: Gastric type AIS: limited to the surface, similar in density and distribution to the normal endocervical glands, overt nuclear atypia. Atypical LEGH: floret-like arrangement with small, round glands surrounding a larger, duct-like structure, typically with a superficial location; nuclear atypia is variable but often present. Invasive gastric type adenocarcinoma, minimal deviation type: haphazard gland distribution with variation of gland size and shape, as well as lack of lobular architecture; extension into the deep cervical stroma. It is our hope that these ISGyP-developed recommendations will facilitate the use of the Silva classification for HPV-associated cervical adenocarcinoma and the proper diagnosis of AIS, not only in the setting of patient care, but also in research activities which ultimately will change the staging and management of this disease. It is important to underscore the fact that enough evidence has been accumulated regarding the Silva classification to proceed with prospective studies in close collaboration with our gynecology oncology colleagues to further support the current evidence. |
Exploring the expression of DLL3 in gastroenteropancreatic neuroendocrine neoplasms and its potential diagnostic value | 10c4e767-ced7-4521-92cc-b904db3ec933 | 11770191 | Anatomy[mh] | Gastroenteropancreatic neuroendocrine neoplasms (GEP-NENs) are highly heterogeneous tumors that constituting approximately 65% of all neuroendocrine neoplasms and rank as the second most common gastrointestinal cancer , . Improved imaging techniques and enhanced awareness have contributed to an increased incidence from 1.1/100,000 in 1973 to 6.9/100,000 in 2012 . According to the 2019 WHO grading system, GEP-NENs are categorized as neuroendocrine tumors (NETs), neuroendocrine carcinomas (NECs), or mixed neuroendocrine-non-neuroendocrine neoplasms (MiNENs) based on histological differentiation, mitotic count, and the Ki-67 proliferation index. Well-differentiated NETs are further stratified into Grade1 (G1), Grade2 (G2), and Grade3 (G3) subtypes, whereas NECs with poorly differentiated NECs are classified as small cell NEC (SCNEC) and large cell NEC (LCNEC) . NECs are often non-functional, highly aggressive, and frequently diagnosed late with distant metastasis, resulting in dismal 5-year survival rates below 5% . Owing to the advanced stage and poor prognosis of GEP-NECs, treatments primarily aim to extend survival and enhance quality of life. Within the neuroendocrine spectrum, GEP-NECs share similar molecular and transcriptional profiles with small cell lung carcinoma (SCLC). Therefore, GEP-NECs treatment was similar to that for SCLC. Systemic chemotherapy, typically platinum-based regimens such as etoposide plus cisplatin (EC) and irinotecan plus cisplatin (IC), is the first-line of treatment for metastatic GEP-NECs. Notably, a Ki-67 index > 55% reliably predicts the reactivity of platinum-based chemotherapy . Although some studies have suggested efficacy for agents such as 5-fluorouracil, irinotecan, and oxaliplatin, there are still no standardized second-line regimens. Ongoing trials, such as the one using 5-fluorouracil + leucovorin + irinotecan offer promise (NCT03387592) – . Advances in molecular detection technology and basic experiments have led researchers to investigate the molecular pathways and interactions in neuroendocrine neoplasms, offering potential targets for GEP-NECs. For instance, sunitinib, a multitarget tyrosine kinase inhibitor primarily inhibiting vascular endothelial growth factor receptor and platelet-derived growth factor receptor, elicited positive responses in 11 of 20 patients with GEP-NECs in a phase II trial encompassing 6 NET G3 and 20 NEC cases . Phase II and III clinical trials testing single- or multi-drug combinations targeting various pathways, including XPO1 (NCT02250885), PARP (NCT04209595), and HDAC (NCT05076786), are currently ongoing for GEP-NECs. Among them, the negative regulator of the Notch pathway, DLL3, has drawn significant attention as a potential therapeutic target for neuroendocrine neoplasms , . The Notch pathway, known for its highly conserved signaling cascade, is a critical factor in cellular transformation, including cell proliferation, epithelial-mesenchymal transition, neuroendocrine cell differentiation, chemoresistance, and immune microenvironment modulation . As an inhibitory ligand of the Notch pathway , DLL3 interacts with various Notch receptors (Notch1-4) to facilitate malignant transformation. DLL3 expression is absent or minimal in normal cells and is predominantly localized within the Golgi apparatus and cytoplasmic vesicles. Conversely, in malignant NENs, DLL3 translocates to the cell membrane surface, inhibiting the CIS notch pathway , . ASCL1 , an intrinsic transcription factor in normal cells, regulates DLL3 expression, guiding neuroendocrine cell differentiation and initiating SCLC . ASCL1 activation increases DLL3 expression, enhancing the inhibition of Notch1 signaling . Both the inhibition of Notch1 and the upregulation of ASCL1 contribute to NEN development , . The efficacy of DLL3-targeted therapy has been confirmed in SCLC. Rova-T, a classic antibody-drug conjugate (ADC) in a phase I clinical trial (NCT01901653), has shown efficacy against recurrent or refractory SCLC . However, severe adverse reactions have halted further clinical investigations of Rova-T. Next-generation ADCs aim to enhance tumor cell uptake by refining drug linkers to mitigate toxicity and optimize pharmacokinetics for improved clinical utility , . Tarlatamab, a bispecific T cell engager (TCE) targeting DLL3 on tumor cells and CD3 on T cells, has demonstrated promising results in a phase I trial, with a 13% objective response rate (ORR) and 71% of patients with SCLC experiencing relief for ≥ 6 months . Other candidate drugs, such as HPN328 and AMG 119, have also shown beneficial antitumor responses in clinical and preclinical stages , . Patients with DLL3-expressing SCLC or NEC are currently enrolled in the first human Phase I trial of BI 764,532 (TCE) (NCT04429087). Although DLL3-targeted therapies are designed for tumors expressing DLL3, the immunohistochemical positivity for DLL3 expression on tumor cells was not a required criterion for patient enrollment in the trials. Therefore, DLL3 immunohistochemistry does not have a predictive role in determining eligibility for the trials. Although DLL3 expression was initially identified in SCLC, in vitro investigations have revealed its diverse oncogenic role; elevated expression promotes aggressive behavior through Snail overexpression , . Moreover, DLL3 is highly expressed not only in SCLC but also in lung LCNEC. In a retrospective study of pulmonary LCNEC, over 74% (70/94) of patients expressed DLL3 . DLL3 expression extends beyond lung cancer to various invasive malignancies, including prostate cancer, bladder SCNEC, malignant melanoma, glioblastoma, and medullary thyroid carcinoma – , suggesting its potential as a biomarker for neuroendocrine-origin malignancies. However, DLL3 expression in GEP-NECs is poorly understood. This study aimed to investigate DLL3 expression in GEP-NECs and analyze its clinicopathological correlation and the relationship between DLL3 expression and patient prognosis. Patient selection All available information in this retrospective study was sourced from the Peking University Cancer Hospital. Basic patient information spanning 2010 to 2023, including age, sex, primary tumor site, histological classification, TNM stage, chemotherapy prior to baseline, and some aspects of prognosis, were primarily collected from medical records. Telephone follow-up was specifically used to obtain detailed prognostic information, which may not have been fully documented in the medical records or required updates beyond what was originally recorded. TNM staging refer to the American Joint Committee on Cancer (AJCC) 8th edition .After re-examination by two experienced pathologists, it was ensured that each case met the WHO definition of neuroendocrine tumors: the shape was organ-like, trabecular and palisade-like, the mitotic count and Ki67 index strictly adhered to the standards of different grades, and expressed at least one neuroendocrine marker (Syn, CgA, CD56). Cases that did not meet these conditions were excluded from the study. Cases that lacked sufficient tissue samples for further staining were also excluded. The study protocol received approval from the Medical Ethics Committee of Peking University Cancer Hospital (approval number 2023KT29.), and all patients provided informed consent before tissue sample utilization. Pathological material The sample composition of this experiment is summarized in Supplementary Fig. 1. We obtained 248 tumor tissue samples from primary GEP-NECs (including surgery and biopsy samples), 19 distant metastatic GEP-NECs samples (all were liver metastases) and 9 lymph node metastases. Additionally, 36 GEP-NETs (8 G1, 9 G2, 19 G3) samples and 29 GACs samples were also collected. Each sample underwent formalin fixation and paraffin embedding (FFPE). Some cases had immunohistochemical staining results for three neuroendocrine markers, Ki67, and PD-L1, when included in the study. For cases lacking any markers, additional immunohistochemical staining was carried out on available tumor tissues. The assessment of neuroendocrine marker results follows the 5th edition WHO criteria , the Ki67 index was evaluated as a percentage, and the expression of PD-L1 was evaluated by CPS (22C3, Agilent DaKo, Denmark, 1:50). PD-L1 expression was defined as the ratio of the number of immune-related cells (tumor cells, lymphocytes, macrophages) expressing PD-L1 to the number of all tumor cells, with CPS ≥ 1 defined as positive. Immunohistochemical staining FFPE specimens were sectioned into 4-µm-thick slices using a rotary microtome, followed by incubation in EDTA solution (pH = 8.4) at 95 °C for 36 min. A DLL3 antibody (clone SP347, Roche, ready-to-use) was used for staining with Ventana, whereas ASCL1 antibody (24B72D11.1, BD Biosciences, 1:100) was stained with Leica Bond III. The color reaction was achieved using diaminobenzidine (ZSGB-BIO, Beijing, China). Staining outcomes were evaluated using an optical microscope with the 20× or 40× magnification. Each slide was reviewed by two experienced pathologists. Only when both pathologists agree on the result was it adopted, in cases of disagreement, the specimens were re-evaluated and discussed to achieve a consensus. DLL3 and ASCL1 results were binarily categorized. For DLL3, positive staining was defined as a reaction in ≥ 1% of tumor cells, regardless of intensity, based on criteria established by other studies – . Any punctate, cytoplasmic and/or membranous staining was considered positive, as previously described , . DLL3-high was characterized by ≥ 50% positive tumor cells and ≥ 75% positive tumor cells . For ASCL1, tumor cell positivity was categorized as 0, 1, 2, or 3 (negative, faint, moderate, or strong, respectively). We adopted semi-quantitative algorithms with H-score, which was calculated by multiplying the percentage of positive tumor cells by the corresponding staining intensity. An H-score ≥ 50 is defined as positive . Statistical analysis We conducted all analyses using SPSS software (IBM Corp., Armonk, NY, USA, version 25). The correlation between DLL3 expression and clinicopathological features was evaluated using the chi-squared test or Fisher’s exact test. Median progression-free survival (PFS) and overall survival (OS) were determined using Kaplan–Meier analysis, and survival differences were assessed using log-rank tests based on DLL3 expression status. The statistical significance level was set at p < 0.05. All available information in this retrospective study was sourced from the Peking University Cancer Hospital. Basic patient information spanning 2010 to 2023, including age, sex, primary tumor site, histological classification, TNM stage, chemotherapy prior to baseline, and some aspects of prognosis, were primarily collected from medical records. Telephone follow-up was specifically used to obtain detailed prognostic information, which may not have been fully documented in the medical records or required updates beyond what was originally recorded. TNM staging refer to the American Joint Committee on Cancer (AJCC) 8th edition .After re-examination by two experienced pathologists, it was ensured that each case met the WHO definition of neuroendocrine tumors: the shape was organ-like, trabecular and palisade-like, the mitotic count and Ki67 index strictly adhered to the standards of different grades, and expressed at least one neuroendocrine marker (Syn, CgA, CD56). Cases that did not meet these conditions were excluded from the study. Cases that lacked sufficient tissue samples for further staining were also excluded. The study protocol received approval from the Medical Ethics Committee of Peking University Cancer Hospital (approval number 2023KT29.), and all patients provided informed consent before tissue sample utilization. The sample composition of this experiment is summarized in Supplementary Fig. 1. We obtained 248 tumor tissue samples from primary GEP-NECs (including surgery and biopsy samples), 19 distant metastatic GEP-NECs samples (all were liver metastases) and 9 lymph node metastases. Additionally, 36 GEP-NETs (8 G1, 9 G2, 19 G3) samples and 29 GACs samples were also collected. Each sample underwent formalin fixation and paraffin embedding (FFPE). Some cases had immunohistochemical staining results for three neuroendocrine markers, Ki67, and PD-L1, when included in the study. For cases lacking any markers, additional immunohistochemical staining was carried out on available tumor tissues. The assessment of neuroendocrine marker results follows the 5th edition WHO criteria , the Ki67 index was evaluated as a percentage, and the expression of PD-L1 was evaluated by CPS (22C3, Agilent DaKo, Denmark, 1:50). PD-L1 expression was defined as the ratio of the number of immune-related cells (tumor cells, lymphocytes, macrophages) expressing PD-L1 to the number of all tumor cells, with CPS ≥ 1 defined as positive. FFPE specimens were sectioned into 4-µm-thick slices using a rotary microtome, followed by incubation in EDTA solution (pH = 8.4) at 95 °C for 36 min. A DLL3 antibody (clone SP347, Roche, ready-to-use) was used for staining with Ventana, whereas ASCL1 antibody (24B72D11.1, BD Biosciences, 1:100) was stained with Leica Bond III. The color reaction was achieved using diaminobenzidine (ZSGB-BIO, Beijing, China). Staining outcomes were evaluated using an optical microscope with the 20× or 40× magnification. Each slide was reviewed by two experienced pathologists. Only when both pathologists agree on the result was it adopted, in cases of disagreement, the specimens were re-evaluated and discussed to achieve a consensus. DLL3 and ASCL1 results were binarily categorized. For DLL3, positive staining was defined as a reaction in ≥ 1% of tumor cells, regardless of intensity, based on criteria established by other studies – . Any punctate, cytoplasmic and/or membranous staining was considered positive, as previously described , . DLL3-high was characterized by ≥ 50% positive tumor cells and ≥ 75% positive tumor cells . For ASCL1, tumor cell positivity was categorized as 0, 1, 2, or 3 (negative, faint, moderate, or strong, respectively). We adopted semi-quantitative algorithms with H-score, which was calculated by multiplying the percentage of positive tumor cells by the corresponding staining intensity. An H-score ≥ 50 is defined as positive . We conducted all analyses using SPSS software (IBM Corp., Armonk, NY, USA, version 25). The correlation between DLL3 expression and clinicopathological features was evaluated using the chi-squared test or Fisher’s exact test. Median progression-free survival (PFS) and overall survival (OS) were determined using Kaplan–Meier analysis, and survival differences were assessed using log-rank tests based on DLL3 expression status. The statistical significance level was set at p < 0.05. Patient profile Table provides a summary of the baseline characteristics of patients diagnosed with GEP-NECs. Among the 248 primary GEP-NECs cases, the age range was 23 to 85 years, with a median of 62 years. The study included 182 (73.4%) males and 66 (26.6%) females. The total number of SCNEC, LCNEC, and MiNENs cases were 138 (55.7%), 101 (40.7%), and 9 (3.6%), respectively. Gastric NECs accounted for the largest proportion (148, 59.7%) in this cohort, followed by the esophagus (51, 20.6%), and pancreas (24, 9.7%). T1-stage patients accounted for 3.2% (8/248), while T2-T4-stage patients accounted for 43.1% (107/248). Patients with stage N0 and N1-N3 accounted for 11.7% (29/248) and 33.1% (82/248), respectively; 40.7% (101/248) of patients with M stage were M0 stage, and 5.2% (13/248) were M1 stage. Additionally, 64 (25.8%) patients received chemotherapy before baseline. The rates of Ki67 proliferation index in different intervals 25–50%, 51–75%, and 76–100% were 11.7% (29/248), 39.5% (98/248), and 35.1% (87/248), respectively. In 75.4% (187/248) of cases, at least 2 NE markers were positive, while in 12.5% (31/248) of cases < 2 NE markers were positive. A positive reaction of PD-L1 was shown in 6.0% (15/248) of cases. Relationship between DLL3 expression and clinicopathological features DLL3 staining was positive in the cytomembrane, cytoplasm, and punctate, in which cytoplasmic and membranous staining was diffuse, whereas intermittent perinuclear staining was punctate. However, we identified four cases in which nuclear positivity was observed alongside cytoplasmic and membranous staining, a pattern not previously reported in other studies. The nuclear positivity in these cases may represent a nonspecific or incidental reaction. Representative images are presented in Fig. a. Moreover, DLL3 has a heterogeneous expression pattern in GEP-NECs. Correlations between DLL3 expression and clinicopathological features are presented in Table . Positive expression was observed in 68.1% of SCNEC (94/138) compared to 38.6% of LCNEC (39/101) and 33.3% (3/9) of MiNENs ( p < 0.001). Of note, in MiNENs, the neuroendocrine component exhibited positive DLL3 expression, while the adenocarcinoma component showed negative expression (Fig. b). The expression rate of DLL3 was highest in the esophagus (68.6%, 35/51), followed by the colorectum (53.8%, 7/13), stomach (52.7%, 78/148), pancreas (41.7%, 10/24), and small intestine (25.0%, 2/8). DLL3 showed significant expression differences among different T-stage cases (T1 100.0% vs. T2-T4 53.3%, p = 0.009). Patients who underwent chemotherapy prior to baseline demonstrated a higher prevalence of DLL3 expression compared to individuals who did not undergo previous therapy (67.2% vs. 50.5%, p = 0.015). In the group with < 2 NE positive markers, the expression rate of DLL3 was 38.7%, significantly lower than that in the group with ≥ 2 NE positive markers (56.7%, p = 0.048). No associations were observed between DLL3 expression and sex, N stage, M stage, Ki67 index, or PD-L1 expression. DLL3 expression in metastatic tumors The expression of DLL3 in metastatic tumors is shown in Supplementary Fig. 2. The staining pattern of DLL3 in metastatic lesions was similar to that in primary tumors. In lymph node metastasis and distant metastases, the positive rate of DLL3 was 44.4% and 52.6%, respectively. Chi-square analysis showed no significant difference in the expression of DLL3 between primary tumor, lymph node metastases, and distant metastases ( p = 0.818) (Fig. a). Explore the differential diagnosis value of DLL3 To explore the expression of DLL3 in well-differentiated GEP-NETs, we stained 36 cases of GEP-NETs. Among the GEP-NETs cases, none of the eight cases of NET G1 and nine cases of NET G2 exhibited positive DLL3 staining, while 3 out of 19 (15.8%) cases of NET G3 were positive for DLL3 staining. The DLL3 expression was also heterogeneous, and all of them were positive in the cytoplasm and membrane (Fig. c). Since the morphological overlap between NET G3 and NECs or poorly differentiated adenocarcinoma and NECs often puts pathologists in a dilemma with respect to differential diagnosis, we further assessed the value of DLL3 in differential diagnosis. Therefore, we performed DLL3 staining on additional GACs with different degrees of differentiation. The results show that no positive DLL3 staining was detected in 29 GACs cases, regardless of the degree of differentiation (Fig. c). DLL3 may serve as a useful differential diagnostic tool with a sensitivity of 54.8% and a specificity of 84.2% when differentiating NECs from NET G3, and a sensitivity of 54.8% and a specificity of 100.0% when differentiating NECs from GACs, with a cutoff value of 1% positive tumor cells. However, with the threshold raised to 50% positive tumor cells, the sensitivity declined to 31.9%, while the specificity increased to 94.7% when identifying NECs and NET G3 and to 100.0% when identifying NECs and GACs, reflecting the stricter standard (Table ). ASCL1 expression Almost all positive reactions for ASCL1 were located to the nucleus, with only one case showing an unusual punctate positivity around the nucleus (Fig. a). The expression status of 111 samples of ASCL1 and DLL3 is shown in Fig. b. In addition, it was observed that in samples with co-expression of DLL3 and ASCL1, tumor cells expressing ASCL1 also exhibited DLL3 expression, demonstrating a spatial consistency. (Fig. c). Similarly, its expression was also heterogeneous inside the tumor. Overall, ASCL1 expression was detected in 14.4% (16/111) of GEP-NECs cases. Interestingly, DLL3 expression rate differed between the ASCL1 positive and ASCL1 negative groups ( p = 0.002) (Fig. b). In the ASCL1 positive group, the DLL3 positive rate was 87.5% (14/16), whereas 47.4% (45/95) showed positive DLL3 staining in the ASCL1 negative group. Similarly, DLL3 expression was higher in the ASCL1 positive group among GEP-SCNECs (100.0% vs. 56.8%, p = 0.02) (Fig. c). However, the differential expression of DLL3 was not significant in GEP-LCNECs (33.3% vs. 37.8%, p = 0.687) (Fig. d). Follow-up and prognosis The prognosis information of 199 patients was obtained. The median follow-up time was 19.7 months (range: 0.8–140.4 months). Survival analysis was conducted for different cutoff values of DLL3 expression ( N = 199) (Fig. ). In the DLL3 positive group (with the 1% cutoff value), the median PFS and median OS were 12.3 months (95% CI: 7.9–16.6) and 24.4 months (95% CI: 20.2–28.6), respectively, while in the DLL3 negative group, the median PFS and median OS were 13.4 months (95% CI: 10.5–16.3) and 25.9 months (95% CI: 18.9–32.9), respectively. The log-rank test showed that there was no significant difference in the prognosis between the DLL3 negative group and the DLL3 positive group (Fig. a). Moreover, no significant difference in survival was observed between high and low DLL3 expression groups, regardless of whether the cutoff value defined by high DLL3 expression was 50% or 75% (Fig. b and c). Table provides a summary of the baseline characteristics of patients diagnosed with GEP-NECs. Among the 248 primary GEP-NECs cases, the age range was 23 to 85 years, with a median of 62 years. The study included 182 (73.4%) males and 66 (26.6%) females. The total number of SCNEC, LCNEC, and MiNENs cases were 138 (55.7%), 101 (40.7%), and 9 (3.6%), respectively. Gastric NECs accounted for the largest proportion (148, 59.7%) in this cohort, followed by the esophagus (51, 20.6%), and pancreas (24, 9.7%). T1-stage patients accounted for 3.2% (8/248), while T2-T4-stage patients accounted for 43.1% (107/248). Patients with stage N0 and N1-N3 accounted for 11.7% (29/248) and 33.1% (82/248), respectively; 40.7% (101/248) of patients with M stage were M0 stage, and 5.2% (13/248) were M1 stage. Additionally, 64 (25.8%) patients received chemotherapy before baseline. The rates of Ki67 proliferation index in different intervals 25–50%, 51–75%, and 76–100% were 11.7% (29/248), 39.5% (98/248), and 35.1% (87/248), respectively. In 75.4% (187/248) of cases, at least 2 NE markers were positive, while in 12.5% (31/248) of cases < 2 NE markers were positive. A positive reaction of PD-L1 was shown in 6.0% (15/248) of cases. DLL3 staining was positive in the cytomembrane, cytoplasm, and punctate, in which cytoplasmic and membranous staining was diffuse, whereas intermittent perinuclear staining was punctate. However, we identified four cases in which nuclear positivity was observed alongside cytoplasmic and membranous staining, a pattern not previously reported in other studies. The nuclear positivity in these cases may represent a nonspecific or incidental reaction. Representative images are presented in Fig. a. Moreover, DLL3 has a heterogeneous expression pattern in GEP-NECs. Correlations between DLL3 expression and clinicopathological features are presented in Table . Positive expression was observed in 68.1% of SCNEC (94/138) compared to 38.6% of LCNEC (39/101) and 33.3% (3/9) of MiNENs ( p < 0.001). Of note, in MiNENs, the neuroendocrine component exhibited positive DLL3 expression, while the adenocarcinoma component showed negative expression (Fig. b). The expression rate of DLL3 was highest in the esophagus (68.6%, 35/51), followed by the colorectum (53.8%, 7/13), stomach (52.7%, 78/148), pancreas (41.7%, 10/24), and small intestine (25.0%, 2/8). DLL3 showed significant expression differences among different T-stage cases (T1 100.0% vs. T2-T4 53.3%, p = 0.009). Patients who underwent chemotherapy prior to baseline demonstrated a higher prevalence of DLL3 expression compared to individuals who did not undergo previous therapy (67.2% vs. 50.5%, p = 0.015). In the group with < 2 NE positive markers, the expression rate of DLL3 was 38.7%, significantly lower than that in the group with ≥ 2 NE positive markers (56.7%, p = 0.048). No associations were observed between DLL3 expression and sex, N stage, M stage, Ki67 index, or PD-L1 expression. The expression of DLL3 in metastatic tumors is shown in Supplementary Fig. 2. The staining pattern of DLL3 in metastatic lesions was similar to that in primary tumors. In lymph node metastasis and distant metastases, the positive rate of DLL3 was 44.4% and 52.6%, respectively. Chi-square analysis showed no significant difference in the expression of DLL3 between primary tumor, lymph node metastases, and distant metastases ( p = 0.818) (Fig. a). To explore the expression of DLL3 in well-differentiated GEP-NETs, we stained 36 cases of GEP-NETs. Among the GEP-NETs cases, none of the eight cases of NET G1 and nine cases of NET G2 exhibited positive DLL3 staining, while 3 out of 19 (15.8%) cases of NET G3 were positive for DLL3 staining. The DLL3 expression was also heterogeneous, and all of them were positive in the cytoplasm and membrane (Fig. c). Since the morphological overlap between NET G3 and NECs or poorly differentiated adenocarcinoma and NECs often puts pathologists in a dilemma with respect to differential diagnosis, we further assessed the value of DLL3 in differential diagnosis. Therefore, we performed DLL3 staining on additional GACs with different degrees of differentiation. The results show that no positive DLL3 staining was detected in 29 GACs cases, regardless of the degree of differentiation (Fig. c). DLL3 may serve as a useful differential diagnostic tool with a sensitivity of 54.8% and a specificity of 84.2% when differentiating NECs from NET G3, and a sensitivity of 54.8% and a specificity of 100.0% when differentiating NECs from GACs, with a cutoff value of 1% positive tumor cells. However, with the threshold raised to 50% positive tumor cells, the sensitivity declined to 31.9%, while the specificity increased to 94.7% when identifying NECs and NET G3 and to 100.0% when identifying NECs and GACs, reflecting the stricter standard (Table ). Almost all positive reactions for ASCL1 were located to the nucleus, with only one case showing an unusual punctate positivity around the nucleus (Fig. a). The expression status of 111 samples of ASCL1 and DLL3 is shown in Fig. b. In addition, it was observed that in samples with co-expression of DLL3 and ASCL1, tumor cells expressing ASCL1 also exhibited DLL3 expression, demonstrating a spatial consistency. (Fig. c). Similarly, its expression was also heterogeneous inside the tumor. Overall, ASCL1 expression was detected in 14.4% (16/111) of GEP-NECs cases. Interestingly, DLL3 expression rate differed between the ASCL1 positive and ASCL1 negative groups ( p = 0.002) (Fig. b). In the ASCL1 positive group, the DLL3 positive rate was 87.5% (14/16), whereas 47.4% (45/95) showed positive DLL3 staining in the ASCL1 negative group. Similarly, DLL3 expression was higher in the ASCL1 positive group among GEP-SCNECs (100.0% vs. 56.8%, p = 0.02) (Fig. c). However, the differential expression of DLL3 was not significant in GEP-LCNECs (33.3% vs. 37.8%, p = 0.687) (Fig. d). The prognosis information of 199 patients was obtained. The median follow-up time was 19.7 months (range: 0.8–140.4 months). Survival analysis was conducted for different cutoff values of DLL3 expression ( N = 199) (Fig. ). In the DLL3 positive group (with the 1% cutoff value), the median PFS and median OS were 12.3 months (95% CI: 7.9–16.6) and 24.4 months (95% CI: 20.2–28.6), respectively, while in the DLL3 negative group, the median PFS and median OS were 13.4 months (95% CI: 10.5–16.3) and 25.9 months (95% CI: 18.9–32.9), respectively. The log-rank test showed that there was no significant difference in the prognosis between the DLL3 negative group and the DLL3 positive group (Fig. a). Moreover, no significant difference in survival was observed between high and low DLL3 expression groups, regardless of whether the cutoff value defined by high DLL3 expression was 50% or 75% (Fig. b and c). High levels of DLL3 expression are found in carcinoid subtypes of pulmonary NETs that exhibit significant dendritic cell infiltration, suggesting its potential as a viable target for focused intervention. Previous studies have reported a 65% DLL3 expression rate in 37 patients with LCNEC . Analysis of clinical trial groups revealed DLL3 expression in over 70% of SCLC cases , . A high expression rate of DLL3 was also observed in patients with castration-resistant neuroendocrine prostate cancer . A previous study reported a DLL3 expression rate of 76.9% in 13 cases of gastrointestinal-pancreatic NECs . We first assessed DLL3 positivity in a large sample of GEP-NECs ( n = 248) and found a positivity rate of 54.8%. Our study revealed different DLL3 expression rates between SCNEC and LCNEC, with SCNEC exhibiting a higher positivity rate than LCNEC. Regarding primary site distribution, the esophagus exhibited a high DLL3 expression rate, which was largely attributed to the high proportion of SCNEC in esophageal NECs. Similarly, this cohort showed a high expression rate of DLL3 in T1 stage patients, partly because SCNEC had an absolute advantage in the number of T1-stage patients (7/8). A single-cell transcriptome sequencing study of SCLC tumor cells demonstrated that chemotherapy resulted in decreased expression of therapeutic target genes, including DLL3 . However, our results showed that DLL3 has a higher expression rate in patients receiving chemotherapy before baseline treatment, which prompts future research to focus on whether chemotherapy affects the expression of the DLL3 gene. Several studies have adopted various definitions for DLL3-high expression. When defined as staining in ≥ 50% of tumor cells, the DLL3-high expression rates in SCLC range from 32% to 79.5% , , . With a cutoff value of 75%, the positive rate of high expression was 70% in another study on SCLC . Conversely, high expression rates of DLL3 in lung LCNEC are relatively low at 54% . In this study, we applied 50% and 75% cutoff values to define DLL3-high expression in GEP-NECs, exploring the DLL3-high expression rate for the first time. Although these cutoff values have been frequently used in small cell lung cancer (SCLC), their application to GEP-NECs has not been previously reported. Our results showed a 31.8% expression rate of DLL3 with a cutoff value of 50% and a 25.4% expression rate of DLL3 with a cutoff value of 75%. Since GEP-NECs contains SCNEC and LCNEC, we analyzed the high expression rates in different subtypes (Supplementary Table 1). Irrespective of the cutoff value, the expression rate of DLL3 in SCNEC is always higher than that in LCNEC. Furthermore, the expression rate of DLL3 in the digestive system was lower than that in high-grade neuroendocrine tumors of the lung, regardless of the cutoff value. Recent studies in SCLC indicate significant improvements in ORR and PFS in patients with high expression of DLL3 (defined as positive staining in ≥ 50% of tumor cells); the high expression group demonstrated significant benefits in confirmed objective response (35% versus 0%) and disease control (90% versus 60%) compared to the low expression group of DLL3 . Additionally, in another phase II clinical trial, the ORR was 14.3% in the DLL3 high-expression group, with high expression defined as positive staining in ≥ 75% of tumor cells . In phase I/II studies of multiple cancer types, a higher remission rate was observed with high DLL3 expression . However, some patients with DLL3 expression in < 50% of tumor cells also exhibit stable disease for a prolonged duration , . Hence, in the following clinical phases, the assessment should focus on evaluating the efficacy in patients exhibiting any measurable DLL3 levels. The field of immuno-oncology (IO) has revolutionized the landscape of cancer therapy, actively influencing patient outcomes . PD-L1 has been identified as the first biomarker for anti-PD-1 therapy and is included in the prescribing information for pembrolizumab. Currently, other immune therapy efficacy-related markers include tumor mutational burden (TMB), mismatch repair system deficiency (dMMR), high microsatellite instability (MSI-H), neoantigens, mutations in antigen presentation pathways, and circulating tumor DNA (ctDNA) as indicators for selecting patients who may benefit from immunotherapy . In a study on pancreatic ductal adenocarcinoma, high DLL3 expression was positively correlated with PD-L1/2 expression . Another study in SCLC also observed cases with high DLL3 expression and negative NOTCH1 expression to have a higher PD-L1 expression rate, revealing a favorable prognosis in such SCLC patients . However, our study did not find any correlation between DLL3 and PD-L1 expression in GEP-NECs. Although no significant correlation was observed, in our study cohort, 9 out of 40 (22.5%) patients co-expressed DLL3 and PD-L1, suggesting the feasibility of combined DLL3-targeted therapy and immunotherapy in some patients and providing a theoretical basis for developing new treatment strategies for patients with GEP-NEC, which requires further clinical validation. In recent studies, four subtypes of SCLC have been identified based on the unique expression patterns of four key transcripts: SCLC-A, SCLC-N, SCLC-P, and SCLC-Y . Baine et al. confirmed higher DLL3 expression in ASCL1-high SCLC compared to other groups . Wang et al. classified NECs into five subtypes (ASCL1, NEUROD1, HNF4A, POU2F3, and YAP1) across various tumor sites (including GEP) . Although our study did not further classify subtypes, we first verified the expression of ASCL1 and DLL3 in 111 available tumor tissues, showing a significantly higher DLL3 expression rate in the ASCL1-positive group (likely corresponding to the A subtype) compared to the ASCL1-negative group (likely corresponding to non-A subtypes) at the protein level, confirming a significant correlation. However, this correlation was only evident in small cell GEP-NECs and was not seen in large cell or mixed types, which was quite different from that observed with pulmonary neuroendocrine tumors. The correlation of DLL3 and ASCL1 expression was observed in both SCLC and LCNEC of the lung , , . Nonetheless, these results suggest that patients with the ASCL1 subtype may benefit from DLL3-targeted therapy. Furthermore, the study by Wang et al. revealed that A and N types of GEP-NEC are classified as NE-high types . Similarly, in our cohort, we found that among the 13 ASCL1-positive cases that underwent NE staining in available tissues, 10 cases were positive for ≥ 2 NE markers. Additionally, in tumors positive for ≥ 2 NE markers, the expression rate of DLL3 was higher than that of the subgroup of < 2 NE markers, which is consistent with the results observed in pulmonary neuroendocrine carcinomas . It has been found that a classic pathway in the development of SCLC involves tumor initiation under the premise of bi-allelic mutations in RB1 and TP53 , driven by NOTCH signaling inactivation . While the roles of the DLL3/Notch pathway and the loss of RB1 and TP53 mutations in the development of GEP-NECs have not been extensively studied, in high-grade neuroendocrine tumors of the digestive system, TP53 and RB mutations are the most common genetic alterations, differing from the frequently occurring MEN1 , ATRX/DAXX , and mTOR pathway activation in NET . This may explain why DLL3 is negative in well-differentiated neuroendocrine tumors, but shows positive reactions in some high-grade neuroendocrine tumors, suggesting that part of digestive neuroendocrine tumors may share a similar carcinogenic pathway with SCLC. However, this speculation needs further confirmation in future studies. DLL3 staining in all GACs cases showed negative results regardless of the degree of differentiation. Among the nine cases of MiNENs comprising a mixture of adenocarcinoma and NECs, no positive signal was detected in the adenocarcinoma component. Hence, DLL3 staining may serve as a tool to differentiate gastric NECs from GACs with 100% specificity, particularly in biopsy samples where limited tissue samples may hinder a definitive diagnosis. Based on the classification established by the 2010 World Health Organization, NENs were divided into highly differentiated NETs (NET G1 and NET G2) and poorly differentiated NECs . Despite being highly proliferative, certain tumors progress slowly and have a favorable prognosis, while others follow an aggressive course. To address this heterogeneity, a subgroup termed NET G3 was introduced into the GEP-NENs classification system, representing an intermediate prognosis between NET G1/G2 and NEC . Nevertheless, distinguishing NET G3 with a relatively favorable prognosis from poorly differentiated NECs remains challenging for pathologists. Currently, a comprehensive diagnostic approach combining clinical performance, morphology, immunohistochemistry, and molecular biomarkers is preferred. These biomarkers include RB, DAXX/ATRX, SSTR2, CgA, and p53 . No large-sample studies have confirmed the value of DLL3 in distinguishing NET G3 from NEC. A small-sample study on GEP-NENs found no DLL3 expression in NET G3 (0/5) . Chen et al. utilized various methods, including WES, FISH, qPCR, and IHC, to identify DLL3 expression in three out of eight GEP-NET G3 cases, with no abnormal DLL3 expression in NET G2 . Our study marks the first discovery of DLL3 protein expression in GEP-NET G3 with a relatively large sample size, suggesting the differential diagnostic value of DLL3,and calling for further research on DLL3-targeting agents in NET G3. Additionally, there was no significant difference in DLL3 expression among primary tumors, lymph node metastases, and distant metastases. Therefore, DLL3 staining in metastases appears to reflect the condition of primary tumors, suggesting the feasibility of applying DLL3-targeted therapy in advanced patients with distant organ metastases. In a study on DLL3 expression in SCLC without targeted therapy, patient prognosis was found to be unrelated to DLL3 expression status, with high or low expression (≥ 50% cutoff) showing no impact on survival . Similar findings were observed in the GEP-NECs cohort in our study. However, another study encompassing SCLC, carcinoid, and atypical carcinoid cases noted that DLL3-high expression (≥ 50% positive tumor cells) correlated with improved OS in SCLC ( p = 0.049), without adjusting for age, tumor dimension, and stage . Conversely, a small-sample cohort study of GEP-NENs ( n = 46) observed significantly better PFS and OS in the DLL3-negative group, largely because of the predominance of NET G1/G2/G3 cases in their study cohort . This study was the largest GEP-NECs cohort to date and found that the expression of DLL3 has no relationship with the prognosis of patients with GEP-NECs. This study had some limitations. Being a single-center study with a modest sample size, future research should involve larger cohorts to validate the differences in DLL3 expression across primary tumor sites and further assess the diagnostic performance between NET G3 and NECs in the digestive system. Although cytoplasmic and/or membranous staining was considered positive, as reported by other studies , , the presence of nuclear positivity remains poorly understood. This discrepancy highlights a limitation in our study, as the relationship between DLL3 staining type (cytoplasmic, membranous, punctata and nuclear) and therapeutic response to DLL3-targeted therapy remains unclear. Further investigation is required to elucidate whether different DLL3 staining patterns are associated with varying clinical outcomes. Moreover, the study’s retrospective nature leads to potential biases stemming from progression in clinical practices over time. This highlights the necessity for prospective studies to more effectively clarify the influence of DLL3 expression on patient prognosis. Our study confirmed DLL3 expression in GEP-NECs and found that DLL3 expression was related to the SCNEC subtype and chemotherapy. The DLL3 expression rate in NET G3 supports the application of DLL3-targeted therapy in high-grade NETs of the digestive system. DLL3 IHC was conducive for distinguishing between NET G3 and NECs, as well as GACs and NECs. Finally, we confirmed the correlation between DLL3 and ASCL1 protein expression in GEP-NECs. Our study suggests that DLL3 expression is not a prognostic factor in patients with GEP-NECs. Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 |
Regional anesthesia for pediatric cardiac surgery: a review | 04fa96cd-94bb-4079-9065-4d8d292b3c20 | 11829357 | Surgical Procedures, Operative[mh] | Postoperative pain management is a challenge in pediatric cardiac surgery as inadequate pain control is associated with hemodynamic instability, postoperative respiratory complications, prolonged intubation, and prolonged intensive care unit (ICU) and hospital stay . Considering that the main sources of pain after cardiac surgery include sternotomy or thoracotomy incisions , efforts are being made to develop more refined and less invasive techniques, with the inclusion of minithoracotomy, ministernotomy and thoracoscopy accesses . Several regional anesthesia (RA) techniques have been proposed for pediatric patients undergoing cardiac surgery, ranging from neuraxial approaches to fascial plane blocks . The Enhanced Recovery After Surgery (ERAS) guidelines emphasize the importance of effective perioperative pain management in improving patient outcomes. Within this framework, RA techniques are recommended for their opioid-sparing effects in pediatric cardiac surgery . However, the rapid development of these techniques, especially fascial plane blocks, leave room for several uncertainties especially related to the mechanism of action, appropriate type, volume, concentration and pharmacokinetics of local anesthetics (LA). Moreover, most of the studies have been performed on adult patients and it is debatable if they could be transferred to the pediatric population . Therefore, application of RA techniques in pediatric cardiac surgery is complex and continuously evolving; the aim of this review is to summarize the existing knowledge regarding RA for pediatric patients undergoing cardiac surgery outlining their application methods and anatomical underpinnings, supported by illustrative clinical examples to highlight their practical implementation and effectiveness. Clinical outcomes RA provides a targeted approach for pain control by interrupting the transmission of pain signals at the site of injury, resulting in more efficient and prolonged pain relief than conventional IV analgesia. In 2023, network meta-analysis of over 5,000 adult patients undergoing cardiac surgery revealed that RA techniques provided superior pain control when compared to conventional IV analgesia , similar results were reported for pediatric patients, even if the total population of the meta-analysis comprised only 605 children from 14 RCTs . Targeted approaches of regional techniques facilitate effective pain management and reduce the need for systemic opioids, thereby decreasing the associated side effects . Ineffective postoperative pain management leads to chronic pain, immunosuppression, infections, and impaired wound healing . Moreover, uncontrolled pain in children can lead to both immediate and long-term consequences, including heightened anxiety and potential behavioral disturbances . Traditionally, IV analgesia is used as the primary method for pain control; however, it is associated with adverse effects that can prolong the ICU and hospital stay of the patients . RA is increasingly used for pediatric cardiac surgery in hopes it will enhance the clinical outcomes and reduce complications . Neuraxial techniques, such as the epidural technique, suppress the inflammatory response and sympathetic activity, which is crucial for patients as extracorporeal circulation, hypothermia, and surgical stress exacerbate the inflammatory response . RA reduces opioid consumption and pain score via effective postoperative pain relief , thereby decreasing the incidence of opioid-induced respiratory depression and other adverse events, such as gastrointestinal disturbances, pruritus, opioid hypersensitivity, and oversedation, which complicate postoperative treatment and prolong recovery . Prolonged ICU and hospital stay poses a significant burden on patients and their families, especially on pediatric patients as being away from their parents significantly influences their motivation and cooperation with subsequent treatment . Additionally, pediatric patients exhibit the most intense negative psychological and behavioral outcomes following critical illness and prolonged ICU stay . Prolonged stay also imposes a significant financial strain on healthcare systems due to the need for intensive care and related treatments. Effective pain management alleviates these social and economic burdens while improving patient outcomes . Pain assessment In pediatric cardiac surgery, assessing pain during intensive care follow-up is challenging due to the difficulty of using self-reported tests; therefore, behavioral observation becomes critical. Tools, such as the Face, Legs, Activity, Cry, and Consolability scale and Modified Objective Pain Scale, rely on observing physical signs of distress, such as facial expressions, body movements, crying, and the ability to be comforted . These tools are especially useful for infants and younger children who cannot verbalize their pain and older children with severe illnesses, which limit self-reporting. However, interpreting child behavior is another challenge. Pain behaviors of children are impacted by fear, anxiety, fatigue, and prior experience, complicating their assessment. Moreover, personal bias and cultural background of the observer also influences the interpretation of patient behaviors . Therefore, physiological measures provide another avenue for pain assessment, relying on objective indicators, such as changes in heart rate, blood pressure, and respiratory rate . However, physiological responses are not specific to pain and can be triggered by other factors, such as stress and illness. Therefore, physiological data combined with self-reports and behavioral observations should be used for accurate assessment . Central neuraxial techniques Central neuraxial techniques, including spinal , caudal , and epidural techniques, are used for analgesia during pediatric cardiac surgery. These techniques have demonstrated significant benefits in perioperative pain control. However, their application can cause potential complications, such as epidural hematoma, especially in patients receiving anticoagulation therapy, and unpredictable spread of LAs . Epidural analgesia (EA) remains a cornerstone technique in this context involving the injection of LAs, often combined with opioids, into the epidural space. Beyond its potent analgesic properties, EA also modulates the stress response via sympathetic blockade, stabilizing the hemodynamics during surgery by reducing catecholamine release. This is especially beneficial for pediatric patients, as controlling the body response to surgical stress is critical for maintaining cardiovascular stability in these patients . Paravertebral Block Paravertebral Block (PVB) involves injecting the LAs near the spinal nerves as they exit the intervertebral foramina, targeting the paravertebral space. Ultrasound guidance reduces the risk of complications, such as pleural puncture. Unlike those of EA, effects of PVB are more localized, resulting in fewer systemic effects, especially on hemodynamic stability. Therefore, PVB is a safe option for children with cardiovascular compromise . Fascial Plane Blocks Erector spinae plane (ESP) block ESP block, first described by Forero et al. , is an emerging fascial plane block in which LAs are injected into the fascial plane between the erector spinae muscles and thoracic transverse processes (Fig. ). Although the specific mechanism of ESP block is still under debate , the most likely mechanism is that the administered LA spreads in the craniocaudal direction and into the paravertebral area, blocking the dorsal and ventral rami of the spinal nerves and sympathetic ganglia, facilitating both somatic and visceral sensory blockade. When applied bilaterally, ESP block effectively provides analgesia across specific dermatomes, particularly in the T2–7 regions, thereby covering the entire thorax, however the debase is still ongoing as some cadaveric studies suggests the staining of ventral rami could not be constant . Many studies have demonstrated the efficiency of the ESP block for pediatric cardiac surgery. In an RCT by Gado et al., postoperative opioid consumption and pain scores in the first 24 h were significantly lower in pediatric patients with ESP block than in those without ESP block . Another study reported that bilateral ESP block effectively relieves sternotomy pain, requiring less rescue analgesia and exerting long-lasting effects in the postoperative period . Gustavo et al. compared an ESP block performed at the T5 level with traditional IV opioids and reported that the ESP block was associated with a shorter ICU stay, lower opioid consumption, and faster discharge from the ICU . A systematic review and meta-analysis of five studies, including 384 pediatric cardiac patients—178 of whom received an ESP block—demonstrated notable benefits compared to IV opioid-based analgesia. The analysis revealed that the ESP block significantly reduced intraoperative fentanyl consumption and ICU length of stay in children undergoing cardiac surgery via midline sternotomy. This meta-analysis highlights the potential of the ESP block as a valuable adjunct to existing multimodal analgesia protocols . Mid-transverse process to pleura block (MTPB) and retrolaminar block (RLB) MTPB and RLB are two variants with a supposed mechanism of action similar to the ESP block (Fig. ). They work as the LA spreads through the paravertebral space, simultaneously blocking the spinal nerve roots of the ventral and dorsal rami, as the intercostal nerve is blocked at a level not far from the spinal cord . A cadaveric study demonstrated the extensive spread of MTPB into the paravertebral space, ESP, dorsal and ventral rami of the spinal nerves, sympathetic chain, and intercostal nerves . The injection was targeted midway between the transverse process of the thoracic vertebra and pleura, posterior to the superior costotransverse ligament, causing pleural displacement and spreading into the ESP. Abdelbaser et al. reported that MTPB significantly reduces intraoperative fentanyl consumption by blunting the stress response to various surgical stimuli during pediatric cardiac surgery. Moreover, MTPB significantly reduces the extubation time and duration of ICU stay, thereby decreasing the pain score. Another RCT revealed no significant differences in intraoperative hemodynamics, intraoperative fentanyl consumption, 24-h postoperative fentanyl consumption, postoperative pain score, extubation time, and ICU discharge time between the PVB and MTPB groups, with significantly shorter time required to perform bilateral MTPB than for PVB . RLB is another fascial plane block that involves LA injection between the posterior surface of the thoracic vertebral lamina and overlying paraspinal muscles . ESP block targets the tips of transverse processes, whereas RLB targets the laminae . Spread of LA into the paravertebral and epidural spaces blocks the ventral and dorsal rami of the thoracic spinal nerves and extends laterally into the fascial plane to block the lateral cutaneous branches of intercostal nerves . The potential superiority of RLB over ESP block is probably due to the deeper deposition that may facilitate more consistent diffusion of the LA toward the nerve roots and paravertebral spaces, areas critical for achieving effective somatic and visceral analgesia. A recent network meta-analysis which assessed the RA techniques in pediatric cardiothoracic surgery reported that the largest opioid consumption decrease was in RLB . Ultrasound-guided RLB is theoretically safer than thoracic EA and PVB, as the block needle is inserted away from the pleura and dura . Moreover, no major vessels or nerves exist along the needle path, with only a slight risk of intramuscular hemorrhage . RLB is associated with early extubation and short ICU stay due to the reduction in perioperative opioid consumption after open cardiac surgery via median sternotomy . Interpectoral plane (IPP) and IPP + pectoserratus plane (PSP) blocks IPP (previously known as PECS-I) block was first described by Blanco in 2011 while IPP + PSP (previously known as PECS-II) block was introduced the following year . IPP block involves a single injection of the LA between the pectoralis major and minor muscles, whereas the IPP + PSP block involves an additional injection into the plane between the pectoralis minor and serratus anterior muscles (SAMs) (Fig. ). IPP block targets the pectoral nerves and possibly the intercostal nerves depending on the injection site, whereas the IPP + PSP block targets the long thoracic, thoracodorsal, and lateral branches of the intercostal nerves. Although the IPP block is widely used in other context such as breast surgery, its use for cardiac surgery, particularly in pediatric populations, is still in its early stages. To date, most studies have focused on adult populations, leading to insufficient data on pediatric patients, for this reason authors would recommend caution to readers while implementing these techniques in their clinical practice until further evidence is provided by future studies on the topic. Zachary et al. retrospectively evaluated the use of IPP and IPP + PSP blocks for postoperative pain management following sternotomy in 73 pediatric patients. IPP was performed in 47 patients and IPP + PSP in 26 patients. Notably, 34% of patients did not experience severe pain within the first 24 h post-surgery . IPP + PSP block decreases the pain score, opioid consumption, and agitation score . In a retrospective study conducted by Yang et al. in pediatric patients undergoing pacemaker or defibrillator implantation, IPP + PSP block yielded positive outcomes by decreasing the pain score and opioid requirement . IPP and IPP + PSP blocks are more suitable for minimally invasive surgeries, such as pacemaker and defibrillator placement, than anterior mediastinotomy. Serratus anterior plane (SAP) block SAP block is performed more laterally and posteriorly than the interpectoral block in the axillary region at the level of the fourth or fifth rib. In this block, LA is injected either between the serratus anterior and latissimus dorsi muscles or beneath the SAMs (Fig. ). Depending on whether the injection is deep or superficial, different nerve branches are affected: LA injected deep into the SAM blocks the lateral cutaneous branches of the intercostal nerves, whereas a superficial injection blocks the long thoracic and thoracodorsal nerves and lateral cutaneous branch of the intercostal nerve . SAP block spreads across the T2–9 levels, covering the anterior, lateral, and posterior chest walls. Specifically, SAP block provides effective and safe analgesia during the initial hours of the postoperative period . Therefore, SAP block is more feasible for minimally invasive procedures, specifically for cardiac surgeries involving lateral thoracic incisions. A recent study comparing the efficacy of SAP block, IPP + PSP block, and intercostal nerve block for the management of postoperative thoracotomy pain in pediatric cardiac surgery revealed that while the early postoperative pain scores were comparable across all three groups, the SAP block group showed significantly lower pain scores at 6, 8, 10, and 12 h post-extubation compared to the intercostal nerve block group . These results suggest that SAP block may offer a more sustained analgesic effect compared to intercostal nerve block, with a comparable safety profile, making it a promising option for postoperative pain management in pediatric cardiac surgery. Parasternal intercostal plane (PIP) block Blocking the anterior cutaneous branches of the thoracic nerve is important for effective pain management after pediatric cardiac surgery . As an alternative to central blocks, targeting the terminal branches innervating the median sternotomy area via direct blockade offers a viable approach to control this specific type of pain . Superficial and deep parasternal intercostal blocks, previously known as pecto-intercostal fascial and transversus thoracic plane blocks, respectively, are increasingly recognized as effective RA techniques for pediatric cardiac surgery, particularly for managing the anterior thoracic wall pain . Superficial-PIP (S-PIP) block, a fascial block that provides analgesia in the parasternal region, decreases opioid consumption following median sternotomy by targeting the anterior cutaneous branches of the thoracic intercostal nerves (Th2–6). In this technique, LA is injected into the fascial plane between the pectoralis major and internal intercostal muscles (Fig. ). By effectively managing the somatic pain at the incision site, S-PIP minimizes the requirement for systemic opioids. This is particularly beneficial for pediatric patients susceptible to opioid-related side effects, including nausea and pruritus . Deep-PIP (D-PIP) block, which involves the injection of LA between the internal intercostal and transversus thoracic muscles, targets the anterior branches of the intercostal nerves (Fig. ). Although similar analgesic efficacy of both techniques has been demonstrated in adult patients , to date, no study has compared their effects in pediatric populations. Although cadaveric studies suggest the need for multiple injections for the parasternal block, this remains a topic of ongoing debate . Future studies on dermatomes are necessary to provide definitive conclusions. A systematic review and meta-analysis of six studies including 601 pediatric patients revealed that D-PIP block was associated with a reduction in postoperative modified objective pain scores at 12 h, intraoperative opioid consumption, and postoperative opioid consumption with low evidence. These findings highlight the potential clinical utility of D-PIP as a part of multimodal analgesia strategies in pediatric cardiac surgery . Although complications are minimized when analgesic techniques are performed under ultrasound guidance, it is important to emphasize that the D-PIP block has not been validated for use in pediatric patients and carries significant risks. Notably, the risk of pneumothorax and the potential for injury to the internal thoracic artery represent critical challenges to its application. Due to the close proximity of the internal thoracic artery and the severe consequences that may arise from its injury, the D-PIP block should be considered only as an advanced technique and approached with extreme caution, particularly in pediatric settings . In summary, both the S-PIP and D-PIP blocks are effective alternatives to other regional techniques for pediatric cardiac surgery. They offer targeted analgesia for anterior thoracic pain with a favorable safety profile, making them well-suited for pediatric patients. The block characteristics are summarized in Table , highlighting key distinctions in technique, anatomical targets, and clinical applications of the regional techniques. Current challenges and complications RA poses inherent risks in pediatric patients undergoing cardiac surgery. Although not specifically developed for cardiac surgery, the European Society of Regional Anesthesia has published many documents on the application of RA for the pediatric population . As RA is typically administered while the patient is under general anesthesia, its administration presents unique challenges. Specifically, lack of active participation by the patient and inability to observe the warning signs that an awake patient can exhibit necessitate heightened vigilance for the observation of indirect indicators of LA toxicity. Therefore, every LA injection should be administered in small aliquots (0.1–0.2 mL/kg), with intermittent aspiration and careful monitoring for any changes in the T wave, heart rate, and blood pressure in the immediate minutes following the injection. Any alterations in these parameters should raise the suspicion of intravascular injection until otherwise proven . Determining a one-size-fits-all dosing regimen is challenging as each block exhibits a distinct pharmacokinetic profile. Fascial blocks are characterized by rapid absorption , necessitating careful dosing regimens considering the higher distribution volume and different protein assets in infants and children compared to those in adult patients . Therefore, the lowest possible volume and effective concentration should be used for the pediatric population, adhering to the recommended maximum LA dose per kilogram. The concentrations, main benefits, and disadvantages of LA used in pediatric cardiac surgery are presented in Table . Adhering to these limits helps minimize the risk of systemic toxicity, which is crucial given the potential for severe adverse effects such as cardiac arrhythmias and central nervous system disturbances. However, the lack of pharmacokinetic data impairs the recommendations for dose (both bolus and infusion) in children. Until sufficient pharmacokinetic studies are available to establish the optimal dosing regimens for each RA technique, clinicians must prioritize safety by meticulously calculating doses based on patient weight and closely monitoring for signs of LA toxicity. While adult and pediatric populations share some anatomical and procedural similarities, the clinical evidence base for RA in pediatric cardiac surgery is notably smaller. It is therefore critical to distinguish between findings from adult studies and those specific to pediatric populations, ensuring that evidence is not extrapolated without careful consideration. RA provides a targeted approach for pain control by interrupting the transmission of pain signals at the site of injury, resulting in more efficient and prolonged pain relief than conventional IV analgesia. In 2023, network meta-analysis of over 5,000 adult patients undergoing cardiac surgery revealed that RA techniques provided superior pain control when compared to conventional IV analgesia , similar results were reported for pediatric patients, even if the total population of the meta-analysis comprised only 605 children from 14 RCTs . Targeted approaches of regional techniques facilitate effective pain management and reduce the need for systemic opioids, thereby decreasing the associated side effects . Ineffective postoperative pain management leads to chronic pain, immunosuppression, infections, and impaired wound healing . Moreover, uncontrolled pain in children can lead to both immediate and long-term consequences, including heightened anxiety and potential behavioral disturbances . Traditionally, IV analgesia is used as the primary method for pain control; however, it is associated with adverse effects that can prolong the ICU and hospital stay of the patients . RA is increasingly used for pediatric cardiac surgery in hopes it will enhance the clinical outcomes and reduce complications . Neuraxial techniques, such as the epidural technique, suppress the inflammatory response and sympathetic activity, which is crucial for patients as extracorporeal circulation, hypothermia, and surgical stress exacerbate the inflammatory response . RA reduces opioid consumption and pain score via effective postoperative pain relief , thereby decreasing the incidence of opioid-induced respiratory depression and other adverse events, such as gastrointestinal disturbances, pruritus, opioid hypersensitivity, and oversedation, which complicate postoperative treatment and prolong recovery . Prolonged ICU and hospital stay poses a significant burden on patients and their families, especially on pediatric patients as being away from their parents significantly influences their motivation and cooperation with subsequent treatment . Additionally, pediatric patients exhibit the most intense negative psychological and behavioral outcomes following critical illness and prolonged ICU stay . Prolonged stay also imposes a significant financial strain on healthcare systems due to the need for intensive care and related treatments. Effective pain management alleviates these social and economic burdens while improving patient outcomes . In pediatric cardiac surgery, assessing pain during intensive care follow-up is challenging due to the difficulty of using self-reported tests; therefore, behavioral observation becomes critical. Tools, such as the Face, Legs, Activity, Cry, and Consolability scale and Modified Objective Pain Scale, rely on observing physical signs of distress, such as facial expressions, body movements, crying, and the ability to be comforted . These tools are especially useful for infants and younger children who cannot verbalize their pain and older children with severe illnesses, which limit self-reporting. However, interpreting child behavior is another challenge. Pain behaviors of children are impacted by fear, anxiety, fatigue, and prior experience, complicating their assessment. Moreover, personal bias and cultural background of the observer also influences the interpretation of patient behaviors . Therefore, physiological measures provide another avenue for pain assessment, relying on objective indicators, such as changes in heart rate, blood pressure, and respiratory rate . However, physiological responses are not specific to pain and can be triggered by other factors, such as stress and illness. Therefore, physiological data combined with self-reports and behavioral observations should be used for accurate assessment . Central neuraxial techniques, including spinal , caudal , and epidural techniques, are used for analgesia during pediatric cardiac surgery. These techniques have demonstrated significant benefits in perioperative pain control. However, their application can cause potential complications, such as epidural hematoma, especially in patients receiving anticoagulation therapy, and unpredictable spread of LAs . Epidural analgesia (EA) remains a cornerstone technique in this context involving the injection of LAs, often combined with opioids, into the epidural space. Beyond its potent analgesic properties, EA also modulates the stress response via sympathetic blockade, stabilizing the hemodynamics during surgery by reducing catecholamine release. This is especially beneficial for pediatric patients, as controlling the body response to surgical stress is critical for maintaining cardiovascular stability in these patients . Paravertebral Block (PVB) involves injecting the LAs near the spinal nerves as they exit the intervertebral foramina, targeting the paravertebral space. Ultrasound guidance reduces the risk of complications, such as pleural puncture. Unlike those of EA, effects of PVB are more localized, resulting in fewer systemic effects, especially on hemodynamic stability. Therefore, PVB is a safe option for children with cardiovascular compromise . Erector spinae plane (ESP) block ESP block, first described by Forero et al. , is an emerging fascial plane block in which LAs are injected into the fascial plane between the erector spinae muscles and thoracic transverse processes (Fig. ). Although the specific mechanism of ESP block is still under debate , the most likely mechanism is that the administered LA spreads in the craniocaudal direction and into the paravertebral area, blocking the dorsal and ventral rami of the spinal nerves and sympathetic ganglia, facilitating both somatic and visceral sensory blockade. When applied bilaterally, ESP block effectively provides analgesia across specific dermatomes, particularly in the T2–7 regions, thereby covering the entire thorax, however the debase is still ongoing as some cadaveric studies suggests the staining of ventral rami could not be constant . Many studies have demonstrated the efficiency of the ESP block for pediatric cardiac surgery. In an RCT by Gado et al., postoperative opioid consumption and pain scores in the first 24 h were significantly lower in pediatric patients with ESP block than in those without ESP block . Another study reported that bilateral ESP block effectively relieves sternotomy pain, requiring less rescue analgesia and exerting long-lasting effects in the postoperative period . Gustavo et al. compared an ESP block performed at the T5 level with traditional IV opioids and reported that the ESP block was associated with a shorter ICU stay, lower opioid consumption, and faster discharge from the ICU . A systematic review and meta-analysis of five studies, including 384 pediatric cardiac patients—178 of whom received an ESP block—demonstrated notable benefits compared to IV opioid-based analgesia. The analysis revealed that the ESP block significantly reduced intraoperative fentanyl consumption and ICU length of stay in children undergoing cardiac surgery via midline sternotomy. This meta-analysis highlights the potential of the ESP block as a valuable adjunct to existing multimodal analgesia protocols . Mid-transverse process to pleura block (MTPB) and retrolaminar block (RLB) MTPB and RLB are two variants with a supposed mechanism of action similar to the ESP block (Fig. ). They work as the LA spreads through the paravertebral space, simultaneously blocking the spinal nerve roots of the ventral and dorsal rami, as the intercostal nerve is blocked at a level not far from the spinal cord . A cadaveric study demonstrated the extensive spread of MTPB into the paravertebral space, ESP, dorsal and ventral rami of the spinal nerves, sympathetic chain, and intercostal nerves . The injection was targeted midway between the transverse process of the thoracic vertebra and pleura, posterior to the superior costotransverse ligament, causing pleural displacement and spreading into the ESP. Abdelbaser et al. reported that MTPB significantly reduces intraoperative fentanyl consumption by blunting the stress response to various surgical stimuli during pediatric cardiac surgery. Moreover, MTPB significantly reduces the extubation time and duration of ICU stay, thereby decreasing the pain score. Another RCT revealed no significant differences in intraoperative hemodynamics, intraoperative fentanyl consumption, 24-h postoperative fentanyl consumption, postoperative pain score, extubation time, and ICU discharge time between the PVB and MTPB groups, with significantly shorter time required to perform bilateral MTPB than for PVB . RLB is another fascial plane block that involves LA injection between the posterior surface of the thoracic vertebral lamina and overlying paraspinal muscles . ESP block targets the tips of transverse processes, whereas RLB targets the laminae . Spread of LA into the paravertebral and epidural spaces blocks the ventral and dorsal rami of the thoracic spinal nerves and extends laterally into the fascial plane to block the lateral cutaneous branches of intercostal nerves . The potential superiority of RLB over ESP block is probably due to the deeper deposition that may facilitate more consistent diffusion of the LA toward the nerve roots and paravertebral spaces, areas critical for achieving effective somatic and visceral analgesia. A recent network meta-analysis which assessed the RA techniques in pediatric cardiothoracic surgery reported that the largest opioid consumption decrease was in RLB . Ultrasound-guided RLB is theoretically safer than thoracic EA and PVB, as the block needle is inserted away from the pleura and dura . Moreover, no major vessels or nerves exist along the needle path, with only a slight risk of intramuscular hemorrhage . RLB is associated with early extubation and short ICU stay due to the reduction in perioperative opioid consumption after open cardiac surgery via median sternotomy . Interpectoral plane (IPP) and IPP + pectoserratus plane (PSP) blocks IPP (previously known as PECS-I) block was first described by Blanco in 2011 while IPP + PSP (previously known as PECS-II) block was introduced the following year . IPP block involves a single injection of the LA between the pectoralis major and minor muscles, whereas the IPP + PSP block involves an additional injection into the plane between the pectoralis minor and serratus anterior muscles (SAMs) (Fig. ). IPP block targets the pectoral nerves and possibly the intercostal nerves depending on the injection site, whereas the IPP + PSP block targets the long thoracic, thoracodorsal, and lateral branches of the intercostal nerves. Although the IPP block is widely used in other context such as breast surgery, its use for cardiac surgery, particularly in pediatric populations, is still in its early stages. To date, most studies have focused on adult populations, leading to insufficient data on pediatric patients, for this reason authors would recommend caution to readers while implementing these techniques in their clinical practice until further evidence is provided by future studies on the topic. Zachary et al. retrospectively evaluated the use of IPP and IPP + PSP blocks for postoperative pain management following sternotomy in 73 pediatric patients. IPP was performed in 47 patients and IPP + PSP in 26 patients. Notably, 34% of patients did not experience severe pain within the first 24 h post-surgery . IPP + PSP block decreases the pain score, opioid consumption, and agitation score . In a retrospective study conducted by Yang et al. in pediatric patients undergoing pacemaker or defibrillator implantation, IPP + PSP block yielded positive outcomes by decreasing the pain score and opioid requirement . IPP and IPP + PSP blocks are more suitable for minimally invasive surgeries, such as pacemaker and defibrillator placement, than anterior mediastinotomy. ESP block, first described by Forero et al. , is an emerging fascial plane block in which LAs are injected into the fascial plane between the erector spinae muscles and thoracic transverse processes (Fig. ). Although the specific mechanism of ESP block is still under debate , the most likely mechanism is that the administered LA spreads in the craniocaudal direction and into the paravertebral area, blocking the dorsal and ventral rami of the spinal nerves and sympathetic ganglia, facilitating both somatic and visceral sensory blockade. When applied bilaterally, ESP block effectively provides analgesia across specific dermatomes, particularly in the T2–7 regions, thereby covering the entire thorax, however the debase is still ongoing as some cadaveric studies suggests the staining of ventral rami could not be constant . Many studies have demonstrated the efficiency of the ESP block for pediatric cardiac surgery. In an RCT by Gado et al., postoperative opioid consumption and pain scores in the first 24 h were significantly lower in pediatric patients with ESP block than in those without ESP block . Another study reported that bilateral ESP block effectively relieves sternotomy pain, requiring less rescue analgesia and exerting long-lasting effects in the postoperative period . Gustavo et al. compared an ESP block performed at the T5 level with traditional IV opioids and reported that the ESP block was associated with a shorter ICU stay, lower opioid consumption, and faster discharge from the ICU . A systematic review and meta-analysis of five studies, including 384 pediatric cardiac patients—178 of whom received an ESP block—demonstrated notable benefits compared to IV opioid-based analgesia. The analysis revealed that the ESP block significantly reduced intraoperative fentanyl consumption and ICU length of stay in children undergoing cardiac surgery via midline sternotomy. This meta-analysis highlights the potential of the ESP block as a valuable adjunct to existing multimodal analgesia protocols . MTPB and RLB are two variants with a supposed mechanism of action similar to the ESP block (Fig. ). They work as the LA spreads through the paravertebral space, simultaneously blocking the spinal nerve roots of the ventral and dorsal rami, as the intercostal nerve is blocked at a level not far from the spinal cord . A cadaveric study demonstrated the extensive spread of MTPB into the paravertebral space, ESP, dorsal and ventral rami of the spinal nerves, sympathetic chain, and intercostal nerves . The injection was targeted midway between the transverse process of the thoracic vertebra and pleura, posterior to the superior costotransverse ligament, causing pleural displacement and spreading into the ESP. Abdelbaser et al. reported that MTPB significantly reduces intraoperative fentanyl consumption by blunting the stress response to various surgical stimuli during pediatric cardiac surgery. Moreover, MTPB significantly reduces the extubation time and duration of ICU stay, thereby decreasing the pain score. Another RCT revealed no significant differences in intraoperative hemodynamics, intraoperative fentanyl consumption, 24-h postoperative fentanyl consumption, postoperative pain score, extubation time, and ICU discharge time between the PVB and MTPB groups, with significantly shorter time required to perform bilateral MTPB than for PVB . RLB is another fascial plane block that involves LA injection between the posterior surface of the thoracic vertebral lamina and overlying paraspinal muscles . ESP block targets the tips of transverse processes, whereas RLB targets the laminae . Spread of LA into the paravertebral and epidural spaces blocks the ventral and dorsal rami of the thoracic spinal nerves and extends laterally into the fascial plane to block the lateral cutaneous branches of intercostal nerves . The potential superiority of RLB over ESP block is probably due to the deeper deposition that may facilitate more consistent diffusion of the LA toward the nerve roots and paravertebral spaces, areas critical for achieving effective somatic and visceral analgesia. A recent network meta-analysis which assessed the RA techniques in pediatric cardiothoracic surgery reported that the largest opioid consumption decrease was in RLB . Ultrasound-guided RLB is theoretically safer than thoracic EA and PVB, as the block needle is inserted away from the pleura and dura . Moreover, no major vessels or nerves exist along the needle path, with only a slight risk of intramuscular hemorrhage . RLB is associated with early extubation and short ICU stay due to the reduction in perioperative opioid consumption after open cardiac surgery via median sternotomy . IPP (previously known as PECS-I) block was first described by Blanco in 2011 while IPP + PSP (previously known as PECS-II) block was introduced the following year . IPP block involves a single injection of the LA between the pectoralis major and minor muscles, whereas the IPP + PSP block involves an additional injection into the plane between the pectoralis minor and serratus anterior muscles (SAMs) (Fig. ). IPP block targets the pectoral nerves and possibly the intercostal nerves depending on the injection site, whereas the IPP + PSP block targets the long thoracic, thoracodorsal, and lateral branches of the intercostal nerves. Although the IPP block is widely used in other context such as breast surgery, its use for cardiac surgery, particularly in pediatric populations, is still in its early stages. To date, most studies have focused on adult populations, leading to insufficient data on pediatric patients, for this reason authors would recommend caution to readers while implementing these techniques in their clinical practice until further evidence is provided by future studies on the topic. Zachary et al. retrospectively evaluated the use of IPP and IPP + PSP blocks for postoperative pain management following sternotomy in 73 pediatric patients. IPP was performed in 47 patients and IPP + PSP in 26 patients. Notably, 34% of patients did not experience severe pain within the first 24 h post-surgery . IPP + PSP block decreases the pain score, opioid consumption, and agitation score . In a retrospective study conducted by Yang et al. in pediatric patients undergoing pacemaker or defibrillator implantation, IPP + PSP block yielded positive outcomes by decreasing the pain score and opioid requirement . IPP and IPP + PSP blocks are more suitable for minimally invasive surgeries, such as pacemaker and defibrillator placement, than anterior mediastinotomy. SAP block is performed more laterally and posteriorly than the interpectoral block in the axillary region at the level of the fourth or fifth rib. In this block, LA is injected either between the serratus anterior and latissimus dorsi muscles or beneath the SAMs (Fig. ). Depending on whether the injection is deep or superficial, different nerve branches are affected: LA injected deep into the SAM blocks the lateral cutaneous branches of the intercostal nerves, whereas a superficial injection blocks the long thoracic and thoracodorsal nerves and lateral cutaneous branch of the intercostal nerve . SAP block spreads across the T2–9 levels, covering the anterior, lateral, and posterior chest walls. Specifically, SAP block provides effective and safe analgesia during the initial hours of the postoperative period . Therefore, SAP block is more feasible for minimally invasive procedures, specifically for cardiac surgeries involving lateral thoracic incisions. A recent study comparing the efficacy of SAP block, IPP + PSP block, and intercostal nerve block for the management of postoperative thoracotomy pain in pediatric cardiac surgery revealed that while the early postoperative pain scores were comparable across all three groups, the SAP block group showed significantly lower pain scores at 6, 8, 10, and 12 h post-extubation compared to the intercostal nerve block group . These results suggest that SAP block may offer a more sustained analgesic effect compared to intercostal nerve block, with a comparable safety profile, making it a promising option for postoperative pain management in pediatric cardiac surgery. Blocking the anterior cutaneous branches of the thoracic nerve is important for effective pain management after pediatric cardiac surgery . As an alternative to central blocks, targeting the terminal branches innervating the median sternotomy area via direct blockade offers a viable approach to control this specific type of pain . Superficial and deep parasternal intercostal blocks, previously known as pecto-intercostal fascial and transversus thoracic plane blocks, respectively, are increasingly recognized as effective RA techniques for pediatric cardiac surgery, particularly for managing the anterior thoracic wall pain . Superficial-PIP (S-PIP) block, a fascial block that provides analgesia in the parasternal region, decreases opioid consumption following median sternotomy by targeting the anterior cutaneous branches of the thoracic intercostal nerves (Th2–6). In this technique, LA is injected into the fascial plane between the pectoralis major and internal intercostal muscles (Fig. ). By effectively managing the somatic pain at the incision site, S-PIP minimizes the requirement for systemic opioids. This is particularly beneficial for pediatric patients susceptible to opioid-related side effects, including nausea and pruritus . Deep-PIP (D-PIP) block, which involves the injection of LA between the internal intercostal and transversus thoracic muscles, targets the anterior branches of the intercostal nerves (Fig. ). Although similar analgesic efficacy of both techniques has been demonstrated in adult patients , to date, no study has compared their effects in pediatric populations. Although cadaveric studies suggest the need for multiple injections for the parasternal block, this remains a topic of ongoing debate . Future studies on dermatomes are necessary to provide definitive conclusions. A systematic review and meta-analysis of six studies including 601 pediatric patients revealed that D-PIP block was associated with a reduction in postoperative modified objective pain scores at 12 h, intraoperative opioid consumption, and postoperative opioid consumption with low evidence. These findings highlight the potential clinical utility of D-PIP as a part of multimodal analgesia strategies in pediatric cardiac surgery . Although complications are minimized when analgesic techniques are performed under ultrasound guidance, it is important to emphasize that the D-PIP block has not been validated for use in pediatric patients and carries significant risks. Notably, the risk of pneumothorax and the potential for injury to the internal thoracic artery represent critical challenges to its application. Due to the close proximity of the internal thoracic artery and the severe consequences that may arise from its injury, the D-PIP block should be considered only as an advanced technique and approached with extreme caution, particularly in pediatric settings . In summary, both the S-PIP and D-PIP blocks are effective alternatives to other regional techniques for pediatric cardiac surgery. They offer targeted analgesia for anterior thoracic pain with a favorable safety profile, making them well-suited for pediatric patients. The block characteristics are summarized in Table , highlighting key distinctions in technique, anatomical targets, and clinical applications of the regional techniques. RA poses inherent risks in pediatric patients undergoing cardiac surgery. Although not specifically developed for cardiac surgery, the European Society of Regional Anesthesia has published many documents on the application of RA for the pediatric population . As RA is typically administered while the patient is under general anesthesia, its administration presents unique challenges. Specifically, lack of active participation by the patient and inability to observe the warning signs that an awake patient can exhibit necessitate heightened vigilance for the observation of indirect indicators of LA toxicity. Therefore, every LA injection should be administered in small aliquots (0.1–0.2 mL/kg), with intermittent aspiration and careful monitoring for any changes in the T wave, heart rate, and blood pressure in the immediate minutes following the injection. Any alterations in these parameters should raise the suspicion of intravascular injection until otherwise proven . Determining a one-size-fits-all dosing regimen is challenging as each block exhibits a distinct pharmacokinetic profile. Fascial blocks are characterized by rapid absorption , necessitating careful dosing regimens considering the higher distribution volume and different protein assets in infants and children compared to those in adult patients . Therefore, the lowest possible volume and effective concentration should be used for the pediatric population, adhering to the recommended maximum LA dose per kilogram. The concentrations, main benefits, and disadvantages of LA used in pediatric cardiac surgery are presented in Table . Adhering to these limits helps minimize the risk of systemic toxicity, which is crucial given the potential for severe adverse effects such as cardiac arrhythmias and central nervous system disturbances. However, the lack of pharmacokinetic data impairs the recommendations for dose (both bolus and infusion) in children. Until sufficient pharmacokinetic studies are available to establish the optimal dosing regimens for each RA technique, clinicians must prioritize safety by meticulously calculating doses based on patient weight and closely monitoring for signs of LA toxicity. While adult and pediatric populations share some anatomical and procedural similarities, the clinical evidence base for RA in pediatric cardiac surgery is notably smaller. It is therefore critical to distinguish between findings from adult studies and those specific to pediatric populations, ensuring that evidence is not extrapolated without careful consideration. Although RA is widely used in multimodal analgesia for pediatric cardiac surgery, high-quality studies on its efficiency are scarce, limiting robust evidence supporting its use. However, recent advancements, such as expanding anatomical knowledge and increasing use of ultrasonography-guided fascial plane blocks, have enhanced its application. Current shift toward opioid-sparing techniques with multimodal analgesia marks the growing demand for fast-track cardiac surgery. In such surgeries, RA plays a crucial role for enhanced recovery by facilitating early extubation and decreasing the hospital stay duration. However, further large-scale randomized multicenter studies are essential to standardize the application of RA in pediatric cardiac surgery and provide robust evidence for its safe and effective use for the pediatric population. |
Clinicopathological Parameters and Immunohistochemical Profiles in Correlation with MRI Characteristics in Glioblastomas | 8e14829c-58c6-4ae7-8d5a-a094b3d495f9 | 11642654 | Anatomy[mh] | Glioblastoma is the most aggressive type of malignant intracranial tumor of glial origin in adults, characterized by a high rate of local recurrence. The standard treatment for glioblastoma is multimodal; however, the survival rate typically does not exceed 15 months from the time of diagnosis . The invasive characteristics of glioblastomas are closely associated with the tumor microenvironment. This microenvironment consists of several components, including tumor glial cells, endothelial cells, immune cells, and extracellular matrix elements . Matrix metalloproteinases (MMPs) are zinc-dependent proteolytic enzymes, also referred to as endopeptidases, that are involved in the remodeling and degradation of the extracellular matrix. They degrade components such as collagen, fibronectin, and laminins by utilizing cytokines and growth factors . Matrix metalloproteinase 9 (MMP-9), also known as type IV collagenase, gelatinase, or gelatinase B, plays a key role in the disruption of the blood–brain barrier and facilitates the release of local tumor growth factors associated with the promotion of angiogenesis . Additionally, MMP-9 promotes the proliferation of endothelial cells by releasing vascular endothelial growth factor (VEGF), leading to increased vascular permeability; thus, its pro-angiogenic role in glioblastomas contributes significantly to neovascularization . MMP-9 is also involved in remodeling the tumor microenvironment, which may enhance tumor invasion and progression, and is associated with a poor prognosis . Magnetic resonance imaging (MRI) is widely used as a non-invasive diagnostic tool for central nervous system tumors. MRI plays an advanced role in the management of patients with glioblastoma (GBM) . Imaging data, alongside the extent of surgical resection, use of adjuvant therapy, and levels of tumor biomarkers, can serve as prognostic factors . Anatomical characteristics, such as tumor location and laterality, have been compared with specific molecular alterations in glioblastomas, including VEGF and MGMT . Some studies have shown that glial tumor cells in glioblastomas can also be present in the peritumoral edema area . The development of peritumoral edema is influenced by several factors that act through mechanisms promoting the release of substances that increase vascular permeability, such as vascular endothelial growth factor (VEGF), associated with elevated expression of metalloproteinases, which can further stimulate tumor angiogenesis. Another contributing factor to the onset of peritumoral edema is the overexpression of Aquaporin 4 by tumor glial cells through a mechanism that is not yet fully understood . Regarding angiogenesis, which can contribute to the development of peritumoral edema, several phenotypic and functional characteristics of endothelial cells in glioblastoma have been demonstrated. These cells form tight connections with tumor cells, altering tissue homeostasis towards a microenvironment with a structure and function more favorable to the tumor. In newly formed vessels within the tumor stroma, most endothelial cells are CD31 and CD34 immunopositive and co-express VEGF. In contrast, most endothelial cells in the peritumoral edema area express CD31 and CD34, but lose VEGF immunoexpression. The decreased expression of VEGF in these endothelial cells may be a response to the therapeutic failure of anti-angiogenic drugs . Imaging data, such as tumor size or location associated with peritumoral edema, have been correlated with the degree of tumor invasion or patient survival. However, studies have shown conflicting results on this topic; some consider peritumoral edema to be a poor prognostic factor, while other authors have reported inconclusive data . The aim of this study was to analyze the relationship between tumor angiogenesis in glioblastomas in association with MMP-9 immunoexpression. The results were correlated with the Ki-67 proliferation index, p53 immunoexpression, and the mutational status of IDH1 and ATRX , as well as MRI imaging data. 2.1. MMP-9 Immunoexpression in Glioblastoma In our study, we included 44 patients diagnosed with glioblastoma. Over 50% of the glioblastomas examined exhibited MMP-9 immunoexpression. It can be observed that in some glial cells, the cytoplasmic immunoexpression was variable, with differing reaction intensities . We observed endothelial cells with MMP-9 immunoexpression, particularly in areas where glomeruloid structures were present in the tumor stroma (arrows) . The expression of matrix metalloproteinase 9 (MMP-9) varied widely with patient age. It was observed that as age increased, MMP-9 immunoexpression was more pronounced in the studied glioblastomas. Among patients under 50 years of age, MMP-9 immunoexpression was present in 37.5% (6/16) of cases, whereas 62.5% (10/16) of cases showed no MMP-9 immunoexpression. In contrast, among patients over 65 years of age, there was a notable increase in MMP-9 immunoexpression, with 78.57% (11/14) of cases showing positive staining and only 21.42% (3/14) of cases showing no MMP-9 immunoexpression. The number of cases with or without MMP-9 immunoexpression was equal among patients aged 50 to 65 years. As age increased linearly, no statistically significant correlation was found between MMP-9 immunoexpression and patient age ( p = 0.28, r = 0.16). Regarding gender differences, a predominance of MMP-9 immunoexpression was observed in female patients, at 59.09% (13/22). No statistically significant correlation was found between MMP-9 immunoexpression and either the age group or gender of the patients ( p = 0.07, p = 0.54) . Regarding localization, almost 45.83% of the cases with MMP-9 immunoexpression were found in the temporal lobe (11/24). In the occipital lobe, MMP-9 immunoexpression was observed twice as frequently as cases without MMP-9 immunoexpression. MMP-9-positive cases were more common in the right cerebral hemisphere, at a rate of 58.33% (14/24). Conversely, MMP-9-negative tumors were more frequent in the left hemisphere (12/20, 60%). No statistically significant correlation was found between tumor localization, laterality, and MMP-9 immunoexpression ( p = 0.91, p = 0.22) . The majority of IDH-1 wild-type glioblastomas (57.5%, 23/40) exhibited MMP-9 immunoexpression, in contrast to IDH1 mutant glioblastomas, where MMP-9 immunoexpression was significantly lower (1/4). Regarding ATRX immunoexpression, most ATRX wild-type cases also presented MMP-9 immunoexpression, at a rate of 55.88% (19/34). The loss of ATRX immunoexpression did not affect MMP-9 immunoexpression, with an equal number of cases showing either presence or absence of MMP-9. No statistically significant association was observed between the presence or absence of IDH1 or ATRX mutations and MMP-9 immunoexpression ( p = 0.21, p = 0.74) . In glioblastomas with mutant p53, MMP-9 immunoexpression was absent in most cases, accounting for 63.63% (7/11). In contrast, MMP-9 immunoexpression was more frequently observed in p53 wild-type glioblastomas (60.60%, 20/33). Regarding the Ki-67 index, it was noted that as the Ki-67 index increased, MMP-9-positive GBM became less frequent, indicating a decline in MMP-9 immunoexpression. GBMs with Ki-67 indexes below 5% were more often associated with MMP-9 immunoexpression, at a rate of 73.33% (11/15), with a ratio of MMP-9-positive to -negative cases of 2.75 (11:4). Conversely, in glioblastomas with Ki-67 indexes above 20%, the proportion of MMP-9-positive cases decreased to 41.17% (7/17), and MMP-9-negative GBMs were more common (10/17, 58.82%). No statistically significant association was found between the Ki-67 index or p53 status and MMP-9 immunoexpression ( p = 0.16, p = 0.17) . Considering the microvascular density (MVD) determined by CD34 immunoexpression, we observed that MMP-9 expression was more frequently low or absent in GBM cases with higher MVD-CD34. The median microvascular density measured by CD34 was 1.75% (0.9–4.1) in GBM with MMP-9 expression. In cases without MMP-9 expression, the median microvascular density was higher, at 3.21% (1.7–6.9). Regarding MVD-CD105, cases with MMP-9 expression showed a median microvascular density of 2.65% (0.9–4.9), while those without MMP-9 expression had a slightly higher density of 2.93% (1.3–5.0). No statistically significant association was found between microvascular density measured by CD34 or CD105 and MMP-9 expression ( p = 0.072, p = 0.79) . Additionally, a nearly similar distribution of percentage values of microvascular densities, both by MVD-CD34 and MVD-CD105, was observed in MMP-9-positive and -negative tumors . 2.2. MRI Characteristics of Glioblastomas 2.2.1. Tumor Size and Tumor Volume In more than 65% of the studied cases, the maximum tumor diameter was below 5 cm (29/S44), with a median diameter of 4.63 cm. Tumor volume measurements ranged from 8.75 cm 3 to 124.56 cm 3 , with an average tumor volume of 43.17 cm 3 . Regarding tumor size, most glioblastomas with a maximum diameter of less than 5 cm were found in patients over 50 years old (20/44) ( p = 0.43). In terms of gender, glioblastomas smaller than 5 cm were more common in male patients (18/22). Additionally, the proportion of tumors larger than 5 cm was higher among women (50%, 11/22) compared to men (18.18%, 4/22). A statistically significant association was observed between tumor size and patient gender ( p = 0.02) . The median tumor volume decreased with increasing age, with the highest tumor volumes recorded in patients under 50 years old. The median tumor volume was higher in women (42.1 cm 3 ) compared to men (38.6 cm 3 ). However, no statistically significant association was found between median tumor volume and patient age or gender ( p = 0.859, p = 0.477) . The median tumor volume was highest in GBM located in the frontal lobe, followed by the occipital lobe, with the smallest values recorded in the temporal and parietal lobes, which had similar proportions. The median tumor volume was larger in the left hemisphere, with a value of 45.6 cm 3 . No statistically significant correlation was observed between tumor size, median tumor volume, localization, and laterality ( p = 0.95, p = 0.75, p = 0.39, p = 0.77) ( and ). In 65% (26/40) of the cases, we observed IDH1 wild-type glioblastomas with sizes under 5 cm, while 25% (1/4) of the cases were IDH1 mutant glioblastomas larger than 5 cm. The median tumor volume was higher in IDH1 mutant glioblastomas (49.8 cm 3 ). Nearly 61.76% (21/34) of ATRX wild-type cases had sizes below 5 cm. The ratio of ATRX mutant cases with maximum diameters <5 cm versus >5 cm was 8:2 = 4, while in ATRX wild-type GBM cases, the ratio was 21:13 = 1.6. The median tumor volume was approximately equal in both ATRX wild-type and mutant cases. No statistically significant association was found between tumor size, median tumor volume, and IDH1 or ATRX mutations ( p = 0.68, p = 0.28, p = 0.27, p = 0.85) ( and ). Regarding the p53 mutation, we observed that 66.66% (22/33) of the p53 wild-type cases had tumors smaller than 5 cm, while 36.36% (4/11) of the cases showed the p53 mutation associated with tumor sizes greater than 5 cm. The median tumor volume was higher in p53 mutant glioblastomas (48.6 cm 3 ) compared to p53 wild-type cases. A Ki67 index greater than 5% was more frequently observed in tumors smaller than 5 cm (21/29). The median tumor volume was larger in glioblastomas with a Ki67 index above 20%. Concerning the relationship with MMP-9, cases without MMP-9 expression had relatively higher median tumor volumes (43.5 cm 3 vs. 39.8 cm 3 ) compared to glioblastomas with MMP-9 expression. No statistically significant correlation was found between tumor size; median tumor volume; and Ki67 index, p53 mutation, or MMP-9 expression ( and , ). Regarding microvascular density, no statistically significant correlation was observed between tumor volume and microvascular density measured by either CD34 or CD105. However, the correlation coefficient suggests a positive correlation for MVD-CD34 and a negative correlation for MVD-CD105 ( p = 0.687, p = 0.658) . 2.2.2. Peritumoral Edema Peritumoral edema ranged in size from 7.2 mm to 63.9 mm. The median thickness of peritumoral edema was higher among women (29.1 mm) and in patients aged between 51 and 65 years (30.6 mm). Median values recorded in the temporal lobe were nearly similar to those in the parietal lobe. The right cerebral hemisphere exhibited a higher median value of peritumoral edema compared to the contralateral hemisphere. No statistically significant association was observed between the median values of peritumoral edema and age, sex, tumor localization, or laterality ( and ). Differences in the median values of peritumoral edema based on the presence or absence of the IDH1 mutation were not statistically significant. The median values of peritumoral edema were higher in ATRX mutant cases compared to ATRX wild-type glioblastomas. The median value of peritumoral edema was higher in p53 wild-type glioblastomas and in cases with Ki67 indexes below 5%. Differences in the median value of peritumoral edema concerning the presence or absence of MMP-9 immunoexpression were not significant. No statistically significant association was found between median values of peritumoral edema and ATRX , p53 mutations, or Ki67 index ( and , ). Regarding microvascular density, which was measured by both CD34 and CD105, no statistically significant correlation was observed between peritumoral edema thickness and microvascular density ( p = 0.107, p = 0.563) . Regarding the margin of peritumoral edema, most cases exhibited rounded margins (27/44). Predominantly in the temporal lobe, we observed edema with irregular contours. Irregular margins were more frequently identified in the right hemisphere, whereas the left hemisphere showed more cases of edema with rounded margins. A statistically significant association was observed between laterality and edema shape ( p = 0.03) . Most glioblastomas with IDH1 wild type, p53 wild type, and ATRX wild type exhibited rounded edema margins. The ratio of cases with regular versus irregular edema contours was 2 for GBM with Ki67 index below 20% and 1.12 for those with Ki67 indexes above 20%. MMP-9-positive cases more frequently presented irregular margins (10/24) compared to MMP-9-negative cases (7/20). No statistically significant association was found between age, gender, localization, IDH1 , ATRX , p53 mutations, Ki67 index, and MMP-9 with the characteristics of peritumoral edema shape . The median MVD-CD34 value was higher in GBM with irregular peritumoral edema, whereas MVD-CD105 showed higher median values in GBM associated with regular edema contours. No statistically significant association was observed between peritumoral edema margin characteristics and MVD-CD34 or MVD-CD105 . 2.2.3. Midline Deviation The median value of midline deviation was 6.1 mm. In 15.9% (7/44) of cases, no midline deviation was observed, while in 27.27% (12/44) of cases, the deviation exceeded 10 mm. Midline deviation was greater in males and in patients aged between 50 and 65 years. Higher values were noted in tumors located in the frontal lobe, followed by the parietal lobe. Tumors located in the right cerebral hemisphere were associated with more pronounced midline deviation. A statistically significant association was observed between age groups and midline deviation ( p = 0.006) . IDH1 mutant glioblastomas, p53 mutant glioblastomas, and ATRX mutant glioblastomas exhibited more pronounced midline deviation. Cases with Ki67 indexes above 20% and MMP-9-negative tumors resulted in more severe midline deviation. A statistically significant association was observed between ATRX mutation and midline deviation ( p = 0.035) . There was no statistically significant correlation between midline deviation and the median values of MVD-CD34 or MVD-CD105. However, a negative correlation was noted between the degree of deviation and MVD-CD105, and a positive correlation with MVD-CD34 . Based on the statistical results, this study did not demonstrate a correlation between tumor location and tumor volume, peritumoral edema size, or respective midline deviation . No statistically significant correlation was found between tumor volume and peritumoral edema or tumor volume and midline deviation. However, a positive correlation was observed between the peritumoral edema and the midline deviation . In our study, we included 44 patients diagnosed with glioblastoma. Over 50% of the glioblastomas examined exhibited MMP-9 immunoexpression. It can be observed that in some glial cells, the cytoplasmic immunoexpression was variable, with differing reaction intensities . We observed endothelial cells with MMP-9 immunoexpression, particularly in areas where glomeruloid structures were present in the tumor stroma (arrows) . The expression of matrix metalloproteinase 9 (MMP-9) varied widely with patient age. It was observed that as age increased, MMP-9 immunoexpression was more pronounced in the studied glioblastomas. Among patients under 50 years of age, MMP-9 immunoexpression was present in 37.5% (6/16) of cases, whereas 62.5% (10/16) of cases showed no MMP-9 immunoexpression. In contrast, among patients over 65 years of age, there was a notable increase in MMP-9 immunoexpression, with 78.57% (11/14) of cases showing positive staining and only 21.42% (3/14) of cases showing no MMP-9 immunoexpression. The number of cases with or without MMP-9 immunoexpression was equal among patients aged 50 to 65 years. As age increased linearly, no statistically significant correlation was found between MMP-9 immunoexpression and patient age ( p = 0.28, r = 0.16). Regarding gender differences, a predominance of MMP-9 immunoexpression was observed in female patients, at 59.09% (13/22). No statistically significant correlation was found between MMP-9 immunoexpression and either the age group or gender of the patients ( p = 0.07, p = 0.54) . Regarding localization, almost 45.83% of the cases with MMP-9 immunoexpression were found in the temporal lobe (11/24). In the occipital lobe, MMP-9 immunoexpression was observed twice as frequently as cases without MMP-9 immunoexpression. MMP-9-positive cases were more common in the right cerebral hemisphere, at a rate of 58.33% (14/24). Conversely, MMP-9-negative tumors were more frequent in the left hemisphere (12/20, 60%). No statistically significant correlation was found between tumor localization, laterality, and MMP-9 immunoexpression ( p = 0.91, p = 0.22) . The majority of IDH-1 wild-type glioblastomas (57.5%, 23/40) exhibited MMP-9 immunoexpression, in contrast to IDH1 mutant glioblastomas, where MMP-9 immunoexpression was significantly lower (1/4). Regarding ATRX immunoexpression, most ATRX wild-type cases also presented MMP-9 immunoexpression, at a rate of 55.88% (19/34). The loss of ATRX immunoexpression did not affect MMP-9 immunoexpression, with an equal number of cases showing either presence or absence of MMP-9. No statistically significant association was observed between the presence or absence of IDH1 or ATRX mutations and MMP-9 immunoexpression ( p = 0.21, p = 0.74) . In glioblastomas with mutant p53, MMP-9 immunoexpression was absent in most cases, accounting for 63.63% (7/11). In contrast, MMP-9 immunoexpression was more frequently observed in p53 wild-type glioblastomas (60.60%, 20/33). Regarding the Ki-67 index, it was noted that as the Ki-67 index increased, MMP-9-positive GBM became less frequent, indicating a decline in MMP-9 immunoexpression. GBMs with Ki-67 indexes below 5% were more often associated with MMP-9 immunoexpression, at a rate of 73.33% (11/15), with a ratio of MMP-9-positive to -negative cases of 2.75 (11:4). Conversely, in glioblastomas with Ki-67 indexes above 20%, the proportion of MMP-9-positive cases decreased to 41.17% (7/17), and MMP-9-negative GBMs were more common (10/17, 58.82%). No statistically significant association was found between the Ki-67 index or p53 status and MMP-9 immunoexpression ( p = 0.16, p = 0.17) . Considering the microvascular density (MVD) determined by CD34 immunoexpression, we observed that MMP-9 expression was more frequently low or absent in GBM cases with higher MVD-CD34. The median microvascular density measured by CD34 was 1.75% (0.9–4.1) in GBM with MMP-9 expression. In cases without MMP-9 expression, the median microvascular density was higher, at 3.21% (1.7–6.9). Regarding MVD-CD105, cases with MMP-9 expression showed a median microvascular density of 2.65% (0.9–4.9), while those without MMP-9 expression had a slightly higher density of 2.93% (1.3–5.0). No statistically significant association was found between microvascular density measured by CD34 or CD105 and MMP-9 expression ( p = 0.072, p = 0.79) . Additionally, a nearly similar distribution of percentage values of microvascular densities, both by MVD-CD34 and MVD-CD105, was observed in MMP-9-positive and -negative tumors . 2.2.1. Tumor Size and Tumor Volume In more than 65% of the studied cases, the maximum tumor diameter was below 5 cm (29/S44), with a median diameter of 4.63 cm. Tumor volume measurements ranged from 8.75 cm 3 to 124.56 cm 3 , with an average tumor volume of 43.17 cm 3 . Regarding tumor size, most glioblastomas with a maximum diameter of less than 5 cm were found in patients over 50 years old (20/44) ( p = 0.43). In terms of gender, glioblastomas smaller than 5 cm were more common in male patients (18/22). Additionally, the proportion of tumors larger than 5 cm was higher among women (50%, 11/22) compared to men (18.18%, 4/22). A statistically significant association was observed between tumor size and patient gender ( p = 0.02) . The median tumor volume decreased with increasing age, with the highest tumor volumes recorded in patients under 50 years old. The median tumor volume was higher in women (42.1 cm 3 ) compared to men (38.6 cm 3 ). However, no statistically significant association was found between median tumor volume and patient age or gender ( p = 0.859, p = 0.477) . The median tumor volume was highest in GBM located in the frontal lobe, followed by the occipital lobe, with the smallest values recorded in the temporal and parietal lobes, which had similar proportions. The median tumor volume was larger in the left hemisphere, with a value of 45.6 cm 3 . No statistically significant correlation was observed between tumor size, median tumor volume, localization, and laterality ( p = 0.95, p = 0.75, p = 0.39, p = 0.77) ( and ). In 65% (26/40) of the cases, we observed IDH1 wild-type glioblastomas with sizes under 5 cm, while 25% (1/4) of the cases were IDH1 mutant glioblastomas larger than 5 cm. The median tumor volume was higher in IDH1 mutant glioblastomas (49.8 cm 3 ). Nearly 61.76% (21/34) of ATRX wild-type cases had sizes below 5 cm. The ratio of ATRX mutant cases with maximum diameters <5 cm versus >5 cm was 8:2 = 4, while in ATRX wild-type GBM cases, the ratio was 21:13 = 1.6. The median tumor volume was approximately equal in both ATRX wild-type and mutant cases. No statistically significant association was found between tumor size, median tumor volume, and IDH1 or ATRX mutations ( p = 0.68, p = 0.28, p = 0.27, p = 0.85) ( and ). Regarding the p53 mutation, we observed that 66.66% (22/33) of the p53 wild-type cases had tumors smaller than 5 cm, while 36.36% (4/11) of the cases showed the p53 mutation associated with tumor sizes greater than 5 cm. The median tumor volume was higher in p53 mutant glioblastomas (48.6 cm 3 ) compared to p53 wild-type cases. A Ki67 index greater than 5% was more frequently observed in tumors smaller than 5 cm (21/29). The median tumor volume was larger in glioblastomas with a Ki67 index above 20%. Concerning the relationship with MMP-9, cases without MMP-9 expression had relatively higher median tumor volumes (43.5 cm 3 vs. 39.8 cm 3 ) compared to glioblastomas with MMP-9 expression. No statistically significant correlation was found between tumor size; median tumor volume; and Ki67 index, p53 mutation, or MMP-9 expression ( and , ). Regarding microvascular density, no statistically significant correlation was observed between tumor volume and microvascular density measured by either CD34 or CD105. However, the correlation coefficient suggests a positive correlation for MVD-CD34 and a negative correlation for MVD-CD105 ( p = 0.687, p = 0.658) . 2.2.2. Peritumoral Edema Peritumoral edema ranged in size from 7.2 mm to 63.9 mm. The median thickness of peritumoral edema was higher among women (29.1 mm) and in patients aged between 51 and 65 years (30.6 mm). Median values recorded in the temporal lobe were nearly similar to those in the parietal lobe. The right cerebral hemisphere exhibited a higher median value of peritumoral edema compared to the contralateral hemisphere. No statistically significant association was observed between the median values of peritumoral edema and age, sex, tumor localization, or laterality ( and ). Differences in the median values of peritumoral edema based on the presence or absence of the IDH1 mutation were not statistically significant. The median values of peritumoral edema were higher in ATRX mutant cases compared to ATRX wild-type glioblastomas. The median value of peritumoral edema was higher in p53 wild-type glioblastomas and in cases with Ki67 indexes below 5%. Differences in the median value of peritumoral edema concerning the presence or absence of MMP-9 immunoexpression were not significant. No statistically significant association was found between median values of peritumoral edema and ATRX , p53 mutations, or Ki67 index ( and , ). Regarding microvascular density, which was measured by both CD34 and CD105, no statistically significant correlation was observed between peritumoral edema thickness and microvascular density ( p = 0.107, p = 0.563) . Regarding the margin of peritumoral edema, most cases exhibited rounded margins (27/44). Predominantly in the temporal lobe, we observed edema with irregular contours. Irregular margins were more frequently identified in the right hemisphere, whereas the left hemisphere showed more cases of edema with rounded margins. A statistically significant association was observed between laterality and edema shape ( p = 0.03) . Most glioblastomas with IDH1 wild type, p53 wild type, and ATRX wild type exhibited rounded edema margins. The ratio of cases with regular versus irregular edema contours was 2 for GBM with Ki67 index below 20% and 1.12 for those with Ki67 indexes above 20%. MMP-9-positive cases more frequently presented irregular margins (10/24) compared to MMP-9-negative cases (7/20). No statistically significant association was found between age, gender, localization, IDH1 , ATRX , p53 mutations, Ki67 index, and MMP-9 with the characteristics of peritumoral edema shape . The median MVD-CD34 value was higher in GBM with irregular peritumoral edema, whereas MVD-CD105 showed higher median values in GBM associated with regular edema contours. No statistically significant association was observed between peritumoral edema margin characteristics and MVD-CD34 or MVD-CD105 . 2.2.3. Midline Deviation The median value of midline deviation was 6.1 mm. In 15.9% (7/44) of cases, no midline deviation was observed, while in 27.27% (12/44) of cases, the deviation exceeded 10 mm. Midline deviation was greater in males and in patients aged between 50 and 65 years. Higher values were noted in tumors located in the frontal lobe, followed by the parietal lobe. Tumors located in the right cerebral hemisphere were associated with more pronounced midline deviation. A statistically significant association was observed between age groups and midline deviation ( p = 0.006) . IDH1 mutant glioblastomas, p53 mutant glioblastomas, and ATRX mutant glioblastomas exhibited more pronounced midline deviation. Cases with Ki67 indexes above 20% and MMP-9-negative tumors resulted in more severe midline deviation. A statistically significant association was observed between ATRX mutation and midline deviation ( p = 0.035) . There was no statistically significant correlation between midline deviation and the median values of MVD-CD34 or MVD-CD105. However, a negative correlation was noted between the degree of deviation and MVD-CD105, and a positive correlation with MVD-CD34 . Based on the statistical results, this study did not demonstrate a correlation between tumor location and tumor volume, peritumoral edema size, or respective midline deviation . No statistically significant correlation was found between tumor volume and peritumoral edema or tumor volume and midline deviation. However, a positive correlation was observed between the peritumoral edema and the midline deviation . In more than 65% of the studied cases, the maximum tumor diameter was below 5 cm (29/S44), with a median diameter of 4.63 cm. Tumor volume measurements ranged from 8.75 cm 3 to 124.56 cm 3 , with an average tumor volume of 43.17 cm 3 . Regarding tumor size, most glioblastomas with a maximum diameter of less than 5 cm were found in patients over 50 years old (20/44) ( p = 0.43). In terms of gender, glioblastomas smaller than 5 cm were more common in male patients (18/22). Additionally, the proportion of tumors larger than 5 cm was higher among women (50%, 11/22) compared to men (18.18%, 4/22). A statistically significant association was observed between tumor size and patient gender ( p = 0.02) . The median tumor volume decreased with increasing age, with the highest tumor volumes recorded in patients under 50 years old. The median tumor volume was higher in women (42.1 cm 3 ) compared to men (38.6 cm 3 ). However, no statistically significant association was found between median tumor volume and patient age or gender ( p = 0.859, p = 0.477) . The median tumor volume was highest in GBM located in the frontal lobe, followed by the occipital lobe, with the smallest values recorded in the temporal and parietal lobes, which had similar proportions. The median tumor volume was larger in the left hemisphere, with a value of 45.6 cm 3 . No statistically significant correlation was observed between tumor size, median tumor volume, localization, and laterality ( p = 0.95, p = 0.75, p = 0.39, p = 0.77) ( and ). In 65% (26/40) of the cases, we observed IDH1 wild-type glioblastomas with sizes under 5 cm, while 25% (1/4) of the cases were IDH1 mutant glioblastomas larger than 5 cm. The median tumor volume was higher in IDH1 mutant glioblastomas (49.8 cm 3 ). Nearly 61.76% (21/34) of ATRX wild-type cases had sizes below 5 cm. The ratio of ATRX mutant cases with maximum diameters <5 cm versus >5 cm was 8:2 = 4, while in ATRX wild-type GBM cases, the ratio was 21:13 = 1.6. The median tumor volume was approximately equal in both ATRX wild-type and mutant cases. No statistically significant association was found between tumor size, median tumor volume, and IDH1 or ATRX mutations ( p = 0.68, p = 0.28, p = 0.27, p = 0.85) ( and ). Regarding the p53 mutation, we observed that 66.66% (22/33) of the p53 wild-type cases had tumors smaller than 5 cm, while 36.36% (4/11) of the cases showed the p53 mutation associated with tumor sizes greater than 5 cm. The median tumor volume was higher in p53 mutant glioblastomas (48.6 cm 3 ) compared to p53 wild-type cases. A Ki67 index greater than 5% was more frequently observed in tumors smaller than 5 cm (21/29). The median tumor volume was larger in glioblastomas with a Ki67 index above 20%. Concerning the relationship with MMP-9, cases without MMP-9 expression had relatively higher median tumor volumes (43.5 cm 3 vs. 39.8 cm 3 ) compared to glioblastomas with MMP-9 expression. No statistically significant correlation was found between tumor size; median tumor volume; and Ki67 index, p53 mutation, or MMP-9 expression ( and , ). Regarding microvascular density, no statistically significant correlation was observed between tumor volume and microvascular density measured by either CD34 or CD105. However, the correlation coefficient suggests a positive correlation for MVD-CD34 and a negative correlation for MVD-CD105 ( p = 0.687, p = 0.658) . Peritumoral edema ranged in size from 7.2 mm to 63.9 mm. The median thickness of peritumoral edema was higher among women (29.1 mm) and in patients aged between 51 and 65 years (30.6 mm). Median values recorded in the temporal lobe were nearly similar to those in the parietal lobe. The right cerebral hemisphere exhibited a higher median value of peritumoral edema compared to the contralateral hemisphere. No statistically significant association was observed between the median values of peritumoral edema and age, sex, tumor localization, or laterality ( and ). Differences in the median values of peritumoral edema based on the presence or absence of the IDH1 mutation were not statistically significant. The median values of peritumoral edema were higher in ATRX mutant cases compared to ATRX wild-type glioblastomas. The median value of peritumoral edema was higher in p53 wild-type glioblastomas and in cases with Ki67 indexes below 5%. Differences in the median value of peritumoral edema concerning the presence or absence of MMP-9 immunoexpression were not significant. No statistically significant association was found between median values of peritumoral edema and ATRX , p53 mutations, or Ki67 index ( and , ). Regarding microvascular density, which was measured by both CD34 and CD105, no statistically significant correlation was observed between peritumoral edema thickness and microvascular density ( p = 0.107, p = 0.563) . Regarding the margin of peritumoral edema, most cases exhibited rounded margins (27/44). Predominantly in the temporal lobe, we observed edema with irregular contours. Irregular margins were more frequently identified in the right hemisphere, whereas the left hemisphere showed more cases of edema with rounded margins. A statistically significant association was observed between laterality and edema shape ( p = 0.03) . Most glioblastomas with IDH1 wild type, p53 wild type, and ATRX wild type exhibited rounded edema margins. The ratio of cases with regular versus irregular edema contours was 2 for GBM with Ki67 index below 20% and 1.12 for those with Ki67 indexes above 20%. MMP-9-positive cases more frequently presented irregular margins (10/24) compared to MMP-9-negative cases (7/20). No statistically significant association was found between age, gender, localization, IDH1 , ATRX , p53 mutations, Ki67 index, and MMP-9 with the characteristics of peritumoral edema shape . The median MVD-CD34 value was higher in GBM with irregular peritumoral edema, whereas MVD-CD105 showed higher median values in GBM associated with regular edema contours. No statistically significant association was observed between peritumoral edema margin characteristics and MVD-CD34 or MVD-CD105 . The median value of midline deviation was 6.1 mm. In 15.9% (7/44) of cases, no midline deviation was observed, while in 27.27% (12/44) of cases, the deviation exceeded 10 mm. Midline deviation was greater in males and in patients aged between 50 and 65 years. Higher values were noted in tumors located in the frontal lobe, followed by the parietal lobe. Tumors located in the right cerebral hemisphere were associated with more pronounced midline deviation. A statistically significant association was observed between age groups and midline deviation ( p = 0.006) . IDH1 mutant glioblastomas, p53 mutant glioblastomas, and ATRX mutant glioblastomas exhibited more pronounced midline deviation. Cases with Ki67 indexes above 20% and MMP-9-negative tumors resulted in more severe midline deviation. A statistically significant association was observed between ATRX mutation and midline deviation ( p = 0.035) . There was no statistically significant correlation between midline deviation and the median values of MVD-CD34 or MVD-CD105. However, a negative correlation was noted between the degree of deviation and MVD-CD105, and a positive correlation with MVD-CD34 . Based on the statistical results, this study did not demonstrate a correlation between tumor location and tumor volume, peritumoral edema size, or respective midline deviation . No statistically significant correlation was found between tumor volume and peritumoral edema or tumor volume and midline deviation. However, a positive correlation was observed between the peritumoral edema and the midline deviation . Glioblastoma (GBM) is the most common primary malignant brain tumor , typically affecting patients over 60 years of age, and is more frequently diagnosed in men. Locationally, the frontal lobe is the most commonly affected, followed by the temporal lobe . Pathologically, glioblastomas are characterized by increased mitotic activity, aggressive invasive behavior, central necrosis, and pronounced angiogenesis. According to the WHO classification, they are classified as grade IV gliomas . Additionally, radiological parameters obtained through preoperative magnetic resonance imaging (MRI) have been shown to have prognostic value, including necrosis volume and peritumoral edema volume . Glioblastomas are lesions that exhibit heterogeneous signal on T1- or T2-weighted MRI sequences, irregular margins, and central areas of necrosis and/or hemorrhage, along with peritumoral edema. The imaging characteristics of a lesion with irregular infiltrative margins, heterogeneous signal, and peritumoral edema are particularly suggestive of a high-grade diffuse glioma . Recurrences are quite common in glioblastomas, and they are associated with detectable peritumoral edema on T2-FLAIR MRI sequences . The edema appears to be a response to angiogenic factors released in the tumor microenvironment, which increase vascular permeability. As tumor proliferation exceeds the native blood supply, resulting ischemia stimulates additional release of angiogenic factors that promote vascular proliferation . In the current study, MMP-9 expression was present in 54.5% of the glioblastomas examined. While we could not demonstrate a statistically significant association between MMP-9 expression and the clinical–pathological parameters studied in glioblastomas, it is noteworthy that MMP-9 expression was more frequently observed in elderly patients; in glioblastomas located in the right hemisphere; in glioblastomas that were IDH1 wild type, ATRX wild type, and p53 wild type; as well as in tumors with lower MVD-CD34 and MVD-CD105. In terms of the imaging characteristics of GBM, the peritumoral characteristic was associated with irregular margins in approximately 39% of cases. Significant associations were found between more pronounced midline deviation and factors such as age and the presence of ATRX mutation. Irregular margins of peritumoral edema were significantly associated with tumor laterality. The presence of edema in the contralateral hemisphere was significantly correlated with increased MVD-CD105 values. In the study conducted by Roux et al., the average tumor volume was 37.1 cm 3 , whereas in our study, it was 43.17 cm 3 . In total, 63% of the glioblastomas studied by Roux et al. were IDH1 wild type, compared to over 90% in our study . In contrast, the study by Yu et al. reported a lower median tumor volume of 32.4 cm 3 . In the study by Wu et al., peritumoral edema with irregular margins was found in 66.7% of GBM cases, whereas in our study, this proportion was only 38.63%. Tumor size exceeding 50 mm was observed in 57.5% of their cases, while in our study, this was only 34%. They described a statistically significant correlation between the degree of peritumoral edema and the form of the edema, in contrast with our study. No significant correlation was found between the degree of edema and sex, patient age, or tumor size, which is similar to our findings. However, in our study, we observed a significant association between tumor laterality and the shape of edema. Major edema and necrosis on MRI are significant prognostic indicators of shorter overall survival (OS) . The increase in matrix metalloproteinase levels promotes the progression of brain tumors, especially glioblastomas. The tumor microenvironment plays a critical role in shaping the prognosis of glioblastoma, with matrix metalloproteinase 9 (MMP-9) serving as a key regulator within this microenvironment . Clinical and experimental studies have demonstrated a correlation between elevated MMP levels and the invasive nature of brain tumors . The results of the study by Xue et al. showed a significant enhancement in cellular proliferation in glioblastomas with MMP-9 overexpression . Li et al. analyzed the immunoexpression of MMP-9 in gliomas of various grades, finding that MMP-9 expression was correlated with tumor grade. MMP-9 expression was significantly higher in grade IV gliomas (glioblastomas) compared to grade II and III gliomas. Additionally, patients with glioblastomas showing low MMP-9 expression were generally younger and had higher rates of MGMT promoter methylation and IDH1 mutations compared to patients with glioblastomas with high MMP-9 expression. Patients with low MMP-9 expression had a longer overall survival (OS) compared to those with high MMP-9 expression . In the study conducted by Zhang et al., gliomas (including 32 glioblastomas) were characterized by frequently having a tumor diameter greater than 30 mm. They found that high levels of MMP-2 in gliomas were associated with a poorer prognosis. Regarding tumor size and grade, positive MMP-2 expression was closely associated with MRI signal uniformity. Severe peritumoral edema was present in more than half of the cases, similar to our study. They observed a statistically significant association between MMP-2 expression in gliomas and tumor diameter as well as peritumoral edema, but not with sex, age, or tumor grade . The Ki-67 antigen directly reflects the degree of cellular proliferation and is closely associated with tumor progression. High Ki-67 index values are linked to a higher grade of malignancy and poorer prognosis. Lower Ki-67 proliferation index values are correlated with the presence of mutations in the isocitrate dehydrogenase gene. Thus, Ki-67 can be considered a prognostic indicator in glioblastomas. Similarly, the p53 protein is also a marker of poor prognosis; however, Ki-67 has a better correlation with radiomic characteristics of GBM on T2-weighted imaging compared to the p53 protein. Additionally, the Ki-67 marker has shown better predictive value in peritumoral areas . The importance of advanced imaging in studying tumor biology through various imaging sequences, such as rCBV (relative cerebral blood volume) and rOEF (relative oxygen extraction fraction), was explored by Wiestler et al. They found that the maximum relative oxygen extraction fraction (rOEF) was higher in glioblastomas compared to grade 2/3 gliomas according to the WHO classification. The expression of HIF1α, which correlates with rOEF, may reflect the level of hypoxia in tumor tissue and stimulate more pronounced local angiogenesis, a characteristic feature of glioblastomas . Another study demonstrated significant correlations between rCBV (relative cerebral blood volume) and microvascular density in IDH-wildtype glioblastomas (GBM) ( p < 0.001), with rCBV values being 2–2.5 times higher in IDH-wildtype GBM compared to IDH-mutant glioblastomas . Liu et al. found that elevated serum levels of MMP-2 and MMP-9 are correlated with recurrence and show a statistically significant correlation with normal cerebral blood volume (nCBV) and normal cerebral blood flow (nCBF) . Radiogenomics studies the association between molecular phenotypes and specific imaging characteristics to indicate how certain genomic variations might influence the imaging traits of tumors . In the future, non-invasive imaging assessments prior to biopsy or surgery could be an ideal approach for detecting genomic alterations with prognostic significance in GBM. These imaging techniques could also be useful in guiding targeted biopsies . 4.1. Clinical Data This retrospective study included 44 patients diagnosed with glioblastoma at the Department of Pathology, County Emergency Clinical Hospital of Târgu Mureș, between 2014 and 2017. The inclusion criteria were as follows: (1) histopathological confirmation of glioblastoma, without any prior diagnosis or oncological treatment for any type of brain tumor; (2) no history of brain biopsy; (3) availability of tumor tissue in at least two paraffin blocks for the determination of immunoexpression of IDH1-R132H, ATRX, CD34, CD105, MMP-9, Ki67, and p53. The histopathological diagnoses were re-evaluated by a neuropathologist according to the 2016 World Health Organization (WHO) classification of central nervous system tumors. Finally, (4) access to preoperative imaging data obtained through MRI (T1-weighted and contrast-enhanced T1, T2, and T2 FLAIR images) was required. 4.2. Immunohistochemistry Surgical specimens were fixed in formalin, embedded in paraffin, and sectioned at a thickness of 3 μm. The obtained sections underwent standard deparaffinization and rehydration procedures. Endogenous peroxidase activity was blocked using a 10 min treatment with 3% H 2 O 2 . Antigen retrieval was performed by pressure steam boiling for 25 min in a citrate solution (pH 6). The following antibodies were used: mouse monoclonal antibody IDH1R132H, clone IHC132 (BioSB, Santa Barbara, CA, USA), dilution 1:25, incubation 60 min; mouse monoclonal antibody ATRX, clone BSB-108 (BioSB), dilution 1:50, incubation 60 min; rabbit monoclonal antibody MMP-9; rabbit monoclonal antibody CD34, clone EP88 (BioSB), dilution 1:100, incubation 60 min; rabbit monoclonal antibody CD105, clone EP274 (BioSB), dilution 1:200, incubation 60 min; mouse monoclonal antibody Ki67, clone MM1 (Novocastra, Leica Biosystems, Deer Park, IL, USA), dilution 1:150, incubation 60 min; mouse monoclonal antibody p53, clone DO7 (BioSB), dilution 1:800, incubation 60 min. The EnVision Flex/horseradish peroxidase (HRP) secondary system (Agilent - Dako, Santa Clara, CA, USA, 30 min) was used for signal amplification, and 3,3′-diaminobenzidine (DAB) was used as the chromogen for primary antibody detection. The slides were subsequently stained with hematoxylin. 4.3. Slide Evaluation The interpretation of immunohistochemical results was supervised by a neuropathologist. Preliminary examination of the slides was performed using an Olympus BX46 microscope, and the slides were subsequently scanned with a 3DHistech PANORAMIC 1000 scanner (Budapest, Hungary). Cytoplasmic immunoexpression of MMP-9, ranging from yellow-brown to dark brown, was recorded as a positive reaction. For the semi-quantitative evaluation of MMP-9 immunoexpression, we considered: (i) staining intensity [0 points (no staining), 1 point (light staining—light brown), 2 points (moderate staining), and 3 points (marked staining—dark brown)] and (ii) the percentage of stained cells [0 points (no stained cells), 1 point (stained cells <25%), 2 points (stained cells 25–50%), and 3 points (stained cells >50%)]. The total score (0–6 points) was calculated by summing these values. A score of 0–2 points indicated the absence of immunoexpression, while a score of 3–6 points indicated a positive immunoreaction . Immunoexpression of p53 and Ki67 was individually evaluated; the Ki-67 proliferation index was determined as the percentage of stained tumor cells (regardless of intensity) out of 1000 cells. The presence of p53 was assessed using the percentage of immunolabeled cells out of 200 cells across 5 fields. p53 was considered negative (wild type) if the immunostaining was <10% and positive (mutant type) if it was >10% of the examined cells . Microvessel density was determined based on the immunoexpression of CD34 and CD105. In the tumor tissue, four areas with the highest microvessel density were selected, initially with a low-power objective (×40) and subsequently with a high-power objective (×400). For objective quantification of microvascular density in the tumor stroma, the Slideview software (SlideViewer 2.6.0.166179 software together with QuantCenter 2.3.0.143967—by 3DHistech) was used, and the median values of the four analyzed areas were calculated. The immunohistochemical reaction was considered positive if solitary or clustered endothelial cells, whether participating in lumen formation or not, showed a positive reaction . The expression of the IDH1 mutation was determined by evaluating tumor cells that were cytoplasmically stained positive, regardless of staining intensity. Cases where ≥10% of the cells were stained were defined as positive ( IDH1 mutant), while cases where this value did not exceed 10% of tumor cells were considered negative ( IDH1 wild type). In tumor cells, ATRX gene mutations result in the loss of nuclear ATRX immunoexpression ( ATRX loss— ATRX mutant type), whereas ATRX immunoexpression remains preserved in ATRX wild type tumor cells, with endothelial cells serving as the endogenous positive control . 4.4. MRI Sequence Evaluation All patients included in this study underwent a standardized preoperative brain MRI using an OptimaTM MR450w GEM 1.5T scanner (GE Medical System, Waukesha, WI, USA). T2-weighted (T2W), T2-FLAIR, T1W, and T1W-CE images were obtained at the Radiology Department of the Clinical County Emergency Hospital in Târgu Mureș . Using MRI image processing, the tumor volume was determined by manually selecting the region of interest (ROI) with a semi-automatic segmentation method using 3DSlicer 5.6.2 software ( https://www.slicer.org/ ). We also measured the thickness of the peritumoral edema, defined the characteristics of the peritumoral edema margins, assessed the presence of edema in the contralateral hemisphere, and determined the midline shift, similar to protocols described by Wu et al., Palpan et al., and Long et al. . 4.5. Statistical Analysis Descriptive and inferential statistics were performed. The normality of the distribution of continuous variables was tested using the Shapiro–Wilk test. Continuous variables were expressed as medians (25th percentile, 75th percentile), and medians were compared using the Mann–Whitney test. Categorical variables were presented as frequencies, and between-group comparisons were performed using the Chi-square test. A value of p < 0.05 was considered significant. Statistical analyses were conducted using IBM SPSS Statistics 22 software (IBM Corporation, New York, NY, USA). 4.6. Limitations of the Study The primary limitation of our study is the small sample size. The analyzed cohort did not provide postoperative or treatment information, so the overall survival rate or the number of recurrences cannot be assessed. In the future, we will likely expand this study to include other MMPs (such as MMP-2) and their serum levels, and the study will have a larger number of participants. We considered IDH1-R132H to be the most relevant mutant variant within codon R132, while R132S, R132C, R132G, and R132L were less common. 4.7. Ethics Committee This study was approved by the Ethics Committee of the Clinical County Emergency Hospital Târgu Mureș under reference number 24494/16 October 2020. This retrospective study included 44 patients diagnosed with glioblastoma at the Department of Pathology, County Emergency Clinical Hospital of Târgu Mureș, between 2014 and 2017. The inclusion criteria were as follows: (1) histopathological confirmation of glioblastoma, without any prior diagnosis or oncological treatment for any type of brain tumor; (2) no history of brain biopsy; (3) availability of tumor tissue in at least two paraffin blocks for the determination of immunoexpression of IDH1-R132H, ATRX, CD34, CD105, MMP-9, Ki67, and p53. The histopathological diagnoses were re-evaluated by a neuropathologist according to the 2016 World Health Organization (WHO) classification of central nervous system tumors. Finally, (4) access to preoperative imaging data obtained through MRI (T1-weighted and contrast-enhanced T1, T2, and T2 FLAIR images) was required. Surgical specimens were fixed in formalin, embedded in paraffin, and sectioned at a thickness of 3 μm. The obtained sections underwent standard deparaffinization and rehydration procedures. Endogenous peroxidase activity was blocked using a 10 min treatment with 3% H 2 O 2 . Antigen retrieval was performed by pressure steam boiling for 25 min in a citrate solution (pH 6). The following antibodies were used: mouse monoclonal antibody IDH1R132H, clone IHC132 (BioSB, Santa Barbara, CA, USA), dilution 1:25, incubation 60 min; mouse monoclonal antibody ATRX, clone BSB-108 (BioSB), dilution 1:50, incubation 60 min; rabbit monoclonal antibody MMP-9; rabbit monoclonal antibody CD34, clone EP88 (BioSB), dilution 1:100, incubation 60 min; rabbit monoclonal antibody CD105, clone EP274 (BioSB), dilution 1:200, incubation 60 min; mouse monoclonal antibody Ki67, clone MM1 (Novocastra, Leica Biosystems, Deer Park, IL, USA), dilution 1:150, incubation 60 min; mouse monoclonal antibody p53, clone DO7 (BioSB), dilution 1:800, incubation 60 min. The EnVision Flex/horseradish peroxidase (HRP) secondary system (Agilent - Dako, Santa Clara, CA, USA, 30 min) was used for signal amplification, and 3,3′-diaminobenzidine (DAB) was used as the chromogen for primary antibody detection. The slides were subsequently stained with hematoxylin. The interpretation of immunohistochemical results was supervised by a neuropathologist. Preliminary examination of the slides was performed using an Olympus BX46 microscope, and the slides were subsequently scanned with a 3DHistech PANORAMIC 1000 scanner (Budapest, Hungary). Cytoplasmic immunoexpression of MMP-9, ranging from yellow-brown to dark brown, was recorded as a positive reaction. For the semi-quantitative evaluation of MMP-9 immunoexpression, we considered: (i) staining intensity [0 points (no staining), 1 point (light staining—light brown), 2 points (moderate staining), and 3 points (marked staining—dark brown)] and (ii) the percentage of stained cells [0 points (no stained cells), 1 point (stained cells <25%), 2 points (stained cells 25–50%), and 3 points (stained cells >50%)]. The total score (0–6 points) was calculated by summing these values. A score of 0–2 points indicated the absence of immunoexpression, while a score of 3–6 points indicated a positive immunoreaction . Immunoexpression of p53 and Ki67 was individually evaluated; the Ki-67 proliferation index was determined as the percentage of stained tumor cells (regardless of intensity) out of 1000 cells. The presence of p53 was assessed using the percentage of immunolabeled cells out of 200 cells across 5 fields. p53 was considered negative (wild type) if the immunostaining was <10% and positive (mutant type) if it was >10% of the examined cells . Microvessel density was determined based on the immunoexpression of CD34 and CD105. In the tumor tissue, four areas with the highest microvessel density were selected, initially with a low-power objective (×40) and subsequently with a high-power objective (×400). For objective quantification of microvascular density in the tumor stroma, the Slideview software (SlideViewer 2.6.0.166179 software together with QuantCenter 2.3.0.143967—by 3DHistech) was used, and the median values of the four analyzed areas were calculated. The immunohistochemical reaction was considered positive if solitary or clustered endothelial cells, whether participating in lumen formation or not, showed a positive reaction . The expression of the IDH1 mutation was determined by evaluating tumor cells that were cytoplasmically stained positive, regardless of staining intensity. Cases where ≥10% of the cells were stained were defined as positive ( IDH1 mutant), while cases where this value did not exceed 10% of tumor cells were considered negative ( IDH1 wild type). In tumor cells, ATRX gene mutations result in the loss of nuclear ATRX immunoexpression ( ATRX loss— ATRX mutant type), whereas ATRX immunoexpression remains preserved in ATRX wild type tumor cells, with endothelial cells serving as the endogenous positive control . All patients included in this study underwent a standardized preoperative brain MRI using an OptimaTM MR450w GEM 1.5T scanner (GE Medical System, Waukesha, WI, USA). T2-weighted (T2W), T2-FLAIR, T1W, and T1W-CE images were obtained at the Radiology Department of the Clinical County Emergency Hospital in Târgu Mureș . Using MRI image processing, the tumor volume was determined by manually selecting the region of interest (ROI) with a semi-automatic segmentation method using 3DSlicer 5.6.2 software ( https://www.slicer.org/ ). We also measured the thickness of the peritumoral edema, defined the characteristics of the peritumoral edema margins, assessed the presence of edema in the contralateral hemisphere, and determined the midline shift, similar to protocols described by Wu et al., Palpan et al., and Long et al. . Descriptive and inferential statistics were performed. The normality of the distribution of continuous variables was tested using the Shapiro–Wilk test. Continuous variables were expressed as medians (25th percentile, 75th percentile), and medians were compared using the Mann–Whitney test. Categorical variables were presented as frequencies, and between-group comparisons were performed using the Chi-square test. A value of p < 0.05 was considered significant. Statistical analyses were conducted using IBM SPSS Statistics 22 software (IBM Corporation, New York, NY, USA). The primary limitation of our study is the small sample size. The analyzed cohort did not provide postoperative or treatment information, so the overall survival rate or the number of recurrences cannot be assessed. In the future, we will likely expand this study to include other MMPs (such as MMP-2) and their serum levels, and the study will have a larger number of participants. We considered IDH1-R132H to be the most relevant mutant variant within codon R132, while R132S, R132C, R132G, and R132L were less common. This study was approved by the Ethics Committee of the Clinical County Emergency Hospital Târgu Mureș under reference number 24494/16 October 2020. The immunoexpression of MMP-9, which plays a role in remodeling the tumor microenvironment, was present in approximately half of the studied glioblastomas. It is noteworthy that MMP-9 expression was more frequently observed in elderly patients; in glioblastomas located in the right hemisphere; in glioblastomas that were IDH1 wild type, ATRX wild type, and p53 wild type; as well as in tumors with lower MVD-CD34 and MVD-CD105. Comparing the imaging data with the immunohistochemical results, we observed that the median values of MVD-CD34 and MVD-CD105 were higher in cases with extensive peritumoral edema in the contralateral hemisphere. Additionally, ATRX mutations were frequently associated with a more pronounced deviation of the median line. Further studies with larger sample sizes are needed to statistically validate these associations between the imaging and histopathological characteristics of glioblastomas. |
Effect of femtosecond laser-assisted cataract surgery for cataracts after pars plana vitrectomy: a prospective randomized controlled study | 37dedb54-c877-432e-b690-3f8bed561b0e | 11834226 | Surgical Procedures, Operative[mh] | Since pars plana vitrectomy (PPV) was introduced into clinical practice in the 1970s, continuous improvements and refinements in surgical equipment and techniques have made it a common surgical method for treating vitreoretinal diseases . The widespread application of PPV, along with the use of silicone oil or gas tamponade, has made cataracts the most common complication after PPV . Cataract surgery post-PPV is more challenging and more prone to complications. The incision used in conventional phacoemulsification surgery (CPS) is small and well sealed, effectively reducing the risk of low IOP. CPS combined with intraocular lens (IOL) implantation is the preferred surgical method for treating post-PPV cataracts. Since Nagy et al. first applied femtosecond lasers to cataract surgery in 2009, femtosecond laser-assisted cataract surgery (FLACS) has been widely recognized for its precision, safety, and efficiency, making FLACS one of the preferred choices for cataract treatment today . Currently, there are few reports on the use of FLACS for treating post-PPV cataracts in China. This study employs a prospective randomized controlled trial to analyze the outcomes of FLACS and CPS combined with IOL implantation in the treatment of post-PPV cataracts, assessing the effectiveness of FLACS in treating post-PPV cataracts. Study design From January 2023 to June 2024, cataract patients who had undergone post-PPV and met the inclusion criteria of this study at Chengdu Aidi Eye Hospital were selected as the study subjects. All patients who underwent cataract surgery were fully informed about the two surgical methods, FLACS and CPS, including their costs, advantages, and disadvantages. After receiving a complete explanation of the study, they signed informed consent for surgery. The patients were randomly divided into two groups: the FLACS group, in which FLACS combined with IOL implantation was used; and the CPS group, in which the CPS combined with IOL implantation was used. The cost difference due to the surgical method is covered by research funding. This clinical study was approved by the Ethics Committee of Chengdu Aidi Eye Hospital and adheres to the Declaration of Helsinki. All patients provided informed consent and signed the informed consent form. The inclusion criteria were as follows: age ≥ 18 years; previous history of PPV surgery, more than 6 months postoperative; presence of significant lens opacity, best corrected visual acuity (BCVA) below 0.4 logMAR; anterior segment condition meeting the surgical requirements of FLACS; preoperative examination showing no definite macular edema (DME) or retinal redetachment; and ability to actively cooperate to complete femtosecond laser operation. The exclusion criteria for patients were as follows: PPV surgery due to ocular trauma or recurrent retinal detachment; a history of corneal refractive surgery; silicone oil eye post-PPV; presence of significant silicone oil in the anterior chamber; obvious lens ectopia; intraocular pressure (IOP) > 21 mmHg not controlled; and cataract postoperative follow-up shorter than three months. Preoperative examination and preparation The baseline characteristics and demographic details were recorded. All patients underwent routine preoperative examinations, including uncorrected distance visual acuity (UCVA) and BCVA, IOP (Topcon CT-80 A, Japan), slit-lamp and fundus examinations, anterior segment photography, corneal topography, corneal endothelial cell count (Sowin SW-7000, China), A/B ultrasound (Sowin SW-2100, China), IOLMaster 700 (Zeiss, Germany), macular OCT (Spectralis HRA + OCT, Germany), and wide-angle fundus photography (Daytona (P200T), UK). The axial length (AL) and IOL diopter were measured via the IOLMaster 700. If measurement was not possible, A/B ultrasound measurement was used. A foldable intraocular lens was selected, and the target diopter was set between − 0.30 D and − 3.00 D on the basis of the patient’s needs and AL. Surgical methods and medication All surgeries were performed by the same experienced cataract surgeon via the same phacoemulsification machine (Bausch & Lomb BL11110, USA). The control group was subjected to the CPS method. After pupil dilation, patients were placed in a supine position. Proparacaine hydrochloride eye drops (S.A. ALCON-COUVREURN. V, Belgium) were used for topical anesthesia. After the eyelid speculum was applied, the conjunctival sac was disinfected with 0.5% povidone-iodine. A side incision and a 2.2 mm main incision were routinely made. Medical sodium hyaluronate gel (Bloomage Bio, China) was injected into the anterior chamber. If necessary, indocyanine green injection (Dandong Medical Innovation, China) was used for anterior capsule staining. CCC capsulorhexis was performed, with a diameter of approximately 5.2 ∼ 5.5 mm. Thorough hydrodissection and hydrodelineation were conducted. The phacoemulsification machine parameters were personalized on the basis of the axial length and anterior chamber stability. The lens nucleus was emulsified and aspirated, and the lens cortex was irrigated and aspirated. The posterior capsule was polished, and a foldable intraocular lens was implanted into the capsular bag. The medical sodium hyaluronate gel in the anterior chamber and capsular bag was aspirated, and the incisions were watertight. Tobramycin and dexamethasone ophthalmic ointment (ALCON CUSI s.a., Spain) was applied to the conjunctival sac, and the operated eye was bandaged. The FLACS method was used for the observation group. After pupil dilation, the eye was routinely disinfected, and the Catalys ophthalmic femtosecond laser system (Johnson & Johnson Vision, USA) was used to perform capsulorhexis, prechopping, and incision creation. The parameters were set to a capsulorhexis diameter of 5.2 ∼ 5.5 mm, centered relative to the capsule bag. An appropriate prechopping mode was selected on the basis of the hardness of the lens nucleus. After the operation, the eye was reinfected under a surgical microscope. After the eyelid was opened with a speculum, the conjunctival sac was disinfected with 0.5% povidone-iodine. The side and main incisions were separated, and medical sodium hyaluronate gel was injected into the anterior chamber. The cut anterior capsule membrane was removed via a manual capsulorhexis technique. For incomplete capsulorhexis, a secondary capsulorhexis was performed. The remaining steps were the same as those for the control group. Perioperative medication Preoperative levofloxacin eye drops (Santen-China, China) and pranoprofen eye drops (Senju Pharmaceutical Co., Ltd., Japan) were used. Postoperatively, tobramycin and dexamethasone eye drops were tapered weekly and combined with pranoprofen eye drops for 1 month. If the postoperative IOP was ≥ 25 mmHg, fluid release from an auxiliary incision or combined with intraocular pressure-lowering drugs was used to control IOP. The patients were subsequently followed up on postoperative days 1, 1 week, 1 month, and 3 months. Observation indicators All preoperative and postoperative examinations were performed by the same personnel using the same equipment. The following preoperative data were collected from all patients: age, sex, lens nucleus hardness (using the Emery-Little classification), BCVA, IOP, corneal endothelial cell density (ECD), AL, intraoperative complications, average phacoemulsification energy (AVE), effective phacoemulsification time (EPT), BCVA at 1 day, 1 week, 1 month, and 3 months postoperatively, and IOP and ECD 3 months postoperatively. BCVA after Nd: YAG laser (LPULSA SYL-9000, Taiwan, China) treatment was included in the statistics for those with posterior capsule plaque or posterior capsular opacification (PCO) at least 1 month postoperatively. Statistical methods Statistical analysis was carried out via the Statistical Package for Social Sciences (SPSS 19.0 version). Cross-tabulation chi-square tests were used for sex, lens nucleus hardness grading, primary disease conditions, and intraoperative and postoperative complications. Measurement data that met the homogeneity of variance criteria are expressed as the means ± standard deviations (means ± SDs). Independent samples t-tests were used for age, ECD, IOP, AL, AVE, and EPT. The preoperative and postoperative BCVAs of the two groups were compared via the Wilcoxon rank-sum test. Statistical significance was set at P < 0.05 for all tests. From January 2023 to June 2024, cataract patients who had undergone post-PPV and met the inclusion criteria of this study at Chengdu Aidi Eye Hospital were selected as the study subjects. All patients who underwent cataract surgery were fully informed about the two surgical methods, FLACS and CPS, including their costs, advantages, and disadvantages. After receiving a complete explanation of the study, they signed informed consent for surgery. The patients were randomly divided into two groups: the FLACS group, in which FLACS combined with IOL implantation was used; and the CPS group, in which the CPS combined with IOL implantation was used. The cost difference due to the surgical method is covered by research funding. This clinical study was approved by the Ethics Committee of Chengdu Aidi Eye Hospital and adheres to the Declaration of Helsinki. All patients provided informed consent and signed the informed consent form. The inclusion criteria were as follows: age ≥ 18 years; previous history of PPV surgery, more than 6 months postoperative; presence of significant lens opacity, best corrected visual acuity (BCVA) below 0.4 logMAR; anterior segment condition meeting the surgical requirements of FLACS; preoperative examination showing no definite macular edema (DME) or retinal redetachment; and ability to actively cooperate to complete femtosecond laser operation. The exclusion criteria for patients were as follows: PPV surgery due to ocular trauma or recurrent retinal detachment; a history of corneal refractive surgery; silicone oil eye post-PPV; presence of significant silicone oil in the anterior chamber; obvious lens ectopia; intraocular pressure (IOP) > 21 mmHg not controlled; and cataract postoperative follow-up shorter than three months. The baseline characteristics and demographic details were recorded. All patients underwent routine preoperative examinations, including uncorrected distance visual acuity (UCVA) and BCVA, IOP (Topcon CT-80 A, Japan), slit-lamp and fundus examinations, anterior segment photography, corneal topography, corneal endothelial cell count (Sowin SW-7000, China), A/B ultrasound (Sowin SW-2100, China), IOLMaster 700 (Zeiss, Germany), macular OCT (Spectralis HRA + OCT, Germany), and wide-angle fundus photography (Daytona (P200T), UK). The axial length (AL) and IOL diopter were measured via the IOLMaster 700. If measurement was not possible, A/B ultrasound measurement was used. A foldable intraocular lens was selected, and the target diopter was set between − 0.30 D and − 3.00 D on the basis of the patient’s needs and AL. All surgeries were performed by the same experienced cataract surgeon via the same phacoemulsification machine (Bausch & Lomb BL11110, USA). The control group was subjected to the CPS method. After pupil dilation, patients were placed in a supine position. Proparacaine hydrochloride eye drops (S.A. ALCON-COUVREURN. V, Belgium) were used for topical anesthesia. After the eyelid speculum was applied, the conjunctival sac was disinfected with 0.5% povidone-iodine. A side incision and a 2.2 mm main incision were routinely made. Medical sodium hyaluronate gel (Bloomage Bio, China) was injected into the anterior chamber. If necessary, indocyanine green injection (Dandong Medical Innovation, China) was used for anterior capsule staining. CCC capsulorhexis was performed, with a diameter of approximately 5.2 ∼ 5.5 mm. Thorough hydrodissection and hydrodelineation were conducted. The phacoemulsification machine parameters were personalized on the basis of the axial length and anterior chamber stability. The lens nucleus was emulsified and aspirated, and the lens cortex was irrigated and aspirated. The posterior capsule was polished, and a foldable intraocular lens was implanted into the capsular bag. The medical sodium hyaluronate gel in the anterior chamber and capsular bag was aspirated, and the incisions were watertight. Tobramycin and dexamethasone ophthalmic ointment (ALCON CUSI s.a., Spain) was applied to the conjunctival sac, and the operated eye was bandaged. The FLACS method was used for the observation group. After pupil dilation, the eye was routinely disinfected, and the Catalys ophthalmic femtosecond laser system (Johnson & Johnson Vision, USA) was used to perform capsulorhexis, prechopping, and incision creation. The parameters were set to a capsulorhexis diameter of 5.2 ∼ 5.5 mm, centered relative to the capsule bag. An appropriate prechopping mode was selected on the basis of the hardness of the lens nucleus. After the operation, the eye was reinfected under a surgical microscope. After the eyelid was opened with a speculum, the conjunctival sac was disinfected with 0.5% povidone-iodine. The side and main incisions were separated, and medical sodium hyaluronate gel was injected into the anterior chamber. The cut anterior capsule membrane was removed via a manual capsulorhexis technique. For incomplete capsulorhexis, a secondary capsulorhexis was performed. The remaining steps were the same as those for the control group. Perioperative medication Preoperative levofloxacin eye drops (Santen-China, China) and pranoprofen eye drops (Senju Pharmaceutical Co., Ltd., Japan) were used. Postoperatively, tobramycin and dexamethasone eye drops were tapered weekly and combined with pranoprofen eye drops for 1 month. If the postoperative IOP was ≥ 25 mmHg, fluid release from an auxiliary incision or combined with intraocular pressure-lowering drugs was used to control IOP. The patients were subsequently followed up on postoperative days 1, 1 week, 1 month, and 3 months. Preoperative levofloxacin eye drops (Santen-China, China) and pranoprofen eye drops (Senju Pharmaceutical Co., Ltd., Japan) were used. Postoperatively, tobramycin and dexamethasone eye drops were tapered weekly and combined with pranoprofen eye drops for 1 month. If the postoperative IOP was ≥ 25 mmHg, fluid release from an auxiliary incision or combined with intraocular pressure-lowering drugs was used to control IOP. The patients were subsequently followed up on postoperative days 1, 1 week, 1 month, and 3 months. All preoperative and postoperative examinations were performed by the same personnel using the same equipment. The following preoperative data were collected from all patients: age, sex, lens nucleus hardness (using the Emery-Little classification), BCVA, IOP, corneal endothelial cell density (ECD), AL, intraoperative complications, average phacoemulsification energy (AVE), effective phacoemulsification time (EPT), BCVA at 1 day, 1 week, 1 month, and 3 months postoperatively, and IOP and ECD 3 months postoperatively. BCVA after Nd: YAG laser (LPULSA SYL-9000, Taiwan, China) treatment was included in the statistics for those with posterior capsule plaque or posterior capsular opacification (PCO) at least 1 month postoperatively. Statistical analysis was carried out via the Statistical Package for Social Sciences (SPSS 19.0 version). Cross-tabulation chi-square tests were used for sex, lens nucleus hardness grading, primary disease conditions, and intraoperative and postoperative complications. Measurement data that met the homogeneity of variance criteria are expressed as the means ± standard deviations (means ± SDs). Independent samples t-tests were used for age, ECD, IOP, AL, AVE, and EPT. The preoperative and postoperative BCVAs of the two groups were compared via the Wilcoxon rank-sum test. Statistical significance was set at P < 0.05 for all tests. Basic information There were 115 patients (115 eyes) who underwent cataract surgery after PPV, with 23 eyes excluded from the study. Among them, 8 eyes had significant silicone oil residue in the anterior chamber, 6 eyes were silicone oil-filled, 5 eyes had posterior synechia or pupil dilation < 6 mm, 3 eyes were lost to follow-up postoperatively, and 1 eye underwent ECCE surgery due to abnormal zonules during the operation. Finally, 92 patients (92 eyes) were included in the analysis, with 47 eyes in the FLACS group and 45 eyes in the CPS group. The basic information of the two groups is shown in Table . There were no statistically significant differences between the FLACS group and the CPS group in terms of age, sex, lens nucleus hardness grade, preoperative BCVA, ECD, AL, or IOP ( P > 0.05). Among the primary diseases leading to PPV surgery in the two patient groups, the FLACS group had 36 eyes with rhegmatogenous retinal detachment (RRD), 7 eyes with epiretinal membrane, 3 eyes with diabetic retinopathy, and 1 eye with macular hole. In the CPS group, the primary diseases included 33 eyes with RRD, 7 eyes with diabetic retinopathy, 4 eyes with macular hole, and 1 eye with epiretinal membrane. There was a statistically significant difference in the primary diseases between the two groups ( P < 0.05). Surgical outcomes For those with ALs ≥ 26 mm who met the criteria, capsular tension rings (CTRs) were routinely implanted, and IOLs were implanted within the capsular bag in all eyes. In the FLACS group, incomplete prechopping and capsulorhexis occurred in 3 eyes (3/47, 6.38%), and incomplete lens dislocation occurred in 1 eye (1/47, 2.13%). In the CPS group, incomplete lens dislocation occurred in 2 eyes (2/45, 4.44%), and anterior capsule tears occurred in 1 eye (1/45, 2.22%). There was no significant difference in the incidence of intraoperative complications between the two groups ( P > 0.05) (Table ). Neither group experienced severe intraoperative complications, such as nucleus drop, expulsive choroidal hemorrhage, or retinal detachment. Postoperative conditions and management On the first day after surgery, corneal edema occurred in 11 eyes (11/47, 23.04%) in the FLACS group and 20 eyes (20/45, 44.44%) in the CPS group, with a statistically significant difference ( P < 0.05). The frequency of tobramycin and dexamethasone eye drops was increased, and for eyes with intraocular pressure ≥ 25 mmHg, fluid release from an auxiliary incision or combined treatment with intraocular pressure-lowering drugs was administered. The corneal edema subsided within 1 week postoperatively. Intraoperative findings of posterior capsule plaques or posterior capsule opacification during follow-up were observed in 16 eyes (16/47, 34.04%) in the FLACS group and 15 eyes (15/45, 33.33%) in the CPS group, with no statistically significant difference ( P > 0.05) (Table ). For posterior capsule plaques or PCOs affecting vision, Nd: YAG laser posterior capsulotomy was performed 1 month postoperatively. AVE and EPT The AVEs and EPTs were recorded at the end of each surgery. Compared with that in the CPS group, the AVE in the FLACS group was 20.64% lower, and the EPT was 35.91% lower; these differences were statistically significant ( P < 0.05) (Table ). Postoperative IOP and ECD In the early postoperative period, some patients experienced corneal edema and an uneven distribution of corneal endothelial cells, leading to inaccurate IOP and mean ECD results. Therefore, the results of IOP and the mean ECD 3 months postoperatively were compared. Compared with the preoperative values, both the FLACS and CPS groups presented varying degrees of IOP reduction, with no statistically significant difference in IOP between the two groups ( P > 0.05). The mean ECD of the FLACS group 3 months postoperatively was greater than that of the CPS group, with the mean endothelial cell loss (ECL) being significantly lower than that of the CPS group ( P < 0.01) (Table ). BCVA The distributions of BCVA during the follow-up period for the FLACS group and the CPS group are shown in Table . Both groups showed varying degrees of improvement compared with before surgery. On the first day after surgery, the FLACS group was significantly superior to the CPS group ( P < 0.05), whereas at 1 week, 1 month, and 3 months after surgery, there was no statistically significant difference between the two groups (all P > 0.05). There were 115 patients (115 eyes) who underwent cataract surgery after PPV, with 23 eyes excluded from the study. Among them, 8 eyes had significant silicone oil residue in the anterior chamber, 6 eyes were silicone oil-filled, 5 eyes had posterior synechia or pupil dilation < 6 mm, 3 eyes were lost to follow-up postoperatively, and 1 eye underwent ECCE surgery due to abnormal zonules during the operation. Finally, 92 patients (92 eyes) were included in the analysis, with 47 eyes in the FLACS group and 45 eyes in the CPS group. The basic information of the two groups is shown in Table . There were no statistically significant differences between the FLACS group and the CPS group in terms of age, sex, lens nucleus hardness grade, preoperative BCVA, ECD, AL, or IOP ( P > 0.05). Among the primary diseases leading to PPV surgery in the two patient groups, the FLACS group had 36 eyes with rhegmatogenous retinal detachment (RRD), 7 eyes with epiretinal membrane, 3 eyes with diabetic retinopathy, and 1 eye with macular hole. In the CPS group, the primary diseases included 33 eyes with RRD, 7 eyes with diabetic retinopathy, 4 eyes with macular hole, and 1 eye with epiretinal membrane. There was a statistically significant difference in the primary diseases between the two groups ( P < 0.05). For those with ALs ≥ 26 mm who met the criteria, capsular tension rings (CTRs) were routinely implanted, and IOLs were implanted within the capsular bag in all eyes. In the FLACS group, incomplete prechopping and capsulorhexis occurred in 3 eyes (3/47, 6.38%), and incomplete lens dislocation occurred in 1 eye (1/47, 2.13%). In the CPS group, incomplete lens dislocation occurred in 2 eyes (2/45, 4.44%), and anterior capsule tears occurred in 1 eye (1/45, 2.22%). There was no significant difference in the incidence of intraoperative complications between the two groups ( P > 0.05) (Table ). Neither group experienced severe intraoperative complications, such as nucleus drop, expulsive choroidal hemorrhage, or retinal detachment. On the first day after surgery, corneal edema occurred in 11 eyes (11/47, 23.04%) in the FLACS group and 20 eyes (20/45, 44.44%) in the CPS group, with a statistically significant difference ( P < 0.05). The frequency of tobramycin and dexamethasone eye drops was increased, and for eyes with intraocular pressure ≥ 25 mmHg, fluid release from an auxiliary incision or combined treatment with intraocular pressure-lowering drugs was administered. The corneal edema subsided within 1 week postoperatively. Intraoperative findings of posterior capsule plaques or posterior capsule opacification during follow-up were observed in 16 eyes (16/47, 34.04%) in the FLACS group and 15 eyes (15/45, 33.33%) in the CPS group, with no statistically significant difference ( P > 0.05) (Table ). For posterior capsule plaques or PCOs affecting vision, Nd: YAG laser posterior capsulotomy was performed 1 month postoperatively. The AVEs and EPTs were recorded at the end of each surgery. Compared with that in the CPS group, the AVE in the FLACS group was 20.64% lower, and the EPT was 35.91% lower; these differences were statistically significant ( P < 0.05) (Table ). In the early postoperative period, some patients experienced corneal edema and an uneven distribution of corneal endothelial cells, leading to inaccurate IOP and mean ECD results. Therefore, the results of IOP and the mean ECD 3 months postoperatively were compared. Compared with the preoperative values, both the FLACS and CPS groups presented varying degrees of IOP reduction, with no statistically significant difference in IOP between the two groups ( P > 0.05). The mean ECD of the FLACS group 3 months postoperatively was greater than that of the CPS group, with the mean endothelial cell loss (ECL) being significantly lower than that of the CPS group ( P < 0.01) (Table ). The distributions of BCVA during the follow-up period for the FLACS group and the CPS group are shown in Table . Both groups showed varying degrees of improvement compared with before surgery. On the first day after surgery, the FLACS group was significantly superior to the CPS group ( P < 0.05), whereas at 1 week, 1 month, and 3 months after surgery, there was no statistically significant difference between the two groups (all P > 0.05). The pathogenesis of cataract formation after PPV surgery is still unclear and may be related to multiple inducing factors . The most common type of cataract after PPV surgery is nuclear sclerosis, with an incidence rate of 81% within 6 months and as high as 100% within 2 years . In this study, all lenses in the eyes showed varying degrees of nuclear sclerosis, with 75 eyes (75/92, 81.52%) having a lens nucleus graded as Emery-Little grade III or above and 38 eyes having hard nuclear cataracts at Grade IV or above, accounting for as many as 41.30% (38/92). In this study, the proportion of post-PPV hard nuclear cataracts was relatively high, possibly because (1) post-PPV visual function decreased to varying degrees, causing patients to focus more on retinal recovery and neglecting the impact of lens opacification; (2) poor compliance in some patients led to a prolonged interval before silicone oil removal, and prolonged silicone oil tamponade accelerated lens nucleus hardening; and (3) patients with insufficient postoperative follow-up time failed to detect lens clouding at an early stage, and by the time of long-term consultation, lens nucleus hardening was already very severe. A newly discovered special suspensory ligament inserts from the posterior insertion area of the vitreous zonule through the ciliary body pars plana and connects to the equatorial posterior capsule of the lens . Together with the vitreous zonule and the traditional lens suspensory ligament, it forms the lens suspensory ligament system, which stabilizes and regulates lens movement. The surgical channel of the PPV passes through the ciliary body pars plana, and in the process of clearing the anterior vitreous, it inevitably leads to varying degrees of damage to the lens suspensory ligament system. The abnormal anatomical structure around the lens post-PPV, as well as lens nucleus hardening, increases the difficulty of cataract surgery. The anterior chamber deepens after post-PPV cataracts, and the poor fundus red reflex in hard nuclear cataracts increases the difficulty of capsulorhexis. Compared with manual capsulorhexis, FLACS yields better capsulorhexis results and better postoperative IOL centration . The degree of lens nucleus sclerosis is related to the energy of phacoemulsification and the operation time, and FLACS, through prechopping, can effectively reduce the use of ultrasound energy. Abell RG et al. reported that in routine cataract surgery, femtosecond laser prechopping treatment can significantly reduce EPT, potentially achieving 0 EPT. Asif MI et al. analyzed 60 eyes with LOCS III nuclear grading ≤ NO4 cataracts, and the average CDE in the FLACS group was 48.8% lower than that in the CPS group ( P = 0.012), with the average ECD being better than that in the CPS group ( P = 0.001). Cai L et al. analyzed 86 cataract eyes (45 in the FLACS group and 41 in the CPS group), including 43 with hard nuclei. The results revealed that FLACS patients with hard nuclei had a lower average CDE, shorter ultrasound time, lower average ECL, and faster recovery of central corneal thickness than CPS patients with hard nuclei. Notably, the proportion of silicone oil filling after PPV surgery is relatively high. Emulsified silicone oil can enter the anterior chamber through the zonular space. During manual capsulorhexis, viscoelastic agents can be used to push away the silicone oil in the anterior chamber, but the silicone oil in the anterior chamber affects the penetration of the femtosecond laser, reducing its effectiveness. This study rigorously excluded eyes with significant emulsified silicone oil residue in the anterior chamber. However, in the FLACS group, there were still three eyes with small amounts of emulsified silicone oil in the angle, partially covering the pupillary area during femtosecond laser treatment, resulting in incomplete prechopping and incomplete capsulorhexis. In the FLACS group, 3 eyes (3/47, 6.38%) had incomplete capsulorhexis, all of which were less than one clock hour in range. After manual capsulorhexis completion, no anterior capsule tears occurred. In the CPS group, 1 eye experienced anterior capsule tear (1/45, 2.22%). Using FLACS did not increase the risk of capsulorhexis complications. Incomplete capsulorhexis caused by silicone oil is often accompanied by incomplete prechopping. The 3 eyes with incomplete prechopping and capsulorhexis in the FLACS group were all the same eye, which is also a unique intraoperative complication of FLACS. Although there were 3 eyes with incomplete prechopping in the FLACS group, after prechopping, the AVE and EPT in the FLACS group were still significantly lower than those in the CPS group ( P < 0.05). Compared with that in the CPS group, the AVE in the FLACS group was reduced by 20.64%, and the EPT was reduced by 35.91%, indicating that when post-PPV cataracts are treated with FLACS, even in the presence of incomplete prechopping, ultrasound time and ultrasound energy can still be significantly reduced. In the FLACS group, there was one case of incomplete lens dislocation, and in the CPS group, there were two cases. After implantation of the capsular tension ring, successful in-the-bag IOL implantation was achieved in all patients. There was no significant difference in intraoperative complications between the two groups ( P > 0.05). No severe complications, such as expulsive choroidal hemorrhage or retinal detachment, occurred intraoperatively in either group. Ultrasound time and energy are important factors affecting ECD. When FLACS reduces the ultrasound time and energy, it can effectively reduce the loss of corneal endothelial cells and alleviate corneal edema. Abell RG et al. reported that pretreatment with a femtosecond laser can reduce the average ECL by 36.1%. Krarup T et al. followed patients with one eye who underwent FLACS surgery and the other eye who underwent CPS surgery for 6 months. Compared with the CPS group, the FLACS group presented a 30% reduction in average ECL on day 40 and a 21% reduction on day 180. In this study, the proportion of hard nuclear cataracts with Emery grade IV and above was high (38/92, 41.30%), with prolonged ultrasound time and higher ultrasound energy, which is one of the reasons for the high proportion of postoperative corneal edema (31/92, 33.70%). On the first day after surgery, corneal edema occurred in 11 eyes (11/47, 23.40%) in the FLACS group and in 20 eyes (20/45, 44.44%) in the CPS group, which was a significant difference ( P < 0.05). Three months postoperatively, the ECD in the FLACS group was greater than that in the CPS group, with an average ECL loss of less than that in the CPS group ( P < 0.01), resulting in 46.13% less loss than that in the CPS group. During surgery, we combined the use of dispersive and cohesive viscoelastics, employing the soft shell technique to protect corneal endothelial cells, further reducing the trauma caused by ultrasound energy to corneal endothelial cells. Noor NA et al. utilized the anterior capsule flap created in FLACS to provide additional protection for corneal endothelial cells, especially in eyes with a lower ECD, which is beneficial for reducing the average ECL. In clinical practice, various techniques can be combined to reduce the impact of ultrasound energy on corneal endothelial cells. The good capsulorhexis results of FLACS can reduce IOL tilt and decentration, decrease aberrations, and improve postoperative visual acuity and quality . Chee SP et al. reported that the more precise size, roundness, and centration of capsulorhexis achieved with FLACS result in superior postoperative UAVA compared with CPS. Asif MI et al. reported that in a study focusing on cataracts with LOCS III nuclear grading ≤ NO4, there was no difference in postoperative corrected visual acuity between the FLACS group and the CPS group. In a prospective study by Ewe SY et al. , patients with better economic means underwent FLACS surgery earlier, resulting in baseline BCVA for FLACS being superior to that of CPS, and BCVA at 6 months postoperative was also superior to that of CPS. Some studies have reported a greater chance of early posterior capsule opacification after FLACS . We observed that in some patients, the lens posterior capsule had already developed plaque-like opacification post-PPV. These opacifications may not be completely removed during cataract surgery. Therefore, for posterior capsule plaques or opacifications affecting vision, we performed Nd: YAG laser capsulotomy at least one month after cataract surgery and used the posttreatment BCVA for statistical analysis. In this study, the postoperative BCVA of both groups improved to varying degrees compared with the preoperative BCVA. On the first day after surgery, the BCVA of the FLACS group was better than that of the CPS group ( P < 0.05), which was related to the different degrees of corneal edema in the two groups of patients in the early postoperative period. As the degree of corneal edema subsided, the visual acuity of both groups gradually improved, and there was no significant difference in BCVA between the two groups at one week, one month, or three months postoperatively ( P > 0.05), which is consistent with the results of early studies . FLACS reduces the difficulty of cataract surgery after PPV and minimizes the impact on corneal endothelial cells, benefiting both surgeons and patients. However, compared with CPS, FLACS is not cost-effective . In our study, the cost of FLACS was almost twice that of CPS, and FLACS did not provide superior final visual acuity. Patients cannot directly perceive the advantages of FLACS, emphasizing the need for clinicians to select the appropriate surgical method on the basis of patients’ needs and practical situations. This study has several limitations, including a small sample size and relatively short follow-up duration, which precludes the evaluation of long-term outcomes and complications of FLACS for cataracts after PPV. In summary, compared with CPS, FLACS can effectively reduce ultrasound energy and time, providing patients with faster visual improvement and promoting early postoperative recovery. FLACS is safe and effective in treating post-PPV cataracts. Although there is no difference in final vision, it can be considered a new option for treating post-PPV cataracts. The presence of residual silicone oil in the anterior chamber after PPV surgery may lead to specific outcomes in FLACS. Although it may not affect the surgical results, it is still worth noting. |
Integrated visual and text-based analysis of ophthalmology clinical cases using a large language model | 23a2c36a-1586-4a76-b416-7b6c94f900e9 | 11811221 | Ophthalmology[mh] | Artificial intelligence (AI) has made remarkable advancements in healthcare, and in ophthalmology, where images are central to diagnosis. Deep learning algorithms have demonstrated promising performance in analyzing ocular images . For example, AI applications have been developed for the screening of strabismus , diagnosis of keratoconus and ectatic corneal disease , and cataract classification . Most past applications were trained to analyze only one type of image, mostly retinal or optic nerve-based images , , and did not integrate clinical context. Large Language Models (LLMs) such as GPT-4 by OpenAI have shown impressive abilities in free-text analysis and generation across different healthcare tasks , , and in ophthalmology specifically – . Patient history and complaints, have been established as a major component of patient diagnosis , . However, visual data such as physical examination, imaging tests and pathology, are often critical in patient evaluation . Ophthalmology in particular relies significantly on the integration of the clinical context, physical examination and imaging . Multimodal models, such as GPT-4 with vision capabilities (GPT-4 V), represent a paradigm shift by enabling simultaneous analysis of visual and textual data. The aim of our study was to evaluate the performance of GPT-4 V in diagnosing ocular conditions based on external eye photographs, both with and without supplemental clinical context. Study design A Sheba Medical Center institutional board approval (IRB) was granted to this study (0143-23-SMC(. This was a retrospective study assessing GPT-4 multimodal (GPT-4 V, with image analysis capability) diagnostic performance on ocular images against two non-ophthalmologist physicians, without and with supplemental clinical context. The GPT-4 V model was accessed on November 3rd. Two radiology residents were each presented with the same subset of cases to provide diagnoses. Data collection and diagnostic procedure A series of 40 anonymized ocular images was curated, representing a spectrum of ocular conditions (Table ). Retinal or optic nerve pathologies were not included. The photos were obtained by a phone camera or a slit lamp microscope camera. We included only external images of the eyes. We did not include fundus photography or other ocular imaging technologies such as ocular coherence tomography (OCT), ocular ultrasound, fluorescein angiography or radiology images in this analysis. Cases were selected by a board certified ophthalmologist, to represent various pathologies and match the level of ophthalmology residents. Each participant (GPT-4 V and human physicians) separately received these images in two sequences: initially without and subsequently with added clinical context. Both GPT-4 and the physicians were asked to render a diagnosis for the images first without any clinical context and then with additional clinical information. Clinical context included age, symptoms and relevant medical history. All interactions with GPT-4 V were conducted through the OpenAI web interface. Each inquiry was initiated in a distinct instance to ensure independence of responses. The output without clinical context and the output with clinical context were tested in separate dialogue instances to ensure independence of responses. Specific prompts used were: Outcome measures The primary metric was the accuracy of diagnoses, expressed as a percentage of correct identifications. A qualitative analysis of GPT-4 V answers was also performed. All diagnostic responses from GPT-4 V and the participating physicians were evaluated for accuracy in consensus by two board-certified ophthalmologists. An additional evaluation based on class agreement analysis by anatomical region was performed. For this analysis, each case was categorized into a specific anatomical region, and the agreement with the ground truth, as determined by the board-certified ophthalmologists, was calculated both with and without clinical context. Statistical analysis Statistical computations were conducted using SPSS software for windows version 24.0 by IBM. A Fisher’s exact test was utilized to contrast the performance differences between GPT-4 V and the physicians, and to compare overall accuracy with and without context. We considered P-values less than 0.05 as indicative of statistical significance. A Sheba Medical Center institutional board approval (IRB) was granted to this study (0143-23-SMC(. This was a retrospective study assessing GPT-4 multimodal (GPT-4 V, with image analysis capability) diagnostic performance on ocular images against two non-ophthalmologist physicians, without and with supplemental clinical context. The GPT-4 V model was accessed on November 3rd. Two radiology residents were each presented with the same subset of cases to provide diagnoses. A series of 40 anonymized ocular images was curated, representing a spectrum of ocular conditions (Table ). Retinal or optic nerve pathologies were not included. The photos were obtained by a phone camera or a slit lamp microscope camera. We included only external images of the eyes. We did not include fundus photography or other ocular imaging technologies such as ocular coherence tomography (OCT), ocular ultrasound, fluorescein angiography or radiology images in this analysis. Cases were selected by a board certified ophthalmologist, to represent various pathologies and match the level of ophthalmology residents. Each participant (GPT-4 V and human physicians) separately received these images in two sequences: initially without and subsequently with added clinical context. Both GPT-4 and the physicians were asked to render a diagnosis for the images first without any clinical context and then with additional clinical information. Clinical context included age, symptoms and relevant medical history. All interactions with GPT-4 V were conducted through the OpenAI web interface. Each inquiry was initiated in a distinct instance to ensure independence of responses. The output without clinical context and the output with clinical context were tested in separate dialogue instances to ensure independence of responses. Specific prompts used were: The primary metric was the accuracy of diagnoses, expressed as a percentage of correct identifications. A qualitative analysis of GPT-4 V answers was also performed. All diagnostic responses from GPT-4 V and the participating physicians were evaluated for accuracy in consensus by two board-certified ophthalmologists. An additional evaluation based on class agreement analysis by anatomical region was performed. For this analysis, each case was categorized into a specific anatomical region, and the agreement with the ground truth, as determined by the board-certified ophthalmologists, was calculated both with and without clinical context. Statistical computations were conducted using SPSS software for windows version 24.0 by IBM. A Fisher’s exact test was utilized to contrast the performance differences between GPT-4 V and the physicians, and to compare overall accuracy with and without context. We considered P-values less than 0.05 as indicative of statistical significance. Study cohort and pathological variance The study cohort comprised a diverse array of 40 ocular conditions presented to the AI model and non-ophthalmologist physicians for diagnosis. Mean age of patients included was 54.4 ± 23.2 years. The pathologies included are detailed in Table . Diagnostic accuracy without and with clinical context The diagnostic accuracy of GPT-4 V based on images alone was 47.5% (19/40). In comparison, Physician 1 achieved an accuracy of 60.0% (24/40) under the same conditions. Physician 2 correctly identified 57.5% (23/40) of cases (Table ). When clinical context was included, GPT-4 V’s diagnostic accuracy improved to 67.5% (27/40). Physician 1’s accuracy was 72.5% (29/40), and Physician 2’s accuracy was 67.5% (27/40). There was no statistically significant difference between GPT-4 and physicians’ diagnostic accuracy (Table ). Overall, for all study readers, adding context improved accuracy, as can be seen in Fig. ( p = 0.033). Qualitative analysis of GPT responses Cases in which GPT-4 V was initially wrong, but when provided with clinical context correctly altered the diagnoses included: nevus of Ota, dacryocystitis, Argentinean flag, herpes zoster pseudo-dendrite, thyroid eye disease, iris nevus, cornea foreign body, and ocular perforation. Context-enriched answers showed deeper diagnostic reasoning, and blending clinical history with visual findings. Example cases are detailed in Figs. and . When analyzing the results per anatomical region, regions that are readily visible in external photographs, such as the cornea and eyelids had high class agreement. In contrast, regions that can only be indirectly inferred from external images, such as the lacrimal system and orbit had relatively lower class agreement. A detailed breakdown of class agreement by anatomical region is provided in Table . The study cohort comprised a diverse array of 40 ocular conditions presented to the AI model and non-ophthalmologist physicians for diagnosis. Mean age of patients included was 54.4 ± 23.2 years. The pathologies included are detailed in Table . The diagnostic accuracy of GPT-4 V based on images alone was 47.5% (19/40). In comparison, Physician 1 achieved an accuracy of 60.0% (24/40) under the same conditions. Physician 2 correctly identified 57.5% (23/40) of cases (Table ). When clinical context was included, GPT-4 V’s diagnostic accuracy improved to 67.5% (27/40). Physician 1’s accuracy was 72.5% (29/40), and Physician 2’s accuracy was 67.5% (27/40). There was no statistically significant difference between GPT-4 and physicians’ diagnostic accuracy (Table ). Overall, for all study readers, adding context improved accuracy, as can be seen in Fig. ( p = 0.033). Cases in which GPT-4 V was initially wrong, but when provided with clinical context correctly altered the diagnoses included: nevus of Ota, dacryocystitis, Argentinean flag, herpes zoster pseudo-dendrite, thyroid eye disease, iris nevus, cornea foreign body, and ocular perforation. Context-enriched answers showed deeper diagnostic reasoning, and blending clinical history with visual findings. Example cases are detailed in Figs. and . When analyzing the results per anatomical region, regions that are readily visible in external photographs, such as the cornea and eyelids had high class agreement. In contrast, regions that can only be indirectly inferred from external images, such as the lacrimal system and orbit had relatively lower class agreement. A detailed breakdown of class agreement by anatomical region is provided in Table . This study evaluated multimodal GPT-4 for clinical diagnosis in ophthalmology based on patient ocular images and clinical context. There are several important findings: (1) GPT-4 V showed capability for ocular image analysis, correctly identifying 48.5% (16/33) of cases based on images alone. (2) When clinical context was added, the accuracy of the model improved to 69.7%. (3) GPT-4 V performance on ophthalmology cases was comparable to non-ophthalmology physicians. A multimodal algorithm that synergizes clinical text with images signals a groundbreaking advance in medical image analysis. The integration of visual and textual data imitates a human decision making process. This is relevant to all medical specialties, but is mostly prominent in specialties that rely on pattern recognition, such as ophthalmology, dermatology, radiology and pathology. These are also the specialties that are at the forefront of AI applications in healthcare, with variable algorithms available and being evaluated for medical images processing , . With an added value of textual analysis, LLMs may ultimately surpass current algorithms that analyze images only . Although there are few recent publications on multimodal GPT-4 V, to the best of our knowledge, our study is the first to evaluate actual patient cases instead of relying on images available in medical question repositories . In this study, GPT-4 V and physicians’ performance improved with clinical context. This reinforces the long-standing medical principle that clinical history is key for accurate diagnosis . Consequently, the potential of multimodal LLMs in ophthalmology is vast. With improvement, such an algorithm can be used as a decision support tool for physicians in diagnosis and management planning. It can also be used in research for cohort generation, enabling creation of large datasets that include textual data with images findings. A multimodal LLM might also significantly impact education in ophthalmology. There is a lack of medical students’ and primary care physicians’ education and training on initial management of basic ophthalmic cases . This is supported by the results of our study, with non-ophthalmology physicians achieving diagnostic accuracies ranging between 58 and 79%. An educational tool that can provide detailed explanations of ocular examination and imaging findings could be used to improve basic understanding and recognition of ophthalmic pathologies. Patient education in this field is also lacking. Many patients already seek initial information using online and unreliable ways, which might lead to harm. A potential restricted model focused on patient education, might be able to integrate patient taken external images with a short patient submitted history to provide supervised case specific patient information, and recommended initial management. Despite their tremendous potential, there are challenges with multimodal application of LLMs in ophthalmology. First, images that cannot be anonymized, such as full-face photos, pose significant privacy issues. To address this, accessing the model would require strict security protocols. Furthermore, these models can be susceptible to cyber threats, including adversarial attacks . In addition, the models can potentially perpetuate bias in healthcare based on data from images, such as skin color or gender . This study has several limitations. First, this was a retrospective analysis of cases, chosen subjectively with potential for selection bias. Second, this was a proof-of-concept study, with a small sample size. This study included only external eye photographs and excluded imaging modalities such as OCT, fundoscopy, or radiological imaging. Cases were intentionally curated by a board-certified ophthalmologist to represent a spectrum of common external eye conditions encountered at the resident level, rather than through random sampling. This approach may have introduced a selection bias and limited the representation of diseases requiring specialized imaging, potentially influencing the overall performance. Finally, due to patient privacy concerns, we have only inserted tightly cropped images focused only on the eyes, to preserve anonymity. This may have influenced the model’s diagnostic abilities. To conclude, GPT-4 V at its current stage is not yet suitable for clinical application in ophthalmology. Nonetheless, its ability to simultaneously analyze and integrate visual and textual data is promising. Multimodal large language models like GPT-4 V have significant potential to advance patient care, education and research in ophthalmology. Future studies involving diverse imaging modalities and randomized case selection are essential to provide a more comprehensive evaluation of these models, their strengths, and limitations. |
Genome-wide copy number variations as molecular diagnostic tool for cutaneous intermediate melanocytic lesions: a systematic review and individual patient data meta-analysis | 070cef4c-8464-41f0-8cac-f88332afa8a2 | 8516778 | Pathology[mh] | Cutaneous melanocytic neoplasms include various tumor types with clinical behavior ranging from indolent to invasive . Histopathologic evaluation is usually sufficient for classification as either benign (nevus) or malignant (melanoma). However, a minority displays ambiguous histopathological features, not allowing definite classification. Studies of preneoplastic melanocytic lesions have shown that intermediate stages exist in the progression from nevus to melanoma, associated with the acquisition of pathogenic genomic aberrations . These observations challenge the notion that melanocytic neoplasms can only be benign or malignant. Therefore, one of the significant changes in the recently updated World Health Organization (WHO) classification of skin tumors is the classification of melanocytic tumors in nine pathways with four-step progression models . As such, the group of intermediate tumors has expanded, for which the term “melanocytoma” has been proposed with two different grades. These lesions present a diagnostic challenge even for expert dermatopathologists . Importantly, incorrect classification might result in either preventable disease progression or substantial unnecessary costs, psychological stress, and additional surgery. Therefore, various ancillary cytogenetic techniques are employed to help distinguish nevi from melanomas, based on the fact that melanomas usually harbor copy number variations (CNVs) whereas nevi do not (or show specific isolated abnormalities) . Cytogenetic techniques such as comparative genomic hybridization (CGH) array and single-nucleotide polymorphism (SNP) array can detect CNVs genome-wide, resulting in improved diagnostic accuracy in ambiguous melanocytic lesions compared to FISH . Thus, CNVs might provide a valuable tool to allow accurate classification. However, to what extent intermediate lesions carry CNVs has not been well established yet, and a CNV cut-off value to distinguish them from melanoma is not well defined. Therefore, we performed a systematic review and individual patient data meta-analysis to evaluate the use of CNVs to classify intermediate melanocytic lesions. Search and study selection Embase and PubMed were systematically searched for primary research articles published in English until September 2020, using the terms “ambiguous,” “atypical,” “borderline,” “dysplastic,” “intermediate,” “spitzoid,” “uncertain,” or “unclassified,” paired with major keywords for melanocytic lesions (including melanocytic “lesion,” “tumor,” “proliferation,” “neoplasm,” “nevus,” “nevi,” “melanoma,” “melanocytoma,” “MELTUMP,” “spitz*,” “STUMP”). These results were then overlapped with the MeSH/Emtree terms for DNA copy number variations: “copy number”; “CNA”; “CNV”; “chromosomal aberration, duplication, amplification, deletion, alteration”; “comparative genomic hybridization”; “CGH”; or “SNP array.” After duplicate removal, unique records were screened for eligibility based on title and abstract first and full-text records thereafter by two authors (CE, WB) using Rayyan for systematic reviews . Differences were discussed until consensus was reached or through input from a third author (AJ). Last, backward and forward snowballing of included articles was employed to identify additional articles of interest. Eligibility criteria and outcomes of interest Articles were included when reporting on intermediate cutaneous melanocytic lesions using molecular techniques to identify genome-wide CNVs, such as CGH array and SNP array. Studies using next-generation sequencing (NGS) were included when using panels or computational methods allowing genome-wide copy number calling . Studies using FISH or multiplex ligation-dependent probe amplification (MLPA) were excluded since these techniques do not screen genome-wide for CNVs. Case reports, abstracts, poster presentations, and articles reporting on non-cutaneous melanomas or melanoma cell lines were excluded. The primary outcomes of interest were the number of CNVs and the type of chromosomal aberrations. Secondary outcomes were clinical follow-up, genomic aberrations, and histopathological characteristics. Data collection and CNV count CNVs were identified on individual lesion level. Authors were contacted to obtain individual patient data or additional information if needed. Two authors (CE, AJ) independently performed a CNV count based on the reported chromosomal aberrations using a predefined ruleset. Segmental gains, losses, high-level amplifications, aneuploidy, and polyploidy were each counted as one CNV. Homozygous loss was counted as two CNVs. CNVs considered insignificant in some studies because of their association with generally benign behavior, such as loss of 3p21 ( BAP1 gene) and gain of 11p ( HRAS gene), were included in the CNV count for uniformity. Chromosomal fusions for which both fusion partners were known were counted as one CNV since they result from one translocation event. Copy-neutral loss of heterozygosity (CN-LOH) was registered separately since it is not accompanied by actual copy number changes. In contrast, chromothripsis can comprise many CNVs but constitutes one tumor event. Therefore, chromothripsis was also registered separately. CNV counts were crosschecked against the reported number of CNVs when available. Ambiguities were resolved via contacting corresponding authors, discussion until consensus, or input from a third author (WB). Recategorization and reclassification of lesions All lesions were reviewed in-depth by two authors (CE, WB) and were recategorized and reclassified according to the 2018 WHO classification of skin tumors. Ambiguous lesions were recategorized as either “benign,” “intermediate,” or “malignant” hierarchically based on (1) provided clinical follow-up, (2) WHO definition, and (3) histopathology and additional case information. Ambiguous or benign lesions with metastatic disease beyond regional lymph nodes during follow-up were recategorized as malignant. Positive sentinel lymph node biopsies were not considered sufficient proof of malignancy since a minority of benign lesions occasionally display such behavior . Per WHO definition, BAP1 -inactivated nevi (BIN), deep penetrating nevi (DPN), cellular blue nevi (CBN), and congenital nevi with proliferative nodules (CNPN) were recategorized as low-grade intermediate. BAP1 -inactivated melanocytomas (BIM), deep penetrating melanocytomas (DPM), atypical cellular blue nevi (ACBN), melanocytic tumors of uncertain malignant potential (MELTUMP), and pigmented epithelioid melanocytomas (PEM) were recategorized as high-grade intermediate. Subsequently, all lesions were reclassified according to the nine WHO pathways primarily based on provided genomic data. When distinctive genomic drivers were unavailable, lesions were reclassified based on the evaluation of available histopathology, ancillary tests, and additional case information. Statistical analysis First, we created box plots to describe the data. Although these appeared not normally distributed, we also reported means to allow comparison with previously reported research on CNV counts. Second, we performed Mann-Whitney U tests to determine differences in CNV number between lesion categories and within classifications according to WHO pathway. Third, we created receiver operating characteristic (ROC) curves and calculated the C -statistic or area under the ROC curve (AUC). Fourth, sensitivity and specificity were calculated for a range of CNV cut-offs (0–7). As sensitivity analyses, we performed these analyses for two alternative categorizations of the lesions: (1) initially reported category and (2) considering low-grade intermediate lesions (BIN, CBN, CN with proliferative nodules, and DPN) as benign. Furthermore, we evaluated an alternative CNV count, including chromothripsis and CN-LOH. Also, we evaluated CNV count based on CGH data or SNP data only. Last, we evaluated sensitivity and specificity irrespective of CNV count by interpreting microarray data as positive for malignancy in the presence of CNVs suspect for melanoma, such as homozygous loss of 9p21 ( CDKN2A ) and gain of 11q13 ( CCND1 ), 8q24 ( MYC ), or 6p25 ( RREB1 ). All statistical analyses were performed in SPSS version 26. Embase and PubMed were systematically searched for primary research articles published in English until September 2020, using the terms “ambiguous,” “atypical,” “borderline,” “dysplastic,” “intermediate,” “spitzoid,” “uncertain,” or “unclassified,” paired with major keywords for melanocytic lesions (including melanocytic “lesion,” “tumor,” “proliferation,” “neoplasm,” “nevus,” “nevi,” “melanoma,” “melanocytoma,” “MELTUMP,” “spitz*,” “STUMP”). These results were then overlapped with the MeSH/Emtree terms for DNA copy number variations: “copy number”; “CNA”; “CNV”; “chromosomal aberration, duplication, amplification, deletion, alteration”; “comparative genomic hybridization”; “CGH”; or “SNP array.” After duplicate removal, unique records were screened for eligibility based on title and abstract first and full-text records thereafter by two authors (CE, WB) using Rayyan for systematic reviews . Differences were discussed until consensus was reached or through input from a third author (AJ). Last, backward and forward snowballing of included articles was employed to identify additional articles of interest. Articles were included when reporting on intermediate cutaneous melanocytic lesions using molecular techniques to identify genome-wide CNVs, such as CGH array and SNP array. Studies using next-generation sequencing (NGS) were included when using panels or computational methods allowing genome-wide copy number calling . Studies using FISH or multiplex ligation-dependent probe amplification (MLPA) were excluded since these techniques do not screen genome-wide for CNVs. Case reports, abstracts, poster presentations, and articles reporting on non-cutaneous melanomas or melanoma cell lines were excluded. The primary outcomes of interest were the number of CNVs and the type of chromosomal aberrations. Secondary outcomes were clinical follow-up, genomic aberrations, and histopathological characteristics. CNVs were identified on individual lesion level. Authors were contacted to obtain individual patient data or additional information if needed. Two authors (CE, AJ) independently performed a CNV count based on the reported chromosomal aberrations using a predefined ruleset. Segmental gains, losses, high-level amplifications, aneuploidy, and polyploidy were each counted as one CNV. Homozygous loss was counted as two CNVs. CNVs considered insignificant in some studies because of their association with generally benign behavior, such as loss of 3p21 ( BAP1 gene) and gain of 11p ( HRAS gene), were included in the CNV count for uniformity. Chromosomal fusions for which both fusion partners were known were counted as one CNV since they result from one translocation event. Copy-neutral loss of heterozygosity (CN-LOH) was registered separately since it is not accompanied by actual copy number changes. In contrast, chromothripsis can comprise many CNVs but constitutes one tumor event. Therefore, chromothripsis was also registered separately. CNV counts were crosschecked against the reported number of CNVs when available. Ambiguities were resolved via contacting corresponding authors, discussion until consensus, or input from a third author (WB). All lesions were reviewed in-depth by two authors (CE, WB) and were recategorized and reclassified according to the 2018 WHO classification of skin tumors. Ambiguous lesions were recategorized as either “benign,” “intermediate,” or “malignant” hierarchically based on (1) provided clinical follow-up, (2) WHO definition, and (3) histopathology and additional case information. Ambiguous or benign lesions with metastatic disease beyond regional lymph nodes during follow-up were recategorized as malignant. Positive sentinel lymph node biopsies were not considered sufficient proof of malignancy since a minority of benign lesions occasionally display such behavior . Per WHO definition, BAP1 -inactivated nevi (BIN), deep penetrating nevi (DPN), cellular blue nevi (CBN), and congenital nevi with proliferative nodules (CNPN) were recategorized as low-grade intermediate. BAP1 -inactivated melanocytomas (BIM), deep penetrating melanocytomas (DPM), atypical cellular blue nevi (ACBN), melanocytic tumors of uncertain malignant potential (MELTUMP), and pigmented epithelioid melanocytomas (PEM) were recategorized as high-grade intermediate. Subsequently, all lesions were reclassified according to the nine WHO pathways primarily based on provided genomic data. When distinctive genomic drivers were unavailable, lesions were reclassified based on the evaluation of available histopathology, ancillary tests, and additional case information. First, we created box plots to describe the data. Although these appeared not normally distributed, we also reported means to allow comparison with previously reported research on CNV counts. Second, we performed Mann-Whitney U tests to determine differences in CNV number between lesion categories and within classifications according to WHO pathway. Third, we created receiver operating characteristic (ROC) curves and calculated the C -statistic or area under the ROC curve (AUC). Fourth, sensitivity and specificity were calculated for a range of CNV cut-offs (0–7). As sensitivity analyses, we performed these analyses for two alternative categorizations of the lesions: (1) initially reported category and (2) considering low-grade intermediate lesions (BIN, CBN, CN with proliferative nodules, and DPN) as benign. Furthermore, we evaluated an alternative CNV count, including chromothripsis and CN-LOH. Also, we evaluated CNV count based on CGH data or SNP data only. Last, we evaluated sensitivity and specificity irrespective of CNV count by interpreting microarray data as positive for malignancy in the presence of CNVs suspect for melanoma, such as homozygous loss of 9p21 ( CDKN2A ) and gain of 11q13 ( CCND1 ), 8q24 ( MYC ), or 6p25 ( RREB1 ). All statistical analyses were performed in SPSS version 26. Study selection Figure shows the PRISMA flowchart for study selection . The search yielded 647 hits, of which 432 were unique records. After assessment for eligibility, 25 studies were included in the meta-analysis, and a further six were identified through snowballing. Study characteristics The characteristics of the 31 included studies are listed in Table . All studies were either retrospective ( n =25), prospective ( n =4), or mixed ( n =2) case series. Twenty-two studies used CGH array, six studies used SNP array, and three studies used NGS. In total, data for 431 individual lesions were extracted, of which 252 (58.5%) had been analyzed with CGH array, 144 (33.4%) with SNP array, and 35 (8.1%) with NGS. Recategorization and reclassification of lesions Initially, 113 lesions (26.2%) were presented as benign, 212 (49.2%) as ambiguous, and 106 (24.6%) as malignant. Clinical follow-up was available for 297 lesions (68.9%), of which 140 were ambiguous. Two benign and ten ambiguous lesions were recategorized as malignant based on follow-up with distant metastasis or additional case information. Per WHO-definition, 80 lesions were recategorized as low-grade intermediate (28 BIN, 22 CBN, 15 CNPN, and 15 DPN) and 83 lesions as high-grade intermediate (30 ACBN, 7 BIM, 13 DPM, 16 MELTUMP, and 17 PEM). A total of 81 intermediate lesions could not be specified as either low- or high-grade (76 AST, three melanocytomas with CRTC1 - TRIM11 fusions, and two melanocytomas with NRAS p.Q61R and IDH1 p.R132C mutations). After recategorization, 69 (16.0%) benign, 244 (56.6%) intermediate, and 118 (27.4%) malignant lesions were available for meta-analysis. Distinctive genomic drivers, including ALK , ROS1 , NTRK , BRAF , or MET fusions and mutational status for BAP1 , BRAF, GNA11 , GNAQ , HRAS , and NRAS , were available for 206/431 (47.8%) lesions and 145/244 (59.4%) intermediate lesions. Accordingly, 61/431 (14.1%) lesions were reclassified, mostly “Spitz” lesions carrying a BRAF p.V600E or NRAS p.Q61R mutation and lesions designated “DPN,” “DPM,” or “MELTUMP” carrying a GNAQ p.Q209L or GNA11 p.Q209L mutation. Chromosomal aberrations in intermediate lesions In our dataset, 18/69 (26.1%) of benign, 134/244 (54.9%) of intermediate, and 112/118 (94.9%) of malignant lesions displayed ≥1 CNV. Within intermediate lesions, 43/80 (53.8%) of low-grade, 35/83 (42.2%) of high-grade, and 56/81 (69.5%) of intermediate lesions not otherwise specified (NOS) displayed ≥1 CNV. The most frequently encountered CNVs in intermediate lesions are listed in Table . Loss of 3p spanning the BAP1 gene on 3p21 was most commonly found, all but one (ACBN) harbored by BAP1 -inactivated lesions. The most common gain involved 7q, carried mainly by AST. Chromosomal aberrations known to occur in melanoma frequently were infrequent or absent in intermediate lesions (marked with an asterisk in Table ). Two AST displayed heterozygous loss of 9p21 spanning the CDKN2A gene. One ACBN showed a gain of 8q24 spanning the MYC gene. Aneuploidies were mainly found in BIN/BIM carrying a loss of chromosome 3 and CNPN carrying a loss of chromosome 7 and gain of chromosome 8. Chromothripsis was found in one malignant Spitz tumor (MST) and five intermediate lesions (two AST, one CBN, one CNPN, and one MELTUMP). Of these, clinical follow-up was only available for the CBN and CNPN. The CBN harbored chromothripsis of chromosomes 3 and 7 and 14 additional CNVs, without evidence of disease during a follow-up of 3.8 years. The CNPN harbored chromothripsis of 1p and two additional CNVs, and the patient was disease-free at 3.5 years after excision. CN-LOH was found in eight melanomas and three intermediate lesions (one DPM, one MELTUMP, and one PEM). The DPM carried CN-LOH of 17q12-qter and did not harbor additional CNVs. The MELTUMP carried CN-LOH of chromosome 7 and had 15 additional CNVs. Clinical follow-up for these cases was not available. The PEM carried CN-LOH of the distal part of chromosome 17q and did not harbor any additional CNVs. Short-term clinical follow-up (not specified) did not show any sign of disease. CNV counts after recategorization and reclassification Figures and show the number of CNVs per lesion category and WHO class, respectively. The CNV number in intermediate lesions (median 1, interquartile range [IQR] 0–2) was significantly higher ( p <0.001) compared to that in benign lesions (median 0, IQR 0–1) and significantly lower ( p <0.001) compared to that in malignant lesions (median 6, IQR 4–11) (Fig. ). There was no significant difference between low-grade or high-grade intermediate lesions ( p =0.499). In WHO pathway I, CNV number in BIM (median 1, IQR 1–1.5) was not significantly higher ( p =0.092) compared to that in BIN (median 1, IQR 1–1) and not significantly higher ( p =0.449) in DPM (median 0, IQR 0–1) compared to that in DPN (median 0, IQR 0–0). CNV number in PEM (median 0, IQR 0–0) was significantly lower ( p <0.001) compared to melanomas in PEM (median 4, IQR 4–5). In pathway IV, CNV number in AST (median 1, IQR 0–2) was significantly higher ( p <0.001) compared to that in Spitz nevi (median 0, IQR 0–1) and significantly lower ( p =0.001) compared to that in MSTs (median 5, IQR 4–8). In pathway VII, CNV number in CNPN (median 2, IQR 1–5) was significantly higher ( p <0.001) compared to that in CN (median 0, IQR 0–0) and significantly lower ( p =0.009) compared to melanomas in CN (median 7.5, IQR 5.5–9.5). Last, in pathway VIII, CNV number in CBN (median 0, IQR 0–0) was not significantly different ( p =0.585) from blue nevi (median 0, IQR 0–0) but significantly lower ( p =0.015) compared to that in ACBN (median 0, IQR 0–2). CNV number in ACBN was significantly lower ( p <0.001) compared to melanomas in blue nevi (median 6, IQR 4–8) (Fig. ). Relevant outliers in the intermediate category are shown as yellow dots in Fig. . The most extreme outlier corresponded to a CNPN harboring 22 CNVs, all gains and losses of whole chromosomes. Clinical follow-up was not available for the MELTUMP with 15 CNVs, the DPM with 10 CNVs, and the AST with 7 CNVs. The remaining lesions did not show evidence of disease during the available relatively short follow-up (varying from 14 to 46 months). CNV cut-off value The C -statistic to differentiate between nevi and melanoma was 0.96 (95% CI 0.93–0.99, p <0.001) and between intermediate lesions and melanoma 0.90 (95% CI 0.86–0.94, p <0.001), indicating excellent ability to differentiate using CNV number . In contrast, the CNV number displayed poor ability to differentiate between intermediate and benign lesions ( C -statistic 0.67, 95% CI 0.61–0.73, p <0.001). Figure shows sensitivity and specificity for differentiating intermediate from malignant lesions given various CNV cut-off values. Using a cut-off of ≥3 CNVs, 85% of malignant lesions would be correctly categorized as malignant (sensitivity), and 84% of non-malignant lesions would be correctly classified as non-malignant (specificity). Sensitivity analyses None of the alternative lesion categorizations or alternative CNV counts substantially changed the results for differentiation between intermediate and malignant lesions. The AUC based on the initially reported category was 0.88 (95% CI 0.84–0.92, p <0.001). The AUC when considering low-grade intermediate lesions benign remained 0.90 (95% CI 0.86–0.94, p <0.001). The AUC, when including chromothripsis and CN-LOH in the CNV count, also remained 0.90 (95% CI 0.86–0.94, p <0.001). The AUC based on SNP array or CGH data only decreased to 0.85 (95% CI 0.78–0.92, p <0.001) and increased to 0.94 (95% CI 0.91–0.97, p <0.001), respectively. Including specific CNVs suspect for melanoma as a positive test marker did not substantially change these results. Risk of bias across studies The risk of bias was generally considered low to unknown and constituted mainly selection and information bias. Most studies used archival cases without adequately defining the selection process, creating an unknown selection bias risk. Three studies reported CNVs for selected representative cases. Comprising only 10 cases, we consider the impact of potential selection bias very low. In addition, the detection of CNVs is highly dependent on the type of microarray, resolution, DNA quality, and sample purity. Most studies used archival DNA from formalin-fixed paraffin-embedded (FFPE) tissue and did not report tumor cell percentages, which introduces an unknown risk of information bias. Figure shows the PRISMA flowchart for study selection . The search yielded 647 hits, of which 432 were unique records. After assessment for eligibility, 25 studies were included in the meta-analysis, and a further six were identified through snowballing. The characteristics of the 31 included studies are listed in Table . All studies were either retrospective ( n =25), prospective ( n =4), or mixed ( n =2) case series. Twenty-two studies used CGH array, six studies used SNP array, and three studies used NGS. In total, data for 431 individual lesions were extracted, of which 252 (58.5%) had been analyzed with CGH array, 144 (33.4%) with SNP array, and 35 (8.1%) with NGS. Initially, 113 lesions (26.2%) were presented as benign, 212 (49.2%) as ambiguous, and 106 (24.6%) as malignant. Clinical follow-up was available for 297 lesions (68.9%), of which 140 were ambiguous. Two benign and ten ambiguous lesions were recategorized as malignant based on follow-up with distant metastasis or additional case information. Per WHO-definition, 80 lesions were recategorized as low-grade intermediate (28 BIN, 22 CBN, 15 CNPN, and 15 DPN) and 83 lesions as high-grade intermediate (30 ACBN, 7 BIM, 13 DPM, 16 MELTUMP, and 17 PEM). A total of 81 intermediate lesions could not be specified as either low- or high-grade (76 AST, three melanocytomas with CRTC1 - TRIM11 fusions, and two melanocytomas with NRAS p.Q61R and IDH1 p.R132C mutations). After recategorization, 69 (16.0%) benign, 244 (56.6%) intermediate, and 118 (27.4%) malignant lesions were available for meta-analysis. Distinctive genomic drivers, including ALK , ROS1 , NTRK , BRAF , or MET fusions and mutational status for BAP1 , BRAF, GNA11 , GNAQ , HRAS , and NRAS , were available for 206/431 (47.8%) lesions and 145/244 (59.4%) intermediate lesions. Accordingly, 61/431 (14.1%) lesions were reclassified, mostly “Spitz” lesions carrying a BRAF p.V600E or NRAS p.Q61R mutation and lesions designated “DPN,” “DPM,” or “MELTUMP” carrying a GNAQ p.Q209L or GNA11 p.Q209L mutation. In our dataset, 18/69 (26.1%) of benign, 134/244 (54.9%) of intermediate, and 112/118 (94.9%) of malignant lesions displayed ≥1 CNV. Within intermediate lesions, 43/80 (53.8%) of low-grade, 35/83 (42.2%) of high-grade, and 56/81 (69.5%) of intermediate lesions not otherwise specified (NOS) displayed ≥1 CNV. The most frequently encountered CNVs in intermediate lesions are listed in Table . Loss of 3p spanning the BAP1 gene on 3p21 was most commonly found, all but one (ACBN) harbored by BAP1 -inactivated lesions. The most common gain involved 7q, carried mainly by AST. Chromosomal aberrations known to occur in melanoma frequently were infrequent or absent in intermediate lesions (marked with an asterisk in Table ). Two AST displayed heterozygous loss of 9p21 spanning the CDKN2A gene. One ACBN showed a gain of 8q24 spanning the MYC gene. Aneuploidies were mainly found in BIN/BIM carrying a loss of chromosome 3 and CNPN carrying a loss of chromosome 7 and gain of chromosome 8. Chromothripsis was found in one malignant Spitz tumor (MST) and five intermediate lesions (two AST, one CBN, one CNPN, and one MELTUMP). Of these, clinical follow-up was only available for the CBN and CNPN. The CBN harbored chromothripsis of chromosomes 3 and 7 and 14 additional CNVs, without evidence of disease during a follow-up of 3.8 years. The CNPN harbored chromothripsis of 1p and two additional CNVs, and the patient was disease-free at 3.5 years after excision. CN-LOH was found in eight melanomas and three intermediate lesions (one DPM, one MELTUMP, and one PEM). The DPM carried CN-LOH of 17q12-qter and did not harbor additional CNVs. The MELTUMP carried CN-LOH of chromosome 7 and had 15 additional CNVs. Clinical follow-up for these cases was not available. The PEM carried CN-LOH of the distal part of chromosome 17q and did not harbor any additional CNVs. Short-term clinical follow-up (not specified) did not show any sign of disease. Figures and show the number of CNVs per lesion category and WHO class, respectively. The CNV number in intermediate lesions (median 1, interquartile range [IQR] 0–2) was significantly higher ( p <0.001) compared to that in benign lesions (median 0, IQR 0–1) and significantly lower ( p <0.001) compared to that in malignant lesions (median 6, IQR 4–11) (Fig. ). There was no significant difference between low-grade or high-grade intermediate lesions ( p =0.499). In WHO pathway I, CNV number in BIM (median 1, IQR 1–1.5) was not significantly higher ( p =0.092) compared to that in BIN (median 1, IQR 1–1) and not significantly higher ( p =0.449) in DPM (median 0, IQR 0–1) compared to that in DPN (median 0, IQR 0–0). CNV number in PEM (median 0, IQR 0–0) was significantly lower ( p <0.001) compared to melanomas in PEM (median 4, IQR 4–5). In pathway IV, CNV number in AST (median 1, IQR 0–2) was significantly higher ( p <0.001) compared to that in Spitz nevi (median 0, IQR 0–1) and significantly lower ( p =0.001) compared to that in MSTs (median 5, IQR 4–8). In pathway VII, CNV number in CNPN (median 2, IQR 1–5) was significantly higher ( p <0.001) compared to that in CN (median 0, IQR 0–0) and significantly lower ( p =0.009) compared to melanomas in CN (median 7.5, IQR 5.5–9.5). Last, in pathway VIII, CNV number in CBN (median 0, IQR 0–0) was not significantly different ( p =0.585) from blue nevi (median 0, IQR 0–0) but significantly lower ( p =0.015) compared to that in ACBN (median 0, IQR 0–2). CNV number in ACBN was significantly lower ( p <0.001) compared to melanomas in blue nevi (median 6, IQR 4–8) (Fig. ). Relevant outliers in the intermediate category are shown as yellow dots in Fig. . The most extreme outlier corresponded to a CNPN harboring 22 CNVs, all gains and losses of whole chromosomes. Clinical follow-up was not available for the MELTUMP with 15 CNVs, the DPM with 10 CNVs, and the AST with 7 CNVs. The remaining lesions did not show evidence of disease during the available relatively short follow-up (varying from 14 to 46 months). The C -statistic to differentiate between nevi and melanoma was 0.96 (95% CI 0.93–0.99, p <0.001) and between intermediate lesions and melanoma 0.90 (95% CI 0.86–0.94, p <0.001), indicating excellent ability to differentiate using CNV number . In contrast, the CNV number displayed poor ability to differentiate between intermediate and benign lesions ( C -statistic 0.67, 95% CI 0.61–0.73, p <0.001). Figure shows sensitivity and specificity for differentiating intermediate from malignant lesions given various CNV cut-off values. Using a cut-off of ≥3 CNVs, 85% of malignant lesions would be correctly categorized as malignant (sensitivity), and 84% of non-malignant lesions would be correctly classified as non-malignant (specificity). None of the alternative lesion categorizations or alternative CNV counts substantially changed the results for differentiation between intermediate and malignant lesions. The AUC based on the initially reported category was 0.88 (95% CI 0.84–0.92, p <0.001). The AUC when considering low-grade intermediate lesions benign remained 0.90 (95% CI 0.86–0.94, p <0.001). The AUC, when including chromothripsis and CN-LOH in the CNV count, also remained 0.90 (95% CI 0.86–0.94, p <0.001). The AUC based on SNP array or CGH data only decreased to 0.85 (95% CI 0.78–0.92, p <0.001) and increased to 0.94 (95% CI 0.91–0.97, p <0.001), respectively. Including specific CNVs suspect for melanoma as a positive test marker did not substantially change these results. The risk of bias was generally considered low to unknown and constituted mainly selection and information bias. Most studies used archival cases without adequately defining the selection process, creating an unknown selection bias risk. Three studies reported CNVs for selected representative cases. Comprising only 10 cases, we consider the impact of potential selection bias very low. In addition, the detection of CNVs is highly dependent on the type of microarray, resolution, DNA quality, and sample purity. Most studies used archival DNA from formalin-fixed paraffin-embedded (FFPE) tissue and did not report tumor cell percentages, which introduces an unknown risk of information bias. To the best of our knowledge, this is the first systematic review and individual patient data meta-analysis to assess genome-wide CNVs as a diagnostic tool for intermediate melanocytic lesions, using cytogenetic tests such as SNP array and CGH array. Chromosomal aberrations were found in 55.1% of intermediate lesions. Gains and losses frequently seen in melanoma, such as gain of 1q, 6p, and 7q and loss of 6q and 9p, were uncommon in intermediate lesions. CN-LOH and chromothripsis were only found in intermediate and malignant lesions. Our analysis shows that the median number of CNVs in intermediate lesions is statistically significantly higher compared to that in nevi and lower compared to that in melanoma. Similarly, the number of CNVs significantly increased in WHO pathway IV (Spitz), VII (congenital), and VIII (blue) along the spectrum from nevus to melanoma. In contrast, the CNV number was not statistically different between BIN and BIM and between DPN and DPM. Surprisingly, “high-grade” melanocytomas (ACBN, BIM, DPM, PEM, and MELTUMP) carried CNVs less frequently than “low-grade” melanocytomas (BIN, CBN, CMN with proliferative nodules, and DPN). This observation demonstrates the difficulty of grading intermediate lesions using a four-tier system as is used in the WHO classification. Yet, our results suggest CNVs demonstrate excellent ability to differentiate between intermediate melanocytic lesions and melanoma in clinical practice. A cut-off of ≥3 CNVs corresponded to 85% sensitivity and 84% specificity, and a cut-off of ≥4 CNVs corresponded to 81% sensitivity and 91% specificity, respectively. Several CNV cut-offs for malignancy have previously been suggested. Based on their case series, Maize et al. and Chan et al. suggested a cut-off of ≥3 CNVs and ≥4 CNVs, respectively, and Alomari et al. proposed an algorithm using ≥4 significant CNVs with additional caveats in case of ≤3 CNVs . Our current meta-analysis integrates and expands their data, providing more robust evidence for various cut-offs in the classification of intermediate lesions. Both a cut-off of ≥3 and ≥4 CNVs can be considered, the first having a higher sensitivity (fewer false-negative diagnoses) and the latter having a higher specificity (fewer false-positive diagnoses). Yet, sensitivity might prevail in clinical practice given the potentially disastrous consequences of a false-negative misdiagnosis, even at the cost of a modestly lower specificity and resulting treatment burden. Therefore, we propose a cut-off of ≥3 CNVs as indicative of malignancy. Of note, a minority of melanomas did not harbor CNVs, and benign lesions might carry CNVs with limited prognostic value. In contrast, specific CNVs may also be relevant if present in isolation, such as homozygous loss of 9p21 ( CDKN2A ) . Therefore, CNVs should always be interpreted considering the clinicopathological context. Yet, the contextual interpretation of specific CNVs is difficult in unclassified lesions. For example, loss of 3p21 ( BAP1 ) is insignificant in BIN/BIM but is of major significance in an ACBN or MEBN. It is currently mostly unknown which CNVs are most predictive for malignancy in the various WHO pathways, and this requires additional research. The main strengths of this study include the following. All lesions have been vigorously reviewed based on published and unpublished individual patient data using the 2018 WHO classification of skin tumors. As such, our meta-analysis integrates the most recent clinicopathological and genomic insights to establish CNVs in intermediate lesions. Our analyses provide a better-defined CNV cut-off value for malignancy to support clinical decision-making, based on the largest pooled dataset of intermediate lesions to date. The sensitivity analyses strengthened the robustness of the results. This study has several limitations. First, genome-wide microarray data are difficult to pool since the detection of CNVs is highly dependent on resolution and technical specifications. Second, the detection of CNVs depends on sample quality. Most studies used DNA from archival FFPE blocks and did not report the tumor cell percentages, although the latter should exceed 30% to detect CNVs reliably. Therefore, paucicellular lesions such as PEM and large-cell lesions such as MAP3K8 - SVIL fused ASTs with strong lymphocytic infiltrate render dilution effects and make CNV detection more difficult. In our dataset, a minority of melanomas (5.1%) did not carry any CNVs, and it is unclear if this is due to dilution or truly represents a lack of CNVs. This might have negatively influenced sensitivity, although our ROC analysis still showed excellent discriminative ability. Third, rules for CNV counts are not uniformly defined, and we were only able to count CNVs that were reported or provided by authors. CNVs reported to be attributable to a chromosomal fusion were counted as one CNV. This may have slightly overestimated the CNV number, especially in Spitz tumors, where fusions are a common driver event and probably not relevant for prognosis. Also, a minority of melanomas were reported to harbor extremely high CNV counts (>30), which likely included aberrations not included in our CNV count, such as chromothripsis. However, it is unlikely this substantially affected our results since we performed non-parametric tests. Fourth, clinical outcomes were not available in 34.0% of ambiguous cases. Follow-up with distant metastasis remains the gold standard for proof of malignancy and might still occur years after diagnosis. Yet, clinical follow-up was available for most outliers in the intermediate category. Fifth, distinctive driver mutations were available in 59.4% of intermediate lesions. For the remaining lesions, classification was based on histopathology alone, which is more subjective than genomic data. Last, we established our ROC analysis on one variable (CNV number), whereas ideally, diagnostic evaluation is performed via multivariable analysis, including all available diagnostic information. Nonetheless, it indicated excellent discriminative ability, which supports further research in a dedicated dataset. Despite these limitations, we believe this meta-analysis provides robust results applicable to general dermatopathology practice. To conclude, this systematic review and individual patient data meta-analysis provides a comprehensive overview of CNVs in cutaneous intermediate melanocytic lesions and a diagnostic interpretation of different CNV cut-offs for malignancy, based on the largest pooled cohort of ambiguous melanocytic neoplasms to date. Our results suggest that a cut-off of ≥3 CNVs might represent the optimal trade-off between sensitivity and specificity in clinical practice to distinguish intermediate from malignant lesions. Future research should externally validate this cut-off in a distinct dataset, assess the predictive value of specific CNVs in the various WHO pathways, and correlate genome-wide microarray data with objective genomic and clinical parameters. ESM 1 (SAV 791 kb) |
Evaluation of ophthalmic healthcare professional-led keratoconus management service in the United Kingdom: the Birmingham and Midland Eye Centre (BMEC) study | 9dd96a9f-bd4e-4c6c-b75c-97ce85d5f033 | 11427565 | Ophthalmology[mh] | |
Clinical study on horizontal bone augmentation using an alveolar mucosa-periosteal bone flap | ead1bd62-e879-4857-a627-f67b5073fdab | 11806868 | Dentistry[mh] | Participants The patients with narrow alveolar ridges(Alveolar ridge width < 2 mm) requiring dental implantation were recruited from Oral Implantation Department, Affiliated Haikou Hospital of Central South University Xiangya Medical School & Hainan Provincial Stomatology Center from October 2013 to December 2023. Inclusion criteria: (1) Patients who received MBF surgery to operate bone increment; (2) patients without contraindication in partial or the whole body; (3) patients not taking medicines that affected bone growth within the last 12 months; and (4) patients with complete case information without loss of follow-up. The patients with narrow alveolar ridges(Alveolar ridge width < 2 mm) requiring dental implantation were recruited from Oral Implantation Department, Affiliated Haikou Hospital of Central South University Xiangya Medical School & Hainan Provincial Stomatology Center from October 2013 to December 2023. Inclusion criteria: (1) Patients who received MBF surgery to operate bone increment; (2) patients without contraindication in partial or the whole body; (3) patients not taking medicines that affected bone growth within the last 12 months; and (4) patients with complete case information without loss of follow-up. Instruments: CBCT (New Tom, Italy), piezosurgery (SATELEC Piezotome 2, France), dental implanter (Nobel, Austria) and centrifuge (TD4Z-WS, China), etc. Medical appliances: mainly including Osseoset 200(Nobel, Switzerland), Nobel Active AB(Nobel, Sweden), Bicon toolbox(Bicon, USA) and other relevant appliances. Implants: Nobel Active (Nobel, Sweden), Bicon (Bicon, America), Osstem (Osstem, Korea) and Dentis (Dentis, Korea). Treatment process Periodontal non-surgical treatment: supragingival scaling, subgingival scaling and oral hygiene education were routinely carried out. The PRF preparation was accomplished according to the approach introduced by Cheng Yanan , et al. . The MBF technique was prepared and the implant was inserted using the method presented by Xu Pu , et al. . Stage II operation and missing tooth restoration: at 3–6 months after the insertion of implants, stage II operation was performed. To be specific, the healing cap was removed, and then the healing abutments were inserted for accelerating soft tissue sulci formation. After 2 weeks, the cuffs were in healthy condition, and the missing teeth were restored after taking an impression. Effect evaluation Clinical effect: Whether soft tissue of the cuff lip was full, and whether the fullness on the labial (buccal) side of the edentulous area was restored after the completion of stage II operation and prostheses. Cone beam computed tomography (CT) observation: Cross-section changes of alveolar ridge, alveolar crest width and bone plate thickness on the lip (cheek) side after implantation. Measurement of the alveolar ridge widths before and after implantation: The alveolar ridge widths were determined before implantation and after the final restoration using MBF technique. Then, data were compared. Statistical analysis The SPSS Statistics software (version 17.0) was utilized for data analysis. Furthermore, paired t-tests were conducted to compare the widths of the alveolar ridges before implantation with those following the final restoration, with the outcomes presented as Mean ± Standard Deviation (M ± SD). The threshold for statistical significance was established at P < 0.05. Periodontal non-surgical treatment: supragingival scaling, subgingival scaling and oral hygiene education were routinely carried out. The PRF preparation was accomplished according to the approach introduced by Cheng Yanan , et al. . The MBF technique was prepared and the implant was inserted using the method presented by Xu Pu , et al. . Stage II operation and missing tooth restoration: at 3–6 months after the insertion of implants, stage II operation was performed. To be specific, the healing cap was removed, and then the healing abutments were inserted for accelerating soft tissue sulci formation. After 2 weeks, the cuffs were in healthy condition, and the missing teeth were restored after taking an impression. Clinical effect: Whether soft tissue of the cuff lip was full, and whether the fullness on the labial (buccal) side of the edentulous area was restored after the completion of stage II operation and prostheses. Cone beam computed tomography (CT) observation: Cross-section changes of alveolar ridge, alveolar crest width and bone plate thickness on the lip (cheek) side after implantation. Measurement of the alveolar ridge widths before and after implantation: The alveolar ridge widths were determined before implantation and after the final restoration using MBF technique. Then, data were compared. The SPSS Statistics software (version 17.0) was utilized for data analysis. Furthermore, paired t-tests were conducted to compare the widths of the alveolar ridges before implantation with those following the final restoration, with the outcomes presented as Mean ± Standard Deviation (M ± SD). The threshold for statistical significance was established at P < 0.05. ) During the initial surgery on a narrow alveolar ridge, the surgeon incises through the gum tissue to reach the periosteum, carefully elevates the fibrous bone membrane, and employs a sonic bone cutter to fashion the bone socket along the ridge's crest and sidewalls. This process involves cutting through the cortical bone to access the cancellous bone. Following the bone preparation, the surgical site is meticulously sutured. Approximately 30 days later, the dental implant I-stage surgery is conducted. This procedure involves making an incision through the soft tissue only, avoiding the need for a flap operation. After the incision, the fibrous bone membrane is gently manipulated to increase the ridge crest's width by pushing it towards the labial and buccal sides. The implant is then inserted using standard techniques, and a biomembrane is positioned at the ridge's crest to finalize the closure. Subsequent treatment protocols remain consistent with those of traditional dental implant surgery. The present work recruited 20 cases (age, 17–69 years) and 45 implants. The proportions of female patients (13 patients, 28 implants) and patients with mandibular lesions (12 patients, 28 implants) were higher, while those of male patients (7 patients, 17 implants) and patients with maxillary lesions (8 patients, 17 implants) were lower (Table ). In clinical observation, the alveolar ridge was mostly triangular and its top was narrow before implantation. There were more soft tissues on cuff labial (buccal) side, thereby restoring the fullness of the edentulous area on the labial (buccal) side, and the aesthetic was available after the completion of stage II operation (Fig. ). Cone beam CT examination revealed that the triangular alveolar ridge before operation turned into the trapezoidal alveolar ridge in coronal section after operation, moreover, the bone on the lip (buccal) side of the implant was significantly greater than 1 mm (Fig. ). Meanwhile, the alveolar ridge crest widths were determined before implantation and after the final restoration (Fig. ), and statistical analysis was carried out. According to the results, the alveolar ridge widths were 3.62 ± 0.90 mm before implantation and 6.58 ± 1.16 mm at 3–6 months after implantation. Clearly, the alveolar ridge width elevated by an average of 2.96 ± 1.21 mm. The difference was statistically significant before and after implantation (t = 16.41, P = 1.573E-43 < 0.05) (Table ). Implant-supported restorations represent an optimal solution for replacing missing teeth, with the quality and quantity of the alveolar bone being crucial for achieving the correct implant placement . Several factors may lead to inadequate alveolar ridge width, a condition that predominantly affects female patients and the mandible, as opposed to male patients and the maxilla (refer to Table ). Asian females generally have a smaller stature compared to their European counterparts, and their skeletal structures, including the mandible, are also relatively diminutive. Patients with a narrow mandibular ridge are at a higher risk of experiencing insufficient alveolar ridge width post-tooth extraction. Given that the supporting area is smaller than that of the maxilla, the mandibular alveolar ridge is more susceptible to constriction and resorption following restoration. The traditional osteotomy technique has proven effective in expanding alveolar bone width; however, it carries inherent risks such as fracture, necrosis, and bone resorption. In contrast, the application of a Mucosa-periosteal bone flap (MBF) yields superior results for horizontal alveolar bone augmentation. By positioning the pre-shaped trapezoidal bone block on the buccal aspect, the vascular connection between the mucoperiosteum and the bone block is preserved.This phenomenon ensures the nutritional supply to the bone graft, thereby preventing its necrosis and resorption . Tooth loss often results in the diminishment of alveolar bone, leading to significant atrophy and the formation of a blade-shaped alveolar ridge . Repairing a missing posterior tooth in an atrophic alveolar ridge presents a common clinical challenge . During the clinical observation of this study, it was noted that the alveolar ridge predominantly exhibited a triangular shape with a narrow apex prior to implantation. The application of the MBF technique at the time of implant placement resulted in an increase in the width of the alveolar ridge's apex. After the stage II procedure, the soft tissues on the labial (buccal) side of the cuff were observed to be fully restored, effectively enhancing the fullness on the labial (buccal) aspect of the missing teeth. As a result, the aesthetic outcome was considered satisfactory following the teeth restoration (Fig. ). The cone beam CT examination demonstrated that the implant restoration cases in this study resulted in favorable clinical outcomes. Post-implantation, the alveolar ridge was notably broadened, exhibiting a trapezoidal cross-section, in contrast to its pre-implantation triangular cross-section. Additionally, the bone on the buccal side of the implant was observed to be significantly more than 1 mm. Upon measurement, the alveolar ridge widths were found to be 3.62 ± 0.90 mm prior to implantation and 6.58 ± 1.16 mm following implantation, indicating an increase of 2.96 ± 1.21 mm. The alveolar crest width experienced a significant augmentation post-operation, as depicted in Fig. . These findings suggest that the MBF technique yields positive clinical effects, aligning with the outcomes reported by other researchers . Some researchers have indicated that bone grafting can achieve superior outcomes in the enhancement of alveolar bone . However, it is crucial to create secondary surgical sites to obtain graft material. Implant placement is postponed for 3–6 months, and the graft bone absorption rate reaches 20%-50% at 6 months later . The guided bone regeneration (GBR) procedures carry the risk of membrane exposure and rupture, which can result in infection . Inter-positional augmentation, which includes ridge split and ridge expansion, has achieved optimal results in the maxilla . However, in the mandible, there is an elevated risk of block fracture post-osteotomy due to the dense cortical bone. To minimize the risk of fractures in these patients, apical and vertical horizontal cuts are executed. The ridge split procedure exhibits flexibility when immature bone callus forms at the cortical incision site. Certain scholars suggest that the ridge split can be conducted once the callus has matured (typically at 3 months) at the cortical incision site, resulting in improved primary stability of the implants . In general, the task of cutting cortical bone with manual instruments during the MBF operation is challenging, especially when it comes to achieving precise osteotomy. The employment of motor-driven equipment for bone cutting generates considerable heat, which can prolong healing times due to the adverse effects on surrounding tissues. Consequently, the piezoelectric osteotomy technique presented in this article offers two significant benefits. First and foremost, the precision of the cut is ensured, as the ultrasound frequency used does not harm the soft tissues. Secondly, the procedure is less invasive, leading to improved healing outcomes. This method yields a more predictable healing effect…. In conclusion, the MBF technique consistently improves the width of the alveolar bone and utilizes piezoelectric osteotomy to achieve the desired outcomes. When creating the "mucous bone flap," it is crucial to consider the labial bone thickness to ensure osteogenesis in the lip bone of the implant, thereby enhancing the implant restoration outcomes both in the short and long term . The MBF technique presents a novel strategy for addressing inadequate alveolar ridges. However, the study is not without its limitations, including a limited sample size, absence of a control group, and the necessity for an extended duration to monitor long-term outcomes. Moving forward, subsequent research will persist in comparing case numbers and other bone augmentation methodologies. |
Evaluating Postoperative Morbidity and Outcomes of Robotic-Assisted Esophagectomy in Esophageal Cancer Treatment—A Comprehensive Review on Behalf of TROGSS (The Robotic Global Surgical Society) and EFISDS (European Federation International Society for Digestive Surgery) Joint Working Group | 1adfc151-8fde-441f-b8fa-a816e269faba | 11854120 | Surgical Procedures, Operative[mh] | Esophagectomy remains a cornerstone in the curative treatment of esophageal cancer, the seventh most prevalent malignancy globally, despite significant advancements in multimodal approaches aimed at enhancing patient survival and quality of life . For instance, incorporating neoadjuvant chemotherapy or chemoradiotherapy has demonstrated improved outcomes, with 5-year survival rates reaching up to 50% compared to surgery alone . However, esophagectomy is a technically demanding procedure associated with substantial postoperative morbidity and mortality, which can negatively impact recovery, survival, and quality of life . To mitigate surgical trauma, minimally invasive esophagectomy (MIE), employing thoracoscopic or laparoscopic techniques, has gained widespread acceptance, now constituting over two-thirds of esophageal surgeries . These approaches have emerged as viable alternatives to traditional open surgery, offering significant benefits such as reduced postoperative pain, a lower incidence of pneumonia, and faster recovery times, all without compromising overall survival (OS) or disease-free survival (DFS) . Nonetheless, open esophagectomy continues to be preferred in certain clinical scenarios due to its shorter operative times while maintaining comparable oncologic outcomes . The advent of robotic-assisted surgery has further revolutionized minimally invasive techniques. The da Vinci Surgical System, approved by the U.S. Food and Drug Administration in 2000, marked a breakthrough in surgical innovation . In 2004, Kernstine and colleagues performed the first robotic-assisted esophagectomy, demonstrating its feasibility . Since then, the field of robotic surgery has expanded substantially with the introduction of newer platforms such as the Hugo™ system by Medtronic and the Versius system by CMR Surgical . These advancements have broadened the scope of minimally invasive surgery, providing surgeons with additional tools to optimize outcomes in complex esophageal resections. Despite the growing global adoption of robotic-assisted esophagectomy, it has yet to achieve widespread acceptance as a standard treatment modality for resectable esophageal cancer. Significant barriers to broader implementation include the high costs associated with acquiring and maintaining robotic systems, as well as a paucity of robust evidence unequivocally demonstrating its superiority over conventional surgical techniques . Consequently, many centers have integrated hybrid surgical approaches that combine minimally invasive and open techniques. A prevalent practice involves performing the abdominal phase laparoscopically while utilizing open thoracotomy for the thoracic phase, thereby leveraging the advantages of minimally invasive surgery, such as reduced postoperative pain and shorter hospitalization, while ensuring oncologically effective resections . This review aims to comprehensively assess the clinical and oncological outcomes of robotic-assisted esophagectomy. It seeks to elucidate emerging surgical trends, critically evaluate the advantages and limitations of robotic platforms, and analyze the learning curve associated with adopting this technology in the treatment of resectable esophageal cancer. We conducted a comprehensive literature search using multiple online databases, including PubMed, Embase, Web of Science, and the Cochrane Library, to identify studies on robotic-assisted esophagectomy (RAE) published up to June 2024. The search strategy utilized a combination of subject headings and text words to ensure broad yet focused retrieval of relevant articles. Keywords and medical subject heading (MeSH) terms included but were not limited to “robotic”, “esophagectomy”, “minimally invasive surgery”, “thoracoscopy”, “laparoscopy”, and “esophageal cancer”. In addition, the references of eligible articles were manually reviewed to identify supplementary studies that met the inclusion criteria. We included studies focusing on patients undergoing robotic-assisted esophagectomy for esophageal cancer, evaluating robotic-assisted techniques either as standalone procedures or as part of hybrid approaches. Comparative studies examining conventional approaches, such as minimally invasive esophagectomy (MIE) and open esophagectomy (OE), were also included. Eligible studies reported clinical outcomes such as overall survival and disease-free survival, perioperative metrics like operative time and blood loss, postoperative recovery, and complication rates. Retrospective and prospective studies, as well as case series and cohort studies involving ≥10 patients, were considered, provided they were published in English. Studies were excluded if they were reviews, systematic reviews, meta-analyses, or study protocols; case reports with fewer than 10 patients; studies lacking full-text availability; or non-human studies or those not reporting surgical or oncological outcomes. This focused yet inclusive methodology ensured a comprehensive analysis of the existing evidence base while maintaining relevance to the review objectives. Esophagectomy is widely recognized as one of the most challenging cancer surgeries, alongside pancreatectomy and hepatectomy. Despite advancements in perioperative care, surgical techniques, and anesthetic protocols that have contributed to reduced complication rates, esophagectomy continues to be a formidable procedure . The average five-year survival rate for operable esophageal cancer remains approximately 28%, underscoring the significant impact of surgical complications on patients’ quality of life, particularly in the context of a limited life expectancy . To standardize the reporting and stratification of postoperative complications, several international organizations, including the Esophagectomy Complications Consensus Group (ECCG), the Dutch Upper Gastrointestinal Cancer Audit (DUCA), and the Oesophago-Gastric Anastomosis Audit (OGAA), have established outcome measures . The Clavien–Dindo (CD) classification system remains a universally accepted tool among these groups for categorizing postoperative complications . The Traditional Invasive vs. Minimally Invasive Esophagectomy (TIME) trial in 2012 marked a pivotal shift toward the adoption of minimally invasive techniques in esophagectomy . Subsequent prospective studies, such as those conducted by the DUCA group in 2017, demonstrated improved “textbook outcomes” for minimally invasive esophagectomy compared to the open approach (odds ratio: 1.60 [1.31–1.94], p = 0.004) . Robotic-assisted minimally invasive esophagectomy (RAMIE) has since gained traction due to its promising initial results. However, debates persist regarding its superiority over conventional minimally invasive esophagectomy (cMIE). Two key randomized controlled trials (RCTs) have sought to address this question. The Dutch ROBOT trial aimed to detect a 22% absolute risk reduction in CD grade ≥ 2 complications with RAMIE compared to open transthoracic esophagectomy (OTE). RAMIE demonstrated a lower rate of surgery-related postoperative complications (CD grade ≥2) compared to OTE (59% vs. 80%; RR 0.74, 95% CI: 0.57–0.96, p = 0.02) . In contrast, the multicenter RAMIE trial conducted by a Chinese group compared RAMIE to cMIE for esophageal squamous cell carcinoma. Their findings revealed no significant difference in CD grade ≥ 2 complications between RAMIE (12.2%) and cMIE (10.2%) (RR 1.20, 95% CI: 0.66–2.15, p = 0.551) . Meta-analyses have provided additional insights into postoperative outcomes. Esagian et al. (2022) included five retrospective studies, four prospective studies, and one RCT comparing RAMIE with OTE. The analysis found no statistically significant difference in overall complication rates between the RAMIE group (27.88%) and the OTE group (33.93%) (OR: 0.66, 95% CI: 0.42–1.05, p = 0.08) . Perry et al. further expanded this analysis to include 18,187 patients, offering a broader perspective on short-term outcomes between RAMIE and cMIE . Specific surgical approaches have also been analyzed. Zhou et al. (2022) focused on McKeown esophagectomy, pooling data from seven propensity-matched retrospective studies. Their findings indicated no significant difference in overall complications between RAMIE and cMIE (OR: 1.10, 95% CI: 0.86–1.41, p = 0.46) . Meanwhile, Angeramo et al. conducted a meta-analysis of 60 studies examining Ivor Lewis esophagectomy, reporting significantly lower overall morbidity rates with RAMIE (30%, 95% CI: 24–38%) compared to cMIE (40%, 95% CI: 34–47%; OR: 0.67, 95% CI: 0.58–0.79, p < 0.001) . Real-world data from the OGAA group have highlighted postoperative outcomes on a global scale, encompassing over 137 countries. Their analysis emphasized the critical role of anastomotic leaks and conduit necrosis in influencing patient mortality, providing a comprehensive overview of postoperative risks and outcomes . 3.1. Clinical Recovery Metrics 3.1.1. Length of Hospital Stay The length of hospital stay (LOS) is a critical postoperative outcome indicator, often reflecting the overall recovery and complication rates. In the ROBOT trial, the median ICU stay was reported as one day in both the RAMIE and OTE groups ( p = 0.45). The median hospital stay was slightly shorter in the RAMIE group (14 days) compared to the OTE group (16 days), but this difference was not statistically significant ( p = 0.33) . A meta-analysis by Esagian et al. demonstrated a significantly shorter LOS for RAMIE compared to OTE. The mean LOS was 17.10 ± 9.39 days for RAMIE versus 30.68 ± 23.88 days for OTE, with a weighted mean difference (WMD) of −9.22 days (95% CI: −14.39 to −4.06; p < 0.001) . In the RAMIE trial, the median ICU stay remained consistent at one day across both RAMIE (range 0–15) and MIE (range 0–14) groups ( p = 0.990). Similarly, the median postoperative hospital stay was nine days for both groups (RAMIE: range 6–49; MIE: range 6–82; p = 0.311). Additionally, the readmission rate to the ICU was identical between the groups (1.7% each; p = 0.815) . The meta-analysis by Perry et al., encompassing 26 studies, further supported the shorter LOS associated with RAMIE. The mean LOS was 18.57 days for RAMIE versus 33.11 days for cMIE, yielding a statistically significant mean difference of −3.03 days (95% CI: −4.51 to −1.54; p < 0.0001; I 2 = 96%; p < 0.00001) . Finally, the meta-analysis by Zhou et al. specifically focused on McKeown esophagectomy, highlighting a shorter postoperative hospital stay for RAMIE compared to cMIE (MD = 1.05 days, 95% CI: 0.05–2.05; p = 0.04) . 3.1.2. Functional Recovery Postoperative functional recovery is a critical outcome measure following esophagectomy. In the ROBOT trial, functional recovery was defined as the successful removal of thoracic tubes, absence of intravenous fluid resuscitation requirements, tolerance for solid oral intake, independent mobilization, and adequate pain control with oral analgesics. At postoperative day 14, significantly more patients in the RAMIE group achieved functional recovery compared to those in the OTE group (38/54, 70% vs. 28/55, 51%; RR 1.48, 95% CI: 1.03–2.13; p = 0.04) . These findings underscore the potential advantages of robotic-assisted techniques in promoting faster recovery and reducing the burden of postoperative care. 3.1.3. Postoperative Pain Management In the ROBOT trial, postoperative pain was assessed using a visual analogue scale (VAS) ranging from 1 to 10, with measurements taken preoperatively and daily during the first 14 days post-surgery. Patients in the RAMIE group reported significantly lower mean pain scores compared to the OTE group (1.86 vs. 2.62; p < 0.001). Furthermore, short-term quality-of-life (QoL) outcomes were superior in the RAMIE group at both discharge and six weeks post-discharge. The mean difference in QoL scores between the groups was 13.4 (95% CI: 2.0–24.7; p = 0.02) at discharge and 11.1 (95% CI: 1.0–21.1; p = 0.03) at six weeks post-discharge. Physical functioning also demonstrated significant improvement in the RAMIE group compared to the OTE group, with mean differences of 13.5 (95% CI: 1.2–25.7; p = 0.03) at discharge and 10.7 (95% CI: 0.04–21.4; p = 0.049) at six weeks. These findings highlight the role of robotic-assisted techniques in reducing postoperative discomfort and enhancing the overall recovery experience for patients . 3.2. Complications 3.2.1. Pulmonary Complications Robotic-assisted surgery offers superior anatomical visualization and enhanced dexterity, enabling precise preservation of parasympathetic lung innervation through sparing of vagal branches. This has been linked to a reduction in pulmonary complications . In the ROBOT trial, pulmonary complications were significantly lower in the RAMIE group (17/54 patients, 32%) compared to the OTE group (32/55 patients, 58%), with a relative risk (RR) of 0.54 (95% CI: 0.34–0.85; p = 0.005) . Similarly, the meta-analysis by Esagian et al. revealed a significantly lower overall pulmonary complication rate in the RAMIE group (14.29%, 49/343) versus the OTE group (25.32%, 174/687), with an odds ratio (OR) of 0.38 (95% CI: 0.26–0.56; p < 0.001) . However, in the RAMIE trial, the incidence of pulmonary complications was comparable between RAMIE (13.8%) and MIE (14.7%) (RR 0.94, 95% CI: 0.57–1.56; p = 0.812). No significant differences were observed in the rates of pneumonia (9.9% in RAMIE vs. 11.9% in MIE; p = 0.560) or respiratory failure (4.4% in RAMIE vs. 5.1% in MIE; p = 0.767) . The meta-analysis by Perry et al. analyzed 10,154 patients (RAMIE: 3185; cMIE: 6969) and found no statistically significant difference in pulmonary complication rates between RAMIE (20.13%) and cMIE (22.20%) (RR 0.89, 95% CI: 0.77–1.02; p = 0.10; I 2 = 18%; p = 0.20) . In contrast, the meta-analysis by Zhou et al. reported a considerably lower risk of pneumonia following RAMIE compared to cMIE (OR 0.72, 95% CI: 0.52–1.00; p = 0.05) . Similarly, a meta-analysis comparing Ivor Lewis RAMIE with cMIE found a lower weighted pooled proportion of postoperative pneumonia in RAMIE (8%, 95% CI: 6–9) than in cMIE (10%, 95% CI: 7–13). Patients undergoing RAMIE demonstrated a significantly reduced risk of pneumonia (OR 0.46, 95% CI: 0.35–0.61; p < 0.0001) . 3.2.2. Cardiac Complications Cardiac complications, particularly new-onset atrial fibrillation (AF), are associated with significant postoperative morbidity and may serve as surrogate markers for underlying complications. The reduced incidence of cardiac complications in robotic-assisted minimally invasive esophagectomy (RAMIE) may be attributed to decreased intravascular depletion due to less blood loss, reduced oxidative stress from superior lung ventilation in the prone position, and fewer infectious complications compared to open transthoracic esophagectomy (OTE) . In the ROBOT trial, atrial fibrillation occurred in 22% of patients in the RAMIE group (17/45) compared to 47% in the OTE group (26/55), yielding a relative risk (RR) of 0.47 (95% CI: 0.27–0.83; p = 0.006) . Similarly, the meta-analysis by Esagian et al., which included data from five studies , reported a significantly lower rate of atrial fibrillation in the RAMIE group (6.79%, 29/427) compared to the OTE group (8.46%, 54/638), with an odds ratio (OR) of 0.53 (95% CI: 0.29–0.98; p = 0.04) . Conversely, the RAMIE trial found comparable rates of cardiac complications in the RAMIE and conventional minimally invasive esophagectomy (cMIE) groups (two cases vs. one case, p = 0.8) . The meta-analysis by Perry et al. further supported this finding, with cardiac complication rates of 14.02% (365/2604) in the RAMIE group and 15.74% (823/5228) in the cMIE group. No statistically significant difference was observed between the two groups (RR 1.01, 95% CI: 0.86–1.19; p = 0.88; I 2 = 3%, p = 0.42) . 3.2.3. Anastomotic Leak The Esophageal Complications Consensus Group (ECCG) defines an anastomotic leak as a “full-thickness gastrointestinal defect involving the esophagus, anastomosis, staple line, or conduit irrespective of presentation or method of identification.” The ECCG further classifies leaks into three types based on the intervention required for management . The technical challenges in crafting a gastric conduit during McKeown esophagectomy and creating an intrathoracic anastomosis in Ivor Lewis esophagectomy are exacerbated in conventional minimally invasive esophagectomy (cMIE) due to the limited maneuverability of laparoscopic instruments. Robotic-assisted minimally invasive esophagectomy (RAMIE) addresses these challenges with enhanced ergonomics, tremor filtration, and the precise control offered by EndoWrist instruments . In a retrospective analysis by de Groot et al., 26% of 152 patients undergoing RAMIE with intrathoracic anastomosis experienced anastomotic leaks. Of these 40 patients, 28% had management failure, including seven cases of esophagobronchial fistula, three cases of anastomotic disconnection, and one death due to septic bleeding . Similarly, Khaitan et al., using data from the Society of Thoracic Surgeons General Thoracic Surgery Database, found that RAMIE was independently associated with higher rates of anastomotic leak compared to OTE (adjusted odds ratio [aOR] 1.53, 95% CI: 1.14–2.04) . In a propensity-matched analysis of 1320 RAMIE cases versus cMIE, the rate of anastomotic leaks requiring surgery was higher in the RAMIE group (aOR 1.39, 95% CI: 1.01–1.92; p = 0.045). Excessive handling of the gastric conduit by robotic instruments, coupled with lower case volumes and the learning curve of surgeons, was suggested as a potential reason for the higher leak rates . In the ROBOT trial, type III anastomotic leaks occurred in 22% of patients in the RAMIE arm and 20% in the OTE arm ( p = 0.57) . A meta-analysis by Esagian et al. found no statistically significant difference in leak rates between RAMIE (6.82%; 46/674) and OTE (6.06%; 79/1303), with an odds ratio of 0.93 (95% CI: 0.60–1.44; p = 0.76) . Similarly, the RAMIE trial reported comparable rates of anastomotic leakage between RAMIE (12.2%) and cMIE (11.3%) (RR 1.08; 95% CI: 0.61–1.90; p = 0.801), with only one patient in each group requiring surgical intervention for type III leakage . The meta-analysis by Perry et al. reported anastomotic leak rates of 12.47% (391/3136) in the RAMIE group and 11.43% (785/6866) in the cMIE group, favoring cMIE (RR 1.23; 95% CI: 1.09–1.38; p = 0.0005; I 2 = 0%, p = 0.64) . A meta-analysis comparing Ivor Lewis cMIE and RAMIE found comparable leak rates, with 7% (95% CI: 6–9%) in the cMIE group and 7% (95% CI: 5–11%) in the RAMIE group. The odds of developing an anastomotic leak were similar between the two groups (OR 0.85; 95% CI: 0.65–1.10; p = 0.22) . 3.2.4. Gastric Conduit Necrosis The Esophageal Complications Consensus Group (ECCG) classifies gastric conduit necrosis into three types based on the extent of necrosis and the requirement for surgical management and diversion . In the ROBOT trial, type III gastric conduit necrosis was observed in one patient in the RAMIE arm and two patients in the OTE arm ( p = 1.00) . 3.2.5. Thoracic Duct Chylothorax, a significant complication following esophagectomy, is associated with substantial morbidity, often requiring prolonged management and impacting patient recovery. The identification of the thoracic duct is critical in minimizing the risk of chylothorax, and recent advances in near-infrared fluorescence (NIRF) imaging have improved the precision of this process. The use of NIRF for thoracic duct identification has been well established in non-robotic esophageal surgeries , and its integration into robotic-assisted procedures holds promise for enhancing surgical outcomes. A study by Vecchiato et al. in 2020 proposed an innovative approach for detecting the thoracic duct during robotic or laparoscopic surgery. This technique involved intranodal injection of indocyanine green (ICG) into the inguinal nodes under ultrasound guidance, followed by robotic or laparoscopic detection of the thoracic duct in 21 patients. The procedure was successful in identifying the thoracic duct in all cases, with one intraoperative injury to the duct that was promptly clipped. Importantly, no postoperative chylothorax or adverse reactions at the injection site were observed, underscoring the safety and effectiveness of this technique . In addition, Jardinet et al. demonstrated the feasibility of this approach with intrainguinal node ICG injection, offering easy identification of the thoracic duct during surgery. This video publication further reinforces the potential of NIRF-guided identification to prevent thoracic duct injury and subsequent complications such as chylothorax . These findings suggest that NIRF imaging, combined with the intranodal injection of ICG, is a promising method for safely identifying the thoracic duct in robotic esophagectomy. By minimizing the risk of injury and improving the precision of lymphatic structure identification, this approach may reduce postoperative complications and improve patient outcomes. Further studies are necessary to refine this technique and assess its broader applicability in clinical practice. 3.2.6. Chyle Leak The ECCG defines chyle leaks by classifying them into three types and grades of severity based on daily output . Dezube et al., in a retrospective analysis, compared 70 RAMIE procedures with 277 cMIE procedures, reporting a significantly higher incidence of chyle leaks in the RAMIE group compared to the cMIE group (12.8% vs. 3.6%, p = 0.006) . Notably, McKeown RAMIE was associated with a higher incidence of chyle leaks (33%) than Ivor Lewis RAMIE (4%). In the ROBOT trial, chyle leaks were documented in 17 patients in the RAMIE arm and 12 patients in the OTE arm, with no statistically significant difference ( p = 0.69) . Similarly, the meta-analysis by Esagian et al. showed no significant difference in the incidence of chylothorax between the RAMIE group (5.39%; 29/538) and the OTE group (3.01%; 33/1095), with an odds ratio of 1.31 (95% CI: 0.75–2.29; p = 0.35) . The RAMIE trial also reported comparable rates of chyle leak between RAMIE (five cases) and cMIE (two cases) ( p = 0.449) . Perry et al., in their meta-analysis, found no statistically significant difference in chyle leak rates between the two groups (RR 1.07; 95% CI: 0.72–1.60; p = 0.74; I 2 = 13%, p = 0.30). The leak rates were 2.82% (69/2443) in the RAMIE group and 3.84% (197/5135) in the cMIE group . Similarly, Zhou et al. observed no significant difference in pooled data analysis between RAMIE and cMIE approaches (OR: 1.33; 95% CI: 0.57–3.07; p = 0.51) . 3.2.7. Recurrent Laryngeal Nerve (RLN) Injury During esophagectomy, the recurrent laryngeal nerve (RLN) is vulnerable to injury due to thermal damage, stretching, compression, or vascular compromise, potentially leading to vocal cord palsy. Such injuries can significantly increase the risk of pulmonary complications, ICU readmissions, and prolonged hospital stays . The ECCG defines vocal cord injury as “vocal cord dysfunction post-resection, confirmed and assessed by direct examination”, with severity graded by laterality and intervention needed, categorized into three types . Robotics has been hypothesized to reduce RLN injuries due to improved visualization of the nerve and adjacent vascular structures. A retrospective analysis by Scholtemeijer et al. on McKeown esophagectomies reported an RLN injury incidence of 14% among 266 patients undergoing robotic-assisted minimally invasive esophagectomy (RAMIE) . In the ROBOT trial, type 1 vocal cord injury was observed in five patients in the RAMIE arm and six patients in the OTE arm, with no statistically significant difference ( p = 0.78) . Similarly, the meta-analysis by Esagian et al., which included seven studies, found no significant difference in RLN palsy rates between RAMIE (13.99%; 67/479) and OTE (10.41%; 84/807) groups (OR: 1.31, 95% CI: 0.90–1.90; p = 0.16) . In the RAMIE trial, RAMIE was associated with a higher rate of vocal cord paralysis compared to cMIE (32.6% vs. 27.1%), although the difference was not statistically significant (RR 1.20; 95% CI: 0.87–1.66; p = 0.258) . The largest meta-analysis, conducted by Perry et al., included over 18,000 patients and reported no significant difference in RLN injury rates between RAMIE (8.94%; 237/2652) and cMIE (7.63%; 423/5541). The relative risk was 0.96 (95% CI: 0.82–1.13; p = 0.62; I 2 = 7%; p = 0.36) . 3.2.8. Para-Conduit Diaphragmatic Herniations Para-conduit diaphragmatic herniation is a relatively rare complication following open esophagectomy. However, with the increasing adoption of minimally invasive techniques, the incidence of this complication has significantly risen. This trend is likely attributable to the reduction in adhesions between the gastric conduit and the diaphragmatic hiatus inherent to minimally invasive approaches. In a retrospective analysis by De Silva et al. , the incidence of para-conduit diaphragmatic herniation was found to be significantly higher with minimally invasive techniques (laparoscopic and robotic approaches) compared to open esophagectomy (70.8% vs. 35.5%; p < 0.001). To reduce the risk of herniation, it has been suggested that surgeons place two or three interrupted sutures between the gastric conduit and the right hemidiaphragm after completing the anastomosis. This maneuver may provide additional stability and prevent the herniation of abdominal contents into the thoracic cavity . 3.3. Postoperative Mortality In the ROBOT trial, two patients on the RAMIE arm and one patient on the OTE arm died in the immediate postoperative period, and 30-day and 90-day mortality rates were 2% and 9% in the RAMIE arm vs. 0% and 2% in the OTE arm . In the RAMIE trial, for 30-day mortality, one patient in MIE died from acute cerebral infarction on POD 12. For 90-day mortality, one patient in RAMIE died from severe pneumonia on POD 42 . In the meta-analysis by Perry et al., the 30-day mortality rate for each procedure was 1.63% (44/2707) in the RAMIE group and 1.87% (117/6244) in the cMIE group. There was no statistically significant difference between the two groups (RR 1.03, p = 0.88 [95% CI 0.73, 1.44], I 2 = 0%, p = 0.53). The rate of 90-day mortality was 3.55% (106/2987) in the RAMIE group and 4.84% (336/6946) in the cMIE group, showing no statistically significant difference between the two groups (RR 0.95, p = 0.66 [95% CI 0.77, 1.18], I 2 = 0%, p = 0.93) . In the meta-analysis by Zhou et al., the in-hospital mortality and 90-day mortality had no statistically distinguished difference in a merged data analysis (OR = 0.54, 95 CI [0.14, 2.02] p = 0.36) and (OR = 0.69, 95 CI [0.26, 1.83] p = 0.46), respectively . 3.1.1. Length of Hospital Stay The length of hospital stay (LOS) is a critical postoperative outcome indicator, often reflecting the overall recovery and complication rates. In the ROBOT trial, the median ICU stay was reported as one day in both the RAMIE and OTE groups ( p = 0.45). The median hospital stay was slightly shorter in the RAMIE group (14 days) compared to the OTE group (16 days), but this difference was not statistically significant ( p = 0.33) . A meta-analysis by Esagian et al. demonstrated a significantly shorter LOS for RAMIE compared to OTE. The mean LOS was 17.10 ± 9.39 days for RAMIE versus 30.68 ± 23.88 days for OTE, with a weighted mean difference (WMD) of −9.22 days (95% CI: −14.39 to −4.06; p < 0.001) . In the RAMIE trial, the median ICU stay remained consistent at one day across both RAMIE (range 0–15) and MIE (range 0–14) groups ( p = 0.990). Similarly, the median postoperative hospital stay was nine days for both groups (RAMIE: range 6–49; MIE: range 6–82; p = 0.311). Additionally, the readmission rate to the ICU was identical between the groups (1.7% each; p = 0.815) . The meta-analysis by Perry et al., encompassing 26 studies, further supported the shorter LOS associated with RAMIE. The mean LOS was 18.57 days for RAMIE versus 33.11 days for cMIE, yielding a statistically significant mean difference of −3.03 days (95% CI: −4.51 to −1.54; p < 0.0001; I 2 = 96%; p < 0.00001) . Finally, the meta-analysis by Zhou et al. specifically focused on McKeown esophagectomy, highlighting a shorter postoperative hospital stay for RAMIE compared to cMIE (MD = 1.05 days, 95% CI: 0.05–2.05; p = 0.04) . 3.1.2. Functional Recovery Postoperative functional recovery is a critical outcome measure following esophagectomy. In the ROBOT trial, functional recovery was defined as the successful removal of thoracic tubes, absence of intravenous fluid resuscitation requirements, tolerance for solid oral intake, independent mobilization, and adequate pain control with oral analgesics. At postoperative day 14, significantly more patients in the RAMIE group achieved functional recovery compared to those in the OTE group (38/54, 70% vs. 28/55, 51%; RR 1.48, 95% CI: 1.03–2.13; p = 0.04) . These findings underscore the potential advantages of robotic-assisted techniques in promoting faster recovery and reducing the burden of postoperative care. 3.1.3. Postoperative Pain Management In the ROBOT trial, postoperative pain was assessed using a visual analogue scale (VAS) ranging from 1 to 10, with measurements taken preoperatively and daily during the first 14 days post-surgery. Patients in the RAMIE group reported significantly lower mean pain scores compared to the OTE group (1.86 vs. 2.62; p < 0.001). Furthermore, short-term quality-of-life (QoL) outcomes were superior in the RAMIE group at both discharge and six weeks post-discharge. The mean difference in QoL scores between the groups was 13.4 (95% CI: 2.0–24.7; p = 0.02) at discharge and 11.1 (95% CI: 1.0–21.1; p = 0.03) at six weeks post-discharge. Physical functioning also demonstrated significant improvement in the RAMIE group compared to the OTE group, with mean differences of 13.5 (95% CI: 1.2–25.7; p = 0.03) at discharge and 10.7 (95% CI: 0.04–21.4; p = 0.049) at six weeks. These findings highlight the role of robotic-assisted techniques in reducing postoperative discomfort and enhancing the overall recovery experience for patients . The length of hospital stay (LOS) is a critical postoperative outcome indicator, often reflecting the overall recovery and complication rates. In the ROBOT trial, the median ICU stay was reported as one day in both the RAMIE and OTE groups ( p = 0.45). The median hospital stay was slightly shorter in the RAMIE group (14 days) compared to the OTE group (16 days), but this difference was not statistically significant ( p = 0.33) . A meta-analysis by Esagian et al. demonstrated a significantly shorter LOS for RAMIE compared to OTE. The mean LOS was 17.10 ± 9.39 days for RAMIE versus 30.68 ± 23.88 days for OTE, with a weighted mean difference (WMD) of −9.22 days (95% CI: −14.39 to −4.06; p < 0.001) . In the RAMIE trial, the median ICU stay remained consistent at one day across both RAMIE (range 0–15) and MIE (range 0–14) groups ( p = 0.990). Similarly, the median postoperative hospital stay was nine days for both groups (RAMIE: range 6–49; MIE: range 6–82; p = 0.311). Additionally, the readmission rate to the ICU was identical between the groups (1.7% each; p = 0.815) . The meta-analysis by Perry et al., encompassing 26 studies, further supported the shorter LOS associated with RAMIE. The mean LOS was 18.57 days for RAMIE versus 33.11 days for cMIE, yielding a statistically significant mean difference of −3.03 days (95% CI: −4.51 to −1.54; p < 0.0001; I 2 = 96%; p < 0.00001) . Finally, the meta-analysis by Zhou et al. specifically focused on McKeown esophagectomy, highlighting a shorter postoperative hospital stay for RAMIE compared to cMIE (MD = 1.05 days, 95% CI: 0.05–2.05; p = 0.04) . Postoperative functional recovery is a critical outcome measure following esophagectomy. In the ROBOT trial, functional recovery was defined as the successful removal of thoracic tubes, absence of intravenous fluid resuscitation requirements, tolerance for solid oral intake, independent mobilization, and adequate pain control with oral analgesics. At postoperative day 14, significantly more patients in the RAMIE group achieved functional recovery compared to those in the OTE group (38/54, 70% vs. 28/55, 51%; RR 1.48, 95% CI: 1.03–2.13; p = 0.04) . These findings underscore the potential advantages of robotic-assisted techniques in promoting faster recovery and reducing the burden of postoperative care. In the ROBOT trial, postoperative pain was assessed using a visual analogue scale (VAS) ranging from 1 to 10, with measurements taken preoperatively and daily during the first 14 days post-surgery. Patients in the RAMIE group reported significantly lower mean pain scores compared to the OTE group (1.86 vs. 2.62; p < 0.001). Furthermore, short-term quality-of-life (QoL) outcomes were superior in the RAMIE group at both discharge and six weeks post-discharge. The mean difference in QoL scores between the groups was 13.4 (95% CI: 2.0–24.7; p = 0.02) at discharge and 11.1 (95% CI: 1.0–21.1; p = 0.03) at six weeks post-discharge. Physical functioning also demonstrated significant improvement in the RAMIE group compared to the OTE group, with mean differences of 13.5 (95% CI: 1.2–25.7; p = 0.03) at discharge and 10.7 (95% CI: 0.04–21.4; p = 0.049) at six weeks. These findings highlight the role of robotic-assisted techniques in reducing postoperative discomfort and enhancing the overall recovery experience for patients . 3.2.1. Pulmonary Complications Robotic-assisted surgery offers superior anatomical visualization and enhanced dexterity, enabling precise preservation of parasympathetic lung innervation through sparing of vagal branches. This has been linked to a reduction in pulmonary complications . In the ROBOT trial, pulmonary complications were significantly lower in the RAMIE group (17/54 patients, 32%) compared to the OTE group (32/55 patients, 58%), with a relative risk (RR) of 0.54 (95% CI: 0.34–0.85; p = 0.005) . Similarly, the meta-analysis by Esagian et al. revealed a significantly lower overall pulmonary complication rate in the RAMIE group (14.29%, 49/343) versus the OTE group (25.32%, 174/687), with an odds ratio (OR) of 0.38 (95% CI: 0.26–0.56; p < 0.001) . However, in the RAMIE trial, the incidence of pulmonary complications was comparable between RAMIE (13.8%) and MIE (14.7%) (RR 0.94, 95% CI: 0.57–1.56; p = 0.812). No significant differences were observed in the rates of pneumonia (9.9% in RAMIE vs. 11.9% in MIE; p = 0.560) or respiratory failure (4.4% in RAMIE vs. 5.1% in MIE; p = 0.767) . The meta-analysis by Perry et al. analyzed 10,154 patients (RAMIE: 3185; cMIE: 6969) and found no statistically significant difference in pulmonary complication rates between RAMIE (20.13%) and cMIE (22.20%) (RR 0.89, 95% CI: 0.77–1.02; p = 0.10; I 2 = 18%; p = 0.20) . In contrast, the meta-analysis by Zhou et al. reported a considerably lower risk of pneumonia following RAMIE compared to cMIE (OR 0.72, 95% CI: 0.52–1.00; p = 0.05) . Similarly, a meta-analysis comparing Ivor Lewis RAMIE with cMIE found a lower weighted pooled proportion of postoperative pneumonia in RAMIE (8%, 95% CI: 6–9) than in cMIE (10%, 95% CI: 7–13). Patients undergoing RAMIE demonstrated a significantly reduced risk of pneumonia (OR 0.46, 95% CI: 0.35–0.61; p < 0.0001) . 3.2.2. Cardiac Complications Cardiac complications, particularly new-onset atrial fibrillation (AF), are associated with significant postoperative morbidity and may serve as surrogate markers for underlying complications. The reduced incidence of cardiac complications in robotic-assisted minimally invasive esophagectomy (RAMIE) may be attributed to decreased intravascular depletion due to less blood loss, reduced oxidative stress from superior lung ventilation in the prone position, and fewer infectious complications compared to open transthoracic esophagectomy (OTE) . In the ROBOT trial, atrial fibrillation occurred in 22% of patients in the RAMIE group (17/45) compared to 47% in the OTE group (26/55), yielding a relative risk (RR) of 0.47 (95% CI: 0.27–0.83; p = 0.006) . Similarly, the meta-analysis by Esagian et al., which included data from five studies , reported a significantly lower rate of atrial fibrillation in the RAMIE group (6.79%, 29/427) compared to the OTE group (8.46%, 54/638), with an odds ratio (OR) of 0.53 (95% CI: 0.29–0.98; p = 0.04) . Conversely, the RAMIE trial found comparable rates of cardiac complications in the RAMIE and conventional minimally invasive esophagectomy (cMIE) groups (two cases vs. one case, p = 0.8) . The meta-analysis by Perry et al. further supported this finding, with cardiac complication rates of 14.02% (365/2604) in the RAMIE group and 15.74% (823/5228) in the cMIE group. No statistically significant difference was observed between the two groups (RR 1.01, 95% CI: 0.86–1.19; p = 0.88; I 2 = 3%, p = 0.42) . 3.2.3. Anastomotic Leak The Esophageal Complications Consensus Group (ECCG) defines an anastomotic leak as a “full-thickness gastrointestinal defect involving the esophagus, anastomosis, staple line, or conduit irrespective of presentation or method of identification.” The ECCG further classifies leaks into three types based on the intervention required for management . The technical challenges in crafting a gastric conduit during McKeown esophagectomy and creating an intrathoracic anastomosis in Ivor Lewis esophagectomy are exacerbated in conventional minimally invasive esophagectomy (cMIE) due to the limited maneuverability of laparoscopic instruments. Robotic-assisted minimally invasive esophagectomy (RAMIE) addresses these challenges with enhanced ergonomics, tremor filtration, and the precise control offered by EndoWrist instruments . In a retrospective analysis by de Groot et al., 26% of 152 patients undergoing RAMIE with intrathoracic anastomosis experienced anastomotic leaks. Of these 40 patients, 28% had management failure, including seven cases of esophagobronchial fistula, three cases of anastomotic disconnection, and one death due to septic bleeding . Similarly, Khaitan et al., using data from the Society of Thoracic Surgeons General Thoracic Surgery Database, found that RAMIE was independently associated with higher rates of anastomotic leak compared to OTE (adjusted odds ratio [aOR] 1.53, 95% CI: 1.14–2.04) . In a propensity-matched analysis of 1320 RAMIE cases versus cMIE, the rate of anastomotic leaks requiring surgery was higher in the RAMIE group (aOR 1.39, 95% CI: 1.01–1.92; p = 0.045). Excessive handling of the gastric conduit by robotic instruments, coupled with lower case volumes and the learning curve of surgeons, was suggested as a potential reason for the higher leak rates . In the ROBOT trial, type III anastomotic leaks occurred in 22% of patients in the RAMIE arm and 20% in the OTE arm ( p = 0.57) . A meta-analysis by Esagian et al. found no statistically significant difference in leak rates between RAMIE (6.82%; 46/674) and OTE (6.06%; 79/1303), with an odds ratio of 0.93 (95% CI: 0.60–1.44; p = 0.76) . Similarly, the RAMIE trial reported comparable rates of anastomotic leakage between RAMIE (12.2%) and cMIE (11.3%) (RR 1.08; 95% CI: 0.61–1.90; p = 0.801), with only one patient in each group requiring surgical intervention for type III leakage . The meta-analysis by Perry et al. reported anastomotic leak rates of 12.47% (391/3136) in the RAMIE group and 11.43% (785/6866) in the cMIE group, favoring cMIE (RR 1.23; 95% CI: 1.09–1.38; p = 0.0005; I 2 = 0%, p = 0.64) . A meta-analysis comparing Ivor Lewis cMIE and RAMIE found comparable leak rates, with 7% (95% CI: 6–9%) in the cMIE group and 7% (95% CI: 5–11%) in the RAMIE group. The odds of developing an anastomotic leak were similar between the two groups (OR 0.85; 95% CI: 0.65–1.10; p = 0.22) . 3.2.4. Gastric Conduit Necrosis The Esophageal Complications Consensus Group (ECCG) classifies gastric conduit necrosis into three types based on the extent of necrosis and the requirement for surgical management and diversion . In the ROBOT trial, type III gastric conduit necrosis was observed in one patient in the RAMIE arm and two patients in the OTE arm ( p = 1.00) . 3.2.5. Thoracic Duct Chylothorax, a significant complication following esophagectomy, is associated with substantial morbidity, often requiring prolonged management and impacting patient recovery. The identification of the thoracic duct is critical in minimizing the risk of chylothorax, and recent advances in near-infrared fluorescence (NIRF) imaging have improved the precision of this process. The use of NIRF for thoracic duct identification has been well established in non-robotic esophageal surgeries , and its integration into robotic-assisted procedures holds promise for enhancing surgical outcomes. A study by Vecchiato et al. in 2020 proposed an innovative approach for detecting the thoracic duct during robotic or laparoscopic surgery. This technique involved intranodal injection of indocyanine green (ICG) into the inguinal nodes under ultrasound guidance, followed by robotic or laparoscopic detection of the thoracic duct in 21 patients. The procedure was successful in identifying the thoracic duct in all cases, with one intraoperative injury to the duct that was promptly clipped. Importantly, no postoperative chylothorax or adverse reactions at the injection site were observed, underscoring the safety and effectiveness of this technique . In addition, Jardinet et al. demonstrated the feasibility of this approach with intrainguinal node ICG injection, offering easy identification of the thoracic duct during surgery. This video publication further reinforces the potential of NIRF-guided identification to prevent thoracic duct injury and subsequent complications such as chylothorax . These findings suggest that NIRF imaging, combined with the intranodal injection of ICG, is a promising method for safely identifying the thoracic duct in robotic esophagectomy. By minimizing the risk of injury and improving the precision of lymphatic structure identification, this approach may reduce postoperative complications and improve patient outcomes. Further studies are necessary to refine this technique and assess its broader applicability in clinical practice. 3.2.6. Chyle Leak The ECCG defines chyle leaks by classifying them into three types and grades of severity based on daily output . Dezube et al., in a retrospective analysis, compared 70 RAMIE procedures with 277 cMIE procedures, reporting a significantly higher incidence of chyle leaks in the RAMIE group compared to the cMIE group (12.8% vs. 3.6%, p = 0.006) . Notably, McKeown RAMIE was associated with a higher incidence of chyle leaks (33%) than Ivor Lewis RAMIE (4%). In the ROBOT trial, chyle leaks were documented in 17 patients in the RAMIE arm and 12 patients in the OTE arm, with no statistically significant difference ( p = 0.69) . Similarly, the meta-analysis by Esagian et al. showed no significant difference in the incidence of chylothorax between the RAMIE group (5.39%; 29/538) and the OTE group (3.01%; 33/1095), with an odds ratio of 1.31 (95% CI: 0.75–2.29; p = 0.35) . The RAMIE trial also reported comparable rates of chyle leak between RAMIE (five cases) and cMIE (two cases) ( p = 0.449) . Perry et al., in their meta-analysis, found no statistically significant difference in chyle leak rates between the two groups (RR 1.07; 95% CI: 0.72–1.60; p = 0.74; I 2 = 13%, p = 0.30). The leak rates were 2.82% (69/2443) in the RAMIE group and 3.84% (197/5135) in the cMIE group . Similarly, Zhou et al. observed no significant difference in pooled data analysis between RAMIE and cMIE approaches (OR: 1.33; 95% CI: 0.57–3.07; p = 0.51) . 3.2.7. Recurrent Laryngeal Nerve (RLN) Injury During esophagectomy, the recurrent laryngeal nerve (RLN) is vulnerable to injury due to thermal damage, stretching, compression, or vascular compromise, potentially leading to vocal cord palsy. Such injuries can significantly increase the risk of pulmonary complications, ICU readmissions, and prolonged hospital stays . The ECCG defines vocal cord injury as “vocal cord dysfunction post-resection, confirmed and assessed by direct examination”, with severity graded by laterality and intervention needed, categorized into three types . Robotics has been hypothesized to reduce RLN injuries due to improved visualization of the nerve and adjacent vascular structures. A retrospective analysis by Scholtemeijer et al. on McKeown esophagectomies reported an RLN injury incidence of 14% among 266 patients undergoing robotic-assisted minimally invasive esophagectomy (RAMIE) . In the ROBOT trial, type 1 vocal cord injury was observed in five patients in the RAMIE arm and six patients in the OTE arm, with no statistically significant difference ( p = 0.78) . Similarly, the meta-analysis by Esagian et al., which included seven studies, found no significant difference in RLN palsy rates between RAMIE (13.99%; 67/479) and OTE (10.41%; 84/807) groups (OR: 1.31, 95% CI: 0.90–1.90; p = 0.16) . In the RAMIE trial, RAMIE was associated with a higher rate of vocal cord paralysis compared to cMIE (32.6% vs. 27.1%), although the difference was not statistically significant (RR 1.20; 95% CI: 0.87–1.66; p = 0.258) . The largest meta-analysis, conducted by Perry et al., included over 18,000 patients and reported no significant difference in RLN injury rates between RAMIE (8.94%; 237/2652) and cMIE (7.63%; 423/5541). The relative risk was 0.96 (95% CI: 0.82–1.13; p = 0.62; I 2 = 7%; p = 0.36) . 3.2.8. Para-Conduit Diaphragmatic Herniations Para-conduit diaphragmatic herniation is a relatively rare complication following open esophagectomy. However, with the increasing adoption of minimally invasive techniques, the incidence of this complication has significantly risen. This trend is likely attributable to the reduction in adhesions between the gastric conduit and the diaphragmatic hiatus inherent to minimally invasive approaches. In a retrospective analysis by De Silva et al. , the incidence of para-conduit diaphragmatic herniation was found to be significantly higher with minimally invasive techniques (laparoscopic and robotic approaches) compared to open esophagectomy (70.8% vs. 35.5%; p < 0.001). To reduce the risk of herniation, it has been suggested that surgeons place two or three interrupted sutures between the gastric conduit and the right hemidiaphragm after completing the anastomosis. This maneuver may provide additional stability and prevent the herniation of abdominal contents into the thoracic cavity . Robotic-assisted surgery offers superior anatomical visualization and enhanced dexterity, enabling precise preservation of parasympathetic lung innervation through sparing of vagal branches. This has been linked to a reduction in pulmonary complications . In the ROBOT trial, pulmonary complications were significantly lower in the RAMIE group (17/54 patients, 32%) compared to the OTE group (32/55 patients, 58%), with a relative risk (RR) of 0.54 (95% CI: 0.34–0.85; p = 0.005) . Similarly, the meta-analysis by Esagian et al. revealed a significantly lower overall pulmonary complication rate in the RAMIE group (14.29%, 49/343) versus the OTE group (25.32%, 174/687), with an odds ratio (OR) of 0.38 (95% CI: 0.26–0.56; p < 0.001) . However, in the RAMIE trial, the incidence of pulmonary complications was comparable between RAMIE (13.8%) and MIE (14.7%) (RR 0.94, 95% CI: 0.57–1.56; p = 0.812). No significant differences were observed in the rates of pneumonia (9.9% in RAMIE vs. 11.9% in MIE; p = 0.560) or respiratory failure (4.4% in RAMIE vs. 5.1% in MIE; p = 0.767) . The meta-analysis by Perry et al. analyzed 10,154 patients (RAMIE: 3185; cMIE: 6969) and found no statistically significant difference in pulmonary complication rates between RAMIE (20.13%) and cMIE (22.20%) (RR 0.89, 95% CI: 0.77–1.02; p = 0.10; I 2 = 18%; p = 0.20) . In contrast, the meta-analysis by Zhou et al. reported a considerably lower risk of pneumonia following RAMIE compared to cMIE (OR 0.72, 95% CI: 0.52–1.00; p = 0.05) . Similarly, a meta-analysis comparing Ivor Lewis RAMIE with cMIE found a lower weighted pooled proportion of postoperative pneumonia in RAMIE (8%, 95% CI: 6–9) than in cMIE (10%, 95% CI: 7–13). Patients undergoing RAMIE demonstrated a significantly reduced risk of pneumonia (OR 0.46, 95% CI: 0.35–0.61; p < 0.0001) . Cardiac complications, particularly new-onset atrial fibrillation (AF), are associated with significant postoperative morbidity and may serve as surrogate markers for underlying complications. The reduced incidence of cardiac complications in robotic-assisted minimally invasive esophagectomy (RAMIE) may be attributed to decreased intravascular depletion due to less blood loss, reduced oxidative stress from superior lung ventilation in the prone position, and fewer infectious complications compared to open transthoracic esophagectomy (OTE) . In the ROBOT trial, atrial fibrillation occurred in 22% of patients in the RAMIE group (17/45) compared to 47% in the OTE group (26/55), yielding a relative risk (RR) of 0.47 (95% CI: 0.27–0.83; p = 0.006) . Similarly, the meta-analysis by Esagian et al., which included data from five studies , reported a significantly lower rate of atrial fibrillation in the RAMIE group (6.79%, 29/427) compared to the OTE group (8.46%, 54/638), with an odds ratio (OR) of 0.53 (95% CI: 0.29–0.98; p = 0.04) . Conversely, the RAMIE trial found comparable rates of cardiac complications in the RAMIE and conventional minimally invasive esophagectomy (cMIE) groups (two cases vs. one case, p = 0.8) . The meta-analysis by Perry et al. further supported this finding, with cardiac complication rates of 14.02% (365/2604) in the RAMIE group and 15.74% (823/5228) in the cMIE group. No statistically significant difference was observed between the two groups (RR 1.01, 95% CI: 0.86–1.19; p = 0.88; I 2 = 3%, p = 0.42) . The Esophageal Complications Consensus Group (ECCG) defines an anastomotic leak as a “full-thickness gastrointestinal defect involving the esophagus, anastomosis, staple line, or conduit irrespective of presentation or method of identification.” The ECCG further classifies leaks into three types based on the intervention required for management . The technical challenges in crafting a gastric conduit during McKeown esophagectomy and creating an intrathoracic anastomosis in Ivor Lewis esophagectomy are exacerbated in conventional minimally invasive esophagectomy (cMIE) due to the limited maneuverability of laparoscopic instruments. Robotic-assisted minimally invasive esophagectomy (RAMIE) addresses these challenges with enhanced ergonomics, tremor filtration, and the precise control offered by EndoWrist instruments . In a retrospective analysis by de Groot et al., 26% of 152 patients undergoing RAMIE with intrathoracic anastomosis experienced anastomotic leaks. Of these 40 patients, 28% had management failure, including seven cases of esophagobronchial fistula, three cases of anastomotic disconnection, and one death due to septic bleeding . Similarly, Khaitan et al., using data from the Society of Thoracic Surgeons General Thoracic Surgery Database, found that RAMIE was independently associated with higher rates of anastomotic leak compared to OTE (adjusted odds ratio [aOR] 1.53, 95% CI: 1.14–2.04) . In a propensity-matched analysis of 1320 RAMIE cases versus cMIE, the rate of anastomotic leaks requiring surgery was higher in the RAMIE group (aOR 1.39, 95% CI: 1.01–1.92; p = 0.045). Excessive handling of the gastric conduit by robotic instruments, coupled with lower case volumes and the learning curve of surgeons, was suggested as a potential reason for the higher leak rates . In the ROBOT trial, type III anastomotic leaks occurred in 22% of patients in the RAMIE arm and 20% in the OTE arm ( p = 0.57) . A meta-analysis by Esagian et al. found no statistically significant difference in leak rates between RAMIE (6.82%; 46/674) and OTE (6.06%; 79/1303), with an odds ratio of 0.93 (95% CI: 0.60–1.44; p = 0.76) . Similarly, the RAMIE trial reported comparable rates of anastomotic leakage between RAMIE (12.2%) and cMIE (11.3%) (RR 1.08; 95% CI: 0.61–1.90; p = 0.801), with only one patient in each group requiring surgical intervention for type III leakage . The meta-analysis by Perry et al. reported anastomotic leak rates of 12.47% (391/3136) in the RAMIE group and 11.43% (785/6866) in the cMIE group, favoring cMIE (RR 1.23; 95% CI: 1.09–1.38; p = 0.0005; I 2 = 0%, p = 0.64) . A meta-analysis comparing Ivor Lewis cMIE and RAMIE found comparable leak rates, with 7% (95% CI: 6–9%) in the cMIE group and 7% (95% CI: 5–11%) in the RAMIE group. The odds of developing an anastomotic leak were similar between the two groups (OR 0.85; 95% CI: 0.65–1.10; p = 0.22) . The Esophageal Complications Consensus Group (ECCG) classifies gastric conduit necrosis into three types based on the extent of necrosis and the requirement for surgical management and diversion . In the ROBOT trial, type III gastric conduit necrosis was observed in one patient in the RAMIE arm and two patients in the OTE arm ( p = 1.00) . Chylothorax, a significant complication following esophagectomy, is associated with substantial morbidity, often requiring prolonged management and impacting patient recovery. The identification of the thoracic duct is critical in minimizing the risk of chylothorax, and recent advances in near-infrared fluorescence (NIRF) imaging have improved the precision of this process. The use of NIRF for thoracic duct identification has been well established in non-robotic esophageal surgeries , and its integration into robotic-assisted procedures holds promise for enhancing surgical outcomes. A study by Vecchiato et al. in 2020 proposed an innovative approach for detecting the thoracic duct during robotic or laparoscopic surgery. This technique involved intranodal injection of indocyanine green (ICG) into the inguinal nodes under ultrasound guidance, followed by robotic or laparoscopic detection of the thoracic duct in 21 patients. The procedure was successful in identifying the thoracic duct in all cases, with one intraoperative injury to the duct that was promptly clipped. Importantly, no postoperative chylothorax or adverse reactions at the injection site were observed, underscoring the safety and effectiveness of this technique . In addition, Jardinet et al. demonstrated the feasibility of this approach with intrainguinal node ICG injection, offering easy identification of the thoracic duct during surgery. This video publication further reinforces the potential of NIRF-guided identification to prevent thoracic duct injury and subsequent complications such as chylothorax . These findings suggest that NIRF imaging, combined with the intranodal injection of ICG, is a promising method for safely identifying the thoracic duct in robotic esophagectomy. By minimizing the risk of injury and improving the precision of lymphatic structure identification, this approach may reduce postoperative complications and improve patient outcomes. Further studies are necessary to refine this technique and assess its broader applicability in clinical practice. The ECCG defines chyle leaks by classifying them into three types and grades of severity based on daily output . Dezube et al., in a retrospective analysis, compared 70 RAMIE procedures with 277 cMIE procedures, reporting a significantly higher incidence of chyle leaks in the RAMIE group compared to the cMIE group (12.8% vs. 3.6%, p = 0.006) . Notably, McKeown RAMIE was associated with a higher incidence of chyle leaks (33%) than Ivor Lewis RAMIE (4%). In the ROBOT trial, chyle leaks were documented in 17 patients in the RAMIE arm and 12 patients in the OTE arm, with no statistically significant difference ( p = 0.69) . Similarly, the meta-analysis by Esagian et al. showed no significant difference in the incidence of chylothorax between the RAMIE group (5.39%; 29/538) and the OTE group (3.01%; 33/1095), with an odds ratio of 1.31 (95% CI: 0.75–2.29; p = 0.35) . The RAMIE trial also reported comparable rates of chyle leak between RAMIE (five cases) and cMIE (two cases) ( p = 0.449) . Perry et al., in their meta-analysis, found no statistically significant difference in chyle leak rates between the two groups (RR 1.07; 95% CI: 0.72–1.60; p = 0.74; I 2 = 13%, p = 0.30). The leak rates were 2.82% (69/2443) in the RAMIE group and 3.84% (197/5135) in the cMIE group . Similarly, Zhou et al. observed no significant difference in pooled data analysis between RAMIE and cMIE approaches (OR: 1.33; 95% CI: 0.57–3.07; p = 0.51) . During esophagectomy, the recurrent laryngeal nerve (RLN) is vulnerable to injury due to thermal damage, stretching, compression, or vascular compromise, potentially leading to vocal cord palsy. Such injuries can significantly increase the risk of pulmonary complications, ICU readmissions, and prolonged hospital stays . The ECCG defines vocal cord injury as “vocal cord dysfunction post-resection, confirmed and assessed by direct examination”, with severity graded by laterality and intervention needed, categorized into three types . Robotics has been hypothesized to reduce RLN injuries due to improved visualization of the nerve and adjacent vascular structures. A retrospective analysis by Scholtemeijer et al. on McKeown esophagectomies reported an RLN injury incidence of 14% among 266 patients undergoing robotic-assisted minimally invasive esophagectomy (RAMIE) . In the ROBOT trial, type 1 vocal cord injury was observed in five patients in the RAMIE arm and six patients in the OTE arm, with no statistically significant difference ( p = 0.78) . Similarly, the meta-analysis by Esagian et al., which included seven studies, found no significant difference in RLN palsy rates between RAMIE (13.99%; 67/479) and OTE (10.41%; 84/807) groups (OR: 1.31, 95% CI: 0.90–1.90; p = 0.16) . In the RAMIE trial, RAMIE was associated with a higher rate of vocal cord paralysis compared to cMIE (32.6% vs. 27.1%), although the difference was not statistically significant (RR 1.20; 95% CI: 0.87–1.66; p = 0.258) . The largest meta-analysis, conducted by Perry et al., included over 18,000 patients and reported no significant difference in RLN injury rates between RAMIE (8.94%; 237/2652) and cMIE (7.63%; 423/5541). The relative risk was 0.96 (95% CI: 0.82–1.13; p = 0.62; I 2 = 7%; p = 0.36) . Para-conduit diaphragmatic herniation is a relatively rare complication following open esophagectomy. However, with the increasing adoption of minimally invasive techniques, the incidence of this complication has significantly risen. This trend is likely attributable to the reduction in adhesions between the gastric conduit and the diaphragmatic hiatus inherent to minimally invasive approaches. In a retrospective analysis by De Silva et al. , the incidence of para-conduit diaphragmatic herniation was found to be significantly higher with minimally invasive techniques (laparoscopic and robotic approaches) compared to open esophagectomy (70.8% vs. 35.5%; p < 0.001). To reduce the risk of herniation, it has been suggested that surgeons place two or three interrupted sutures between the gastric conduit and the right hemidiaphragm after completing the anastomosis. This maneuver may provide additional stability and prevent the herniation of abdominal contents into the thoracic cavity . In the ROBOT trial, two patients on the RAMIE arm and one patient on the OTE arm died in the immediate postoperative period, and 30-day and 90-day mortality rates were 2% and 9% in the RAMIE arm vs. 0% and 2% in the OTE arm . In the RAMIE trial, for 30-day mortality, one patient in MIE died from acute cerebral infarction on POD 12. For 90-day mortality, one patient in RAMIE died from severe pneumonia on POD 42 . In the meta-analysis by Perry et al., the 30-day mortality rate for each procedure was 1.63% (44/2707) in the RAMIE group and 1.87% (117/6244) in the cMIE group. There was no statistically significant difference between the two groups (RR 1.03, p = 0.88 [95% CI 0.73, 1.44], I 2 = 0%, p = 0.53). The rate of 90-day mortality was 3.55% (106/2987) in the RAMIE group and 4.84% (336/6946) in the cMIE group, showing no statistically significant difference between the two groups (RR 0.95, p = 0.66 [95% CI 0.77, 1.18], I 2 = 0%, p = 0.93) . In the meta-analysis by Zhou et al., the in-hospital mortality and 90-day mortality had no statistically distinguished difference in a merged data analysis (OR = 0.54, 95 CI [0.14, 2.02] p = 0.36) and (OR = 0.69, 95 CI [0.26, 1.83] p = 0.46), respectively . Fluorescence-guided surgery has revolutionized many surgical specialties by enabling more precise and safer procedures. In esophageal cancer surgery, a particularly complex domain, this technology has become an essential tool for enhancing surgical outcomes. This outlines the current applications of fluorescence in identifying lymphatic structures, performing sentinel node biopsy, visualizing the thoracic duct, and conducting angiography during robotic-assisted esophagectomy. 4.1. Angiography Anastomotic leakage following esophagectomy is a significant complication, occurring in 6-41% of patients and associated with considerable morbidity and mortality . Fluorescence angiography has proven effective in reducing the incidence of such complications by providing real-time vascular visualization. In a study of 30 patients, Sarkaria et al. demonstrated the utility of fluorescence in identifying the termination of the vascular arcade and small transverse vessels under fluorescence, which aided in confirming the vascular supply during the mobilization of the greater curvature and omentum . In a larger cohort of 75 patients, Egberts et al. utilized fluorescence angiography to analyze gastric conduit perfusion during robotic surgery . While the majority of patients benefited from this technique, Hodari et al. reported anastomotic leakage in three patients, even with real-time perfusion assessment . Similar studies conducted by Pötscher et al. and DeLong et al. corroborate the efficacy of fluorescence in detecting perfusion issues, although challenges remain in predicting leaks with absolute certainty. Slooter et al., in their study of 81 patients undergoing Ivory Lewis and McKeown esophagectomies with robotic assistance, found that the time interval between indocyanine green (ICG) injection and conduit tip reinforcement was a significant predictor of outcomes, with a cut-off value of 98 s . In open surgery, Ishikawa et al. proposed a quantitative analysis using three parameters—ingress index at both the tip and 5 cm of the conduit, and ingress time—as key indicators for predicting anastomotic leaks following esophagectomy . These findings suggest that a more quantitative approach to fluorescence angiography could enhance the prediction and prevention of postoperative complications. 4.2. Near-Infrared Fluorescent-Guided Lymphadenectomy and Sentinel Node Biopsy Lymphadenectomy plays a crucial role in achieving optimal oncological outcomes in esophageal cancer surgery. The use of image-guided lymphadenectomy has been well established in non-robotic surgery , and recent advancements in robotic-assisted esophagectomy have integrated near-infrared fluorescence (NIRF) imaging for more precise identification and resection of lymphatic structures. This technique enhances the surgeon’s ability to visualize lymph nodes, especially in areas that are difficult to access or identify through conventional methods. In a study by Hosogi et al., 15 patients undergoing robotic esophagectomy were assessed for NIRF-guided lymphadenectomy. The study found that 80% of patients had NIRF-stained lymph nodes in the right recurrent laryngeal nerve area, and 73% had stained lymph nodes in the left recurrent laryngeal nerve area, highlighting the ability of NIRF imaging to facilitate accurate lymph node mapping during robotic surgery . Furthermore, the prospective ESOMAP feasibility trial, which evaluated robotic-assisted minimally invasive Ivory Lewis esophagectomy, demonstrated the feasibility of intraoperative NIRF-guided lymph node mapping and resection for pathological examination. In a cohort of 20 patients, 5 had no ICG uptake during a standard D2 lymphadenectomy, but notably, the NIRF-guided procedures were significantly shorter compared to non-NIRF procedures, suggesting potential advantages in terms of operative efficiency. The study of NIRF-stained lymph nodes in gastroesophageal junction cancer showed no increase in the number of harvested lymph nodes compared to a historical control group. Additionally, there were no significant differences in operative time, blood loss, or other postoperative complications between the NIRF and non-NIRF groups . These findings suggest that while NIRF-guided lymphadenectomy may improve the precision of lymph node identification, it may not necessarily result in a higher number of harvested nodes or improved clinical outcomes in all cases. In a study by Shiomi et al., 54 patients without preoperative chemotherapy were divided into groups based on NIRF-guided resection, with ICG-positive or -negative lymph nodes and metastasis-positive or -negative nodes. This study revealed that preoperative chemotherapy affected the sensitivity of NIRF in predicting metastatic lymph nodes, indicating that the effectiveness of NIRF imaging may vary depending on the patient’s treatment history . A hybrid approach combining technetium-99m and ICG for sentinel node biopsy (SNB) was also investigated in high-risk pT1c patients. In a cohort of 10 patients, the hybrid tracer successfully identified sentinel nodes in all cases. In one patient, the dissection was found to be incomplete, and in four patients, additional fluorescent lymph nodes were harvested, with micrometastases identified in two cases . Similarly, Overwater et al. applied a hybrid tracer in minimally invasive surgery, including robotic surgery, for pT1b esophageal cancer patients. They found successful sentinel node identification in all five patients, with two cases revealing additional peritumoral fluorescent-only sentinel nodes, though no postoperative pathological metastases were found . These studies underscore the promising role of NIRF in esophageal cancer surgery, particularly for sentinel node biopsy and lymphadenectomy. However, challenges remain, such as the impact of preoperative chemotherapy on the sensitivity of NIRF in detecting metastatic lymph nodes. Further research and refinement of these techniques will be essential to determine their full clinical potential and to standardize their use in esophageal cancer treatment protocols. Anastomotic leakage following esophagectomy is a significant complication, occurring in 6-41% of patients and associated with considerable morbidity and mortality . Fluorescence angiography has proven effective in reducing the incidence of such complications by providing real-time vascular visualization. In a study of 30 patients, Sarkaria et al. demonstrated the utility of fluorescence in identifying the termination of the vascular arcade and small transverse vessels under fluorescence, which aided in confirming the vascular supply during the mobilization of the greater curvature and omentum . In a larger cohort of 75 patients, Egberts et al. utilized fluorescence angiography to analyze gastric conduit perfusion during robotic surgery . While the majority of patients benefited from this technique, Hodari et al. reported anastomotic leakage in three patients, even with real-time perfusion assessment . Similar studies conducted by Pötscher et al. and DeLong et al. corroborate the efficacy of fluorescence in detecting perfusion issues, although challenges remain in predicting leaks with absolute certainty. Slooter et al., in their study of 81 patients undergoing Ivory Lewis and McKeown esophagectomies with robotic assistance, found that the time interval between indocyanine green (ICG) injection and conduit tip reinforcement was a significant predictor of outcomes, with a cut-off value of 98 s . In open surgery, Ishikawa et al. proposed a quantitative analysis using three parameters—ingress index at both the tip and 5 cm of the conduit, and ingress time—as key indicators for predicting anastomotic leaks following esophagectomy . These findings suggest that a more quantitative approach to fluorescence angiography could enhance the prediction and prevention of postoperative complications. Lymphadenectomy plays a crucial role in achieving optimal oncological outcomes in esophageal cancer surgery. The use of image-guided lymphadenectomy has been well established in non-robotic surgery , and recent advancements in robotic-assisted esophagectomy have integrated near-infrared fluorescence (NIRF) imaging for more precise identification and resection of lymphatic structures. This technique enhances the surgeon’s ability to visualize lymph nodes, especially in areas that are difficult to access or identify through conventional methods. In a study by Hosogi et al., 15 patients undergoing robotic esophagectomy were assessed for NIRF-guided lymphadenectomy. The study found that 80% of patients had NIRF-stained lymph nodes in the right recurrent laryngeal nerve area, and 73% had stained lymph nodes in the left recurrent laryngeal nerve area, highlighting the ability of NIRF imaging to facilitate accurate lymph node mapping during robotic surgery . Furthermore, the prospective ESOMAP feasibility trial, which evaluated robotic-assisted minimally invasive Ivory Lewis esophagectomy, demonstrated the feasibility of intraoperative NIRF-guided lymph node mapping and resection for pathological examination. In a cohort of 20 patients, 5 had no ICG uptake during a standard D2 lymphadenectomy, but notably, the NIRF-guided procedures were significantly shorter compared to non-NIRF procedures, suggesting potential advantages in terms of operative efficiency. The study of NIRF-stained lymph nodes in gastroesophageal junction cancer showed no increase in the number of harvested lymph nodes compared to a historical control group. Additionally, there were no significant differences in operative time, blood loss, or other postoperative complications between the NIRF and non-NIRF groups . These findings suggest that while NIRF-guided lymphadenectomy may improve the precision of lymph node identification, it may not necessarily result in a higher number of harvested nodes or improved clinical outcomes in all cases. In a study by Shiomi et al., 54 patients without preoperative chemotherapy were divided into groups based on NIRF-guided resection, with ICG-positive or -negative lymph nodes and metastasis-positive or -negative nodes. This study revealed that preoperative chemotherapy affected the sensitivity of NIRF in predicting metastatic lymph nodes, indicating that the effectiveness of NIRF imaging may vary depending on the patient’s treatment history . A hybrid approach combining technetium-99m and ICG for sentinel node biopsy (SNB) was also investigated in high-risk pT1c patients. In a cohort of 10 patients, the hybrid tracer successfully identified sentinel nodes in all cases. In one patient, the dissection was found to be incomplete, and in four patients, additional fluorescent lymph nodes were harvested, with micrometastases identified in two cases . Similarly, Overwater et al. applied a hybrid tracer in minimally invasive surgery, including robotic surgery, for pT1b esophageal cancer patients. They found successful sentinel node identification in all five patients, with two cases revealing additional peritumoral fluorescent-only sentinel nodes, though no postoperative pathological metastases were found . These studies underscore the promising role of NIRF in esophageal cancer surgery, particularly for sentinel node biopsy and lymphadenectomy. However, challenges remain, such as the impact of preoperative chemotherapy on the sensitivity of NIRF in detecting metastatic lymph nodes. Further research and refinement of these techniques will be essential to determine their full clinical potential and to standardize their use in esophageal cancer treatment protocols. The evolution of robotic surgical techniques has introduced innovative approaches aimed at minimizing surgical trauma while maintaining or improving patient outcomes. One such method, the robotic transhiatal approach (Th-RAMIE), has been designed to eliminate the need for thoracotomy or single-lung ventilation, thereby reducing the associated morbidity. In a prospective trial conducted by Williams et al., 97 patients undergoing Th-RAMIE were compared to 212 patients treated with open transhiatal esophagectomy (THE) for patient-related outcomes. The study demonstrated comparable outcomes between the two groups; however, opioid use at discharge was significantly lower in the Th-RAMIE group (71% vs. 82%; p = 0.03) . These findings suggest that Th-RAMIE offers an advantage in postoperative pain management, potentially enhancing recovery while preserving oncological outcomes. 5.1. Robotic-Assisted Cervical Esophagectomy (RACE) Another promising advancement is the single-port trans-cervical approach, also known as robotic-assisted cervical esophagectomy (RACE). This novel technique is gaining traction as an alternative to the conventional thoracic approach, with the potential to markedly reduce pulmonary complications by completely avoiding thoracic incisions . The articulated robotic arms provide enhanced maneuverability and precision, reducing interference with critical structures such as the recurrent laryngeal nerve. The three-dimensional, magnified field of view inherent to robotic platforms allows for a more accurate assessment of anatomical relationships, which may lead to improved patient outcomes and quality of life. Fujita et al. reported on a case series of ten patients who underwent robot-assisted trans-cervical esophagectomy using a bilateral cervical approach. The short-term postoperative outcomes revealed the following complication rates: arrhythmia in 10.0% of patients, vocal cord paralysis in 10.0%, anastomotic leakage in 20.0%, and no cases of postoperative pneumonia . While the study highlights the feasibility of this approach, it underscores the need for larger studies to further evaluate its safety profile and long-term outcomes. 5.2. Future Implications The ongoing ROBOT-2 and REVATE trials are expected to provide valuable insights into the benefits of robotic approaches in esophageal cancer surgery. The ROBOT-2 trial, which compares robot-assisted minimally invasive thoraco-laparoscopic esophagectomy with conventional minimally invasive esophagectomy for resectable esophageal adenocarcinoma, aims to further elucidate the potential advantages of robotic surgery in terms of postoperative recovery and long-term oncological outcomes . Similarly, the REVATE trial, which contrasts robotic-assisted esophagectomy with video-assisted thoracoscopic esophagectomy (VATS), will help clarify the relative merits of robotic systems in minimizing complications, enhancing surgical precision, and improving functional recovery . These trials play a crucial role in expanding the evidence base regarding the impact of robotic techniques on postoperative outcomes, particularly with regard to complication rates, recovery timelines, and overall patient quality of life. As the data from these studies accumulate, they may refine current clinical practice and offer stronger evidence for the adoption of robotic approaches in esophageal surgery. Another promising advancement is the single-port trans-cervical approach, also known as robotic-assisted cervical esophagectomy (RACE). This novel technique is gaining traction as an alternative to the conventional thoracic approach, with the potential to markedly reduce pulmonary complications by completely avoiding thoracic incisions . The articulated robotic arms provide enhanced maneuverability and precision, reducing interference with critical structures such as the recurrent laryngeal nerve. The three-dimensional, magnified field of view inherent to robotic platforms allows for a more accurate assessment of anatomical relationships, which may lead to improved patient outcomes and quality of life. Fujita et al. reported on a case series of ten patients who underwent robot-assisted trans-cervical esophagectomy using a bilateral cervical approach. The short-term postoperative outcomes revealed the following complication rates: arrhythmia in 10.0% of patients, vocal cord paralysis in 10.0%, anastomotic leakage in 20.0%, and no cases of postoperative pneumonia . While the study highlights the feasibility of this approach, it underscores the need for larger studies to further evaluate its safety profile and long-term outcomes. The ongoing ROBOT-2 and REVATE trials are expected to provide valuable insights into the benefits of robotic approaches in esophageal cancer surgery. The ROBOT-2 trial, which compares robot-assisted minimally invasive thoraco-laparoscopic esophagectomy with conventional minimally invasive esophagectomy for resectable esophageal adenocarcinoma, aims to further elucidate the potential advantages of robotic surgery in terms of postoperative recovery and long-term oncological outcomes . Similarly, the REVATE trial, which contrasts robotic-assisted esophagectomy with video-assisted thoracoscopic esophagectomy (VATS), will help clarify the relative merits of robotic systems in minimizing complications, enhancing surgical precision, and improving functional recovery . These trials play a crucial role in expanding the evidence base regarding the impact of robotic techniques on postoperative outcomes, particularly with regard to complication rates, recovery timelines, and overall patient quality of life. As the data from these studies accumulate, they may refine current clinical practice and offer stronger evidence for the adoption of robotic approaches in esophageal surgery. In conclusion, robotic-assisted approaches to esophageal cancer surgery have shown significant improvements in surgical precision, postoperative recovery, and complication management. The integration of robotic techniques, such as RAMIE, Th-RAMIE, and RACE, along with fluorescence-guided technologies, has enhanced lymphadenectomy, sentinel node biopsy, and thoracic duct identification, contributing to better oncological outcomes and reduced morbidity. Studies consistently demonstrate that robotic surgery offers advantages such as lower postoperative pain, faster functional recovery, and fewer complications like anastomotic leakage and chylothorax, compared to traditional open approaches. As ongoing trials continue to evaluate long-term outcomes, the evidence strongly supports robotic surgery as a superior modality for esophagectomy, offering both clinical benefits and the potential for improved patient quality of life. |
Evidence of traumatic brain injury in headbutting bovids | 0ae99d63-8bdd-42ed-8467-f1c7c5ab63c0 | 9217783 | Pathology[mh] | Traumatic brain injury (TBI) is one of the main causes of neurological deficits and death worldwide, accounting for 2.5 million hospital admissions per year . Human cranial anatomy is vulnerable to coup-contrecoup injuries and TBI is most often the result of a fall, motor vehicle collision, or firearm accident in the U.S. . Military personnel and athletes are especially at risk and have become part of the mounting concern around TBI and its long-term effects . Repetitive brain trauma is particularly dangerous due to its potential link to progressive neurological deterioration . However, while neurodegenerative diseases such as chronic traumatic encephalopathy (CTE) can be suspected during life, they can only be confirmed postmortem , making them difficult to study. TBI pathology is often categorized into primary and secondary injuries, the primary being the acute phase of injury that causes axonal shearing leading to hemorrhage and contusions, while the secondary injury is the result of molecular mechanisms involving cell death and tissue degeneration. Dying and damaged cells release debris that triggers the surrounding microglia and astrocytes to mount an immune response resulting in inflammation . In relation to neuronal death, neurofibrillary tangles (NFTs) form and accumulate in the superficial layers of the cerebral cortex and at the depths of sulci, resulting in axonal instability and impeded neuronal communication . Microscopic imaging of these pathologies using histology provides an indicator on the progression of the disease. In TBI cases without focal lesions, sequelae such as behavioral changes or macroscopic regional brain shrinkage only appear at advanced stages of pathology , adding a layer of difficulty to diagnosis of mild and repetitive TBI. Although the study of brain injury has advanced through animal models, mostly mice and rats, all neuroprotective therapies developed from rodent approaches have failed late-stage clinical trials, possibly due to the many morphofunctional differences between rodent and human brains . Complementing rodent studies with alternative animal models with larger, gyrencephalic brains can increase our knowledge of the translational mechanics of neurodegenerative disease progression . Many male animals of the artiodactyl order (mammals with even-toed hooves and cetaceans) perform head-to-head sexual displays during the reproductive season called the rut, usually sparring with their heads, horns, or antlers as a show of dominance in their social hierarchy. Among artiodactyls, caprines (sheep, goat, and muskox-type animals) exhibit the most extreme form of headbutting behavior. Muskox bulls ( Ovibos moschatus ) run towards each other and bash heads at peak speeds of 50 km/h. Bighorn sheep ( Ovis canadensis ) engage in forceful headbutting displays in mountainous terrain, both animals exerting forces of around 3000 N , aided by some of the proportionally largest horns in extant mammals. It is a common belief that headbutting animals like bighorn sheep are mostly unscathed after headbutting; however, this claim has not been investigated empirically, either through behavioral measures, or anatomically. The biomechanics literature specifically has embraced this assumption with the aim to create biomimicry materials from horns for products such as helmets . However, the anatomical aspects of the modeled animals lack comprehensive analysis . Horns are certainly an important factor in the absorption of shock, but they cannot be the only structures in play, as females with smaller horns and dehorned domestic bovids also headbutt without sustaining apparent injury . Headbutting behavior is not uniform across species or sexes. Bighorn ewes headbutt at lower forces but engage up to four times more often than males , and muskox cows also headbutt on occasion, without engaging in the long, ritualistic rut behavior observed in the males (Jamie Luce, The Musk Ox Farm, personal communication). Finally, bighorn rams often strike each other on a fat pad between their horns and on the forehead, rather than directly on the horns during fights . Field observations have described muskoxen as “acting dazed” or even “bleeding from the nose and ears after the rut” , but these observations were not confirmed anatomically. Whether bovids sustain any damage from these encounters in the form of chronic or acute brain trauma remains untested at both the macroscopic and microscopic levels, and yet sheep are one of the preferred large-animal models for TBI and CTE . In this study, we aim to determine if bighorn sheep and muskoxen naturally sustain TBIs despite their large horns and thick skulls. In that instance, we expect males to follow a pattern resembling CTE in humans, with tau-immunoreactive staining of NFTs and neurites, especially in the superficial neocortical layers and in the depths of the sulci, coupled with activated microglia and astrocytosis. Macroscopic signs of TBI would only be expected to develop in the latest stages of neurodegeneration . Understanding how bovids survive high-force head impacts can address the anatomical and physiological knowledge gaps with regards to current large animal TBI models, as well as providing further insight on the life history of these animals. Moreover, comparing TBI development between bovids and humans could lead to a better understanding of brain injury progression and treatment overall. Specimens A total of nine brain specimens were used in this study: three from muskoxen, four from bighorn sheep, one from a human with late-stage Alzheimer’s disease (AD), and one from a human with CTE, the latter two used as positive controls. The brains of three wild adult muskoxen were collected during a field expedition to Ittoqqortoormiit, Greenland by TMW and all ages were estimated from tooth wear. During the time of brain recovery, the ambient temperature remained always below 7 °C, mitigating tissue deterioration. The male muskox (older adult, collected in summer 2019) was shot in the neck directly after a goring injury to the flank sustained while headbutting another male muskox. One female muskox was shot in the top of the head (middle-aged adult, collected in summer 2018) and the second female muskox was shot in the neck (very old adult, collected in summer 2018), for subsistence hunting (Table ). The male muskox’s skull was approximately 4 cm thick in the frontal region (including sinuses), and 2 cm thick in the parietal cortex. The four bighorn sheep brains were collected as follows. The brain of an adult male bighorn sheep (Bighorn 1, five years old, collected in winter 2020) was acquired from Colorado Parks and Wildlife from a captive research herd after it was humanely euthanized (darted with NalMedA for sedation, euthanized with 20 ml Euthasol IV) due to a leg fracture. Formalin was injected around the brainstem and into the carotid artery for preservation before shipping. The male bighorn sheep’s skull was approximately 3 cm thick in the frontal region (including sinuses) and 1.5 cm thick in the parietal region. The brain of one wild adult female bighorn sheep (Bighorn 2, four years old, collected in winter 2020) was acquired from Utah Fish and Wildlife after euthanasia due to a Mycoplasma infection. The skull of the bighorn ewe was approximately 1.3 cm thick in the frontal region (including sinuses) and 0.8 cm thick in the parietal cortex. Two additional female bighorn sheep brains (Bighorn 3, five years old, collected in fall 1987; Bighorn 4, adult, collected in winter 2008) were archived in our collection and were provided by the Buffalo Zoo in 2003. After the muskoxen and sheep brains were removed from the skull, all samples were fixed in 10% formalin, either hours after death for the muskoxen, or within 36 h after death for the bighorn sheep. As a positive control for presence of phosphorylated tau, an archived human brain specimen with AD (male, 85 years old, Clinical Dementia Rating 3, Mini-Mental State Exam 11, Thal amyloid stage 4, Braak tangle stage V, postmortem interval 11 h, clinical diagnosis: severe cognitive impairment) and one with CTE (male, 69 years old, postmortem interval 9 h, clinical history: repetitive athletic head injuries, post-mortem diagnosis of advanced CTE, moderate cerebrovascular disease—athero-arteriolosclerosis; moderate hypoxic–ischemic encephalopathy, severe postmortem autolysis) were used in this study. Magnetic resonance imaging (MRI) MRI with superior soft tissue visualization was used to image the anatomical structure of the muskox and bighorn sheep brains. Coronal T2-weighed turbo spin-echo images were performed on a whole-body 7 Tesla (7 T) MRI scanner (Siemens Magnetom, Siemens Healthcare, Erlangen Germany) using a 1-channel transmit and 32-channel receive head coil (Nova Medical, Wilmington, Massachusetts) with the following parameters: repetition time (TR) 8000 ms, echo time (TE) 64 ms, number of sections 24, section thickness 1 mm, field-of-view 16 × 14 cm 2 , voxel size 0.5 × 0.5 × 1 mm 3 , scanning duration 6 min 20 s. Human tissue specimen imaging was performed in compliance with all institutional requirements. BLAST analysis The tau-immunoreactive antibodies selected for this study were developed against human antigens. Their specificity in bighorn sheep and muskox tissues required validation to confirm their immunoreactivity. A basic local alignment tool (BLAST) compares protein sequences to sequence databases and calculates the statistical significance. The MAPT gene or protein sequence for tau was not available in bighorn sheep or muskoxen, therefore BLAST was applied to the predicted proteome of the domestic sheep ( Ovis aries ), the closest related species with MAPT sequence available on NCBI. Protein isoforms with the closest sequence homology and molecular weight to the requested sequence were ranked in terms of degree of identity in percent and E value. Homology values over 95% are considered acceptable, values over 85% are considered moderate. Immunohistochemistry and histochemical staining Brain tissue from the human subjects was sampled from Brodmann area 10. In the bovids, tissue was taken from the anterior region of the prefrontal cortex of the right hemisphere (Fig. ), and additionally from the parietal cortex of the right hemisphere for the three muskoxen. Each block was cut into 50 µm-thick sections on a vibratome (Leica VT1000S) and stored in phosphate-buffered saline (PBS, pH 7.0) solution with 0.1% sodium azide. Multiple phosphorylated tau antibodies were tested in this study, as no protocols existed for these bovid species, making it uncertain which epitopes were present in each. Antibodies recognizing ionized calcium-binding adaptor molecule-1 (Iba1) and glial fibrillary acidic protein (GFAP) were used to investigate microglial and astrocytic morphology, respectively. Three antibodies were tested to detect the presence of phosphorylated tau protein (p-tau). Anti-CP13 is a p-tau antibody clone that binds an epitope around Serine 202 (pSer202 tau), the AT8 clone recognizes epitopes around Serine 202 and Threonine 205 (pSer205/Thr205 tau), and the PHF-1 clone was raised against epitopes including Serine 396 and Serine 404 (pSer396/Ser404 tau). Both CP13 and PHF-1 antibody clones were generously supplied by the Davies laboratory (Feinstein Institute for Medical Research, Northwell Health, Manhasset, New York, USA). Anti-neurofilament clone SMI-312 detects medium- and heavy-chain phosphorylated neurofilament proteins (pNFP) and was used to investigate axonal damage. Anti-denatured myelin basic protein (dMBP) was applied to assess possible demyelination. The MOAB-2 antibody clone was applied against amyloid beta protein (Aβ), and anti-TDP43 clone binds an epitope around phosphorylated Serine 409 and Serine 410 (pSer409/410) to detect pathological proteins. Anti-collagen IV was applied to highlight blood vessel morphology. Additional details on antibodies are reported in Table . Sections were stained either as single instances or, in the case of the muskoxen, as series of ten sections, each 500 µm apart for stereological quantification. Primary antibody controls were performed by omitting the primary antibody and assessed that the secondary antibody did not generate non-specific staining (supplementary figures S1 and S2, provided as an online resource). All washes were performed with either PBS (Iba1, GFAP, dMBP, collagen IV) or Tris-buffered saline (TBS, pH 7.0) (AT8, CP13, PHF1, TDP43, MOAB-2, SMI-312) at room temperature on a shaker at 80 rpm for five minutes each. Antigen retrieval for Iba1 was performed by submerging free-floating tissue sections in a 10 mM EDTA solution (pH 8.0) in closed 15-ml tubes. The tubes were immersed in a water bath at 80 °C for 10 min. For GFAP, AT8, MOAB-2, dMBP, and SMI-312, antigen retrieval was performed by submerging free-floating tissue sections in citrate buffer (pH 6.0), then boiled at 100 °C for 10 min, followed by 5 min of cooling down. Antigen retrieval for collagen IV was performed by submerging sections in a 0.5 M acetic acid and 10 mg/ml pepsin solution at 37 °C for 8 min. After antigen retrieval, sections were transferred to 12-well plates using a glass hook and washed three times. Sections were then incubated with 0.3% hydrogen peroxide and 0.3% Triton X-100 in buffer for 30 min at room temperature, to inhibit endogenous peroxidase activity. The sections were then washed four times and blocked in buffer with 5% normal donkey serum (017000121, JacksonImmuno, West Grove, PA, USA) or normal goat serum (for TDP43 and collagen IV. 1002635372, Sigma, St. Louis, MD, USA) for 1 h, followed by an incubation in primary antibody (Iba1, GFAP, CP13, AT8, PHF1, MOAB-2, TDP43, SMI-312, dMBP, or collagen IV) in buffer with 5% normal donkey serum and 0.3% Triton X-100, at 4 °C overnight (Iba1, GFAP, AT8, MOAB-2, dMBP, SMI-312, collagen IV) or for 64 h (CP13, PHF1, TDP43); the control sections were incubated in buffer. After primary antibody incubation, sections were washed three times in buffer and 0.3% Triton X-100, then incubated with the appropriate secondary antibody (biotinylated donkey anti-rabbit, secondary antibody, 1:1000, 715065152, JacksonImmuno; biotinylated donkey anti-mouse, secondary antibody, 1:1000, 715065150, JacksonImmuno, biotinylated goat anti-rat secondary antibody 1:1000, BA9400, Vector laboratories, Burlingame, CA, USA) in a 5% normal donkey or goat serum and 0.3% Triton X-100 solution in buffer at room temperature for one hour (Iba1, GFAP, collagen IV, SMI-312) or two hours (AT8, CP13, PHF1, TDP43, MOAB-2, dMBP). After incubation with the secondary antibody, the sections were washed in buffer four times, then incubated with avidin–biotin solution (according to the manufacturer’s instructions, Vectastain ABC kit, Vector Laboratories, Burlingame, CA, USA) for one hour (Iba1, GFAP, MOAB-2, dMBP, collagen IV, SMI-312) or two hours (AT8, CP13, PHF1, TDP43) at room temperature. The sections were then washed four times, followed by an incubation with DAB peroxidase substrate (Vector Laboratories, according to the manufacturer’s instructions, DAB kit SK-4100) to reveal immunostaining. Samples were then mounted on gelatin-coated slides and left to dry overnight. After drying, the samples were counterstained with cresyl violet, dehydrated through an ethanol gradient, and coverslipped with DPX mounting medium. Luxol Fast-Blue was applied to observe demyelination as follows. Mounted sections were incubated in 0.1% Luxol Fast-Blue solution for one hour at 56 °C, then rinsed in distilled water. Sections were then differentiated in a 0.05% lithium carbonate solution for 30 s, followed by 70% ethanol for 30 s. Sections were rinsed in distilled water and checked microscopically for differentiation. Finally, slides were rinsed and differentiated in 95% ethanol then coverslipped as above. Immunofluorescence The blocking solution was increased from 5 to 10% normal donkey serum and primary antibody incubation was performed as in the immunohistochemistry protocol for both the CP13 and GFAP antibodies. After primary antibody incubation, sections were washed four times, protected from light, and incubated in biotinylated donkey anti-mouse antibody (1:1000, 715065150, JacksonImmuno) and anti-rabbit-AlexaFluor 488 (1:1000, A31570, ThermoFisher Scientific), followed by streptavidin-AlexaFluor 555 (1:500, A21206, ThermoFisher Scientific) in 10% normal donkey serum and 0.1% Triton X-100 in TBS at room temperature for two hours. The sections were washed four times and then incubated in 10% normal donkey serum and 0.1% Triton X-100 in TBS at room temperature for two hours. Sections were then washed four more times, mounted on SuperFrost slides and dried for one hour at 50 °C. Wells were drawn around the sections with an ImmEdge pen and sections were washed for 10 min. Sections were then incubated with TrueBlack (diluted 20 × in 70% ethanol) for 30 s each to reduce autofluorescence, then washed four times. The sections were then incubated with 4′,6-diamino-2-phenylindole dihydrochloride (DAPI, 250 ng/ml) for ten minutes in a humid chamber to stain cell nuclei and then washed a final time, after which they were mounted under Vectashield (H1000, Vector Laboratories) and coverslipped (24 × 50 mm No.1.5 ThermoFisher Scientific). Microscopy and stereology Brightfield microscopy images were taken on an Axiophot brightfield microscope (Carl Zeiss Microscopy, Jena, Germany), with a 10×/0.32 Plan-Apochromat objective using StereoInvestigator (version 11.03, MBF Bioscience, Williston, VT, USA). Fluorescence images were taken on a CLSM 780 confocal microscope (Carl Zeiss Microscopy), using a 20×/0.8 DICII objective and DPSS 561-10 diode and Argon lasers at excitation wavelengths of 461, 555, and 488 nm. Confocal stacks in layers II and III of the cerebral cortex were imaged at 512 × 512 pixel resolution with a Z-step of 1 µm and a pinhole setting of 1 Airy unit for the red wavelength and optimal settings for gain and contrast. Images are presented as maximum intensity projections of the Z-stack using ZenBlue (version 3.3, Carl Zeiss Microscopy). Stereological quantification of tau immunostaining was performed using the optical fractionator workflow probe in StereoInvestigator (magnification × 10, counting frame size 700 × 700 µm, SRS grid layout at 100% of the region of interest, optical dissector height 11 µm with 2 µm of top and bottom guard zones, manual focus), on each muskox specimen in a series of 10 sections, each separated by 500 µm, and counted exhaustively. Cortical layers were manually contoured into layers I, II, III, IV–VI, and white matter. In each layer, different markers were placed for tau-immunoreactive neuropil threads (axonal or dendritic filaments composed of abnormally phosphorylated microtubule-associated tau protein), neuritic thread clusters (circular dense cluster of neuritic threads), and neurons (neuronal cells with tau-immunoreactive inclusions in the cytoplasm). Section contours were aligned manually and the coordinates were exported to create individual and combined heatmaps of tau-immunoreactive neuropil density distribution in Rstudio using the ggplot2 package . Annotated R code and raw data is available on GitHub ( https://github.com/NLAckermans/Ackermans2022BovidTBI.git ). To quantify tau-immunoreactive pathology accumulation in the sulci as opposed to gyri in the muskox specimens, sulcal depths were delimitated as 1/3 of the sulcus as in and pSer202 tau-immunoreactive markers in each region were quantified in StereoInvestigator using the same specifications as above. To quantify pSer202 tau-immunoreactive pathology around blood vessels in the muskoxen as compared to the CTE human control, one section from each individual was subjected to exhaustive counting of vessels larger than 30 µm in diameter at 2.5 × magnification with StereoInvestigator’s Optical Fractionator probe. The percentage of vessels immunostained with pSer202 tau located within 100 µm of the edge of the vessel and the average distance from the edge of the vessel were calculated and reported. A total of nine brain specimens were used in this study: three from muskoxen, four from bighorn sheep, one from a human with late-stage Alzheimer’s disease (AD), and one from a human with CTE, the latter two used as positive controls. The brains of three wild adult muskoxen were collected during a field expedition to Ittoqqortoormiit, Greenland by TMW and all ages were estimated from tooth wear. During the time of brain recovery, the ambient temperature remained always below 7 °C, mitigating tissue deterioration. The male muskox (older adult, collected in summer 2019) was shot in the neck directly after a goring injury to the flank sustained while headbutting another male muskox. One female muskox was shot in the top of the head (middle-aged adult, collected in summer 2018) and the second female muskox was shot in the neck (very old adult, collected in summer 2018), for subsistence hunting (Table ). The male muskox’s skull was approximately 4 cm thick in the frontal region (including sinuses), and 2 cm thick in the parietal cortex. The four bighorn sheep brains were collected as follows. The brain of an adult male bighorn sheep (Bighorn 1, five years old, collected in winter 2020) was acquired from Colorado Parks and Wildlife from a captive research herd after it was humanely euthanized (darted with NalMedA for sedation, euthanized with 20 ml Euthasol IV) due to a leg fracture. Formalin was injected around the brainstem and into the carotid artery for preservation before shipping. The male bighorn sheep’s skull was approximately 3 cm thick in the frontal region (including sinuses) and 1.5 cm thick in the parietal region. The brain of one wild adult female bighorn sheep (Bighorn 2, four years old, collected in winter 2020) was acquired from Utah Fish and Wildlife after euthanasia due to a Mycoplasma infection. The skull of the bighorn ewe was approximately 1.3 cm thick in the frontal region (including sinuses) and 0.8 cm thick in the parietal cortex. Two additional female bighorn sheep brains (Bighorn 3, five years old, collected in fall 1987; Bighorn 4, adult, collected in winter 2008) were archived in our collection and were provided by the Buffalo Zoo in 2003. After the muskoxen and sheep brains were removed from the skull, all samples were fixed in 10% formalin, either hours after death for the muskoxen, or within 36 h after death for the bighorn sheep. As a positive control for presence of phosphorylated tau, an archived human brain specimen with AD (male, 85 years old, Clinical Dementia Rating 3, Mini-Mental State Exam 11, Thal amyloid stage 4, Braak tangle stage V, postmortem interval 11 h, clinical diagnosis: severe cognitive impairment) and one with CTE (male, 69 years old, postmortem interval 9 h, clinical history: repetitive athletic head injuries, post-mortem diagnosis of advanced CTE, moderate cerebrovascular disease—athero-arteriolosclerosis; moderate hypoxic–ischemic encephalopathy, severe postmortem autolysis) were used in this study. MRI with superior soft tissue visualization was used to image the anatomical structure of the muskox and bighorn sheep brains. Coronal T2-weighed turbo spin-echo images were performed on a whole-body 7 Tesla (7 T) MRI scanner (Siemens Magnetom, Siemens Healthcare, Erlangen Germany) using a 1-channel transmit and 32-channel receive head coil (Nova Medical, Wilmington, Massachusetts) with the following parameters: repetition time (TR) 8000 ms, echo time (TE) 64 ms, number of sections 24, section thickness 1 mm, field-of-view 16 × 14 cm 2 , voxel size 0.5 × 0.5 × 1 mm 3 , scanning duration 6 min 20 s. Human tissue specimen imaging was performed in compliance with all institutional requirements. The tau-immunoreactive antibodies selected for this study were developed against human antigens. Their specificity in bighorn sheep and muskox tissues required validation to confirm their immunoreactivity. A basic local alignment tool (BLAST) compares protein sequences to sequence databases and calculates the statistical significance. The MAPT gene or protein sequence for tau was not available in bighorn sheep or muskoxen, therefore BLAST was applied to the predicted proteome of the domestic sheep ( Ovis aries ), the closest related species with MAPT sequence available on NCBI. Protein isoforms with the closest sequence homology and molecular weight to the requested sequence were ranked in terms of degree of identity in percent and E value. Homology values over 95% are considered acceptable, values over 85% are considered moderate. Brain tissue from the human subjects was sampled from Brodmann area 10. In the bovids, tissue was taken from the anterior region of the prefrontal cortex of the right hemisphere (Fig. ), and additionally from the parietal cortex of the right hemisphere for the three muskoxen. Each block was cut into 50 µm-thick sections on a vibratome (Leica VT1000S) and stored in phosphate-buffered saline (PBS, pH 7.0) solution with 0.1% sodium azide. Multiple phosphorylated tau antibodies were tested in this study, as no protocols existed for these bovid species, making it uncertain which epitopes were present in each. Antibodies recognizing ionized calcium-binding adaptor molecule-1 (Iba1) and glial fibrillary acidic protein (GFAP) were used to investigate microglial and astrocytic morphology, respectively. Three antibodies were tested to detect the presence of phosphorylated tau protein (p-tau). Anti-CP13 is a p-tau antibody clone that binds an epitope around Serine 202 (pSer202 tau), the AT8 clone recognizes epitopes around Serine 202 and Threonine 205 (pSer205/Thr205 tau), and the PHF-1 clone was raised against epitopes including Serine 396 and Serine 404 (pSer396/Ser404 tau). Both CP13 and PHF-1 antibody clones were generously supplied by the Davies laboratory (Feinstein Institute for Medical Research, Northwell Health, Manhasset, New York, USA). Anti-neurofilament clone SMI-312 detects medium- and heavy-chain phosphorylated neurofilament proteins (pNFP) and was used to investigate axonal damage. Anti-denatured myelin basic protein (dMBP) was applied to assess possible demyelination. The MOAB-2 antibody clone was applied against amyloid beta protein (Aβ), and anti-TDP43 clone binds an epitope around phosphorylated Serine 409 and Serine 410 (pSer409/410) to detect pathological proteins. Anti-collagen IV was applied to highlight blood vessel morphology. Additional details on antibodies are reported in Table . Sections were stained either as single instances or, in the case of the muskoxen, as series of ten sections, each 500 µm apart for stereological quantification. Primary antibody controls were performed by omitting the primary antibody and assessed that the secondary antibody did not generate non-specific staining (supplementary figures S1 and S2, provided as an online resource). All washes were performed with either PBS (Iba1, GFAP, dMBP, collagen IV) or Tris-buffered saline (TBS, pH 7.0) (AT8, CP13, PHF1, TDP43, MOAB-2, SMI-312) at room temperature on a shaker at 80 rpm for five minutes each. Antigen retrieval for Iba1 was performed by submerging free-floating tissue sections in a 10 mM EDTA solution (pH 8.0) in closed 15-ml tubes. The tubes were immersed in a water bath at 80 °C for 10 min. For GFAP, AT8, MOAB-2, dMBP, and SMI-312, antigen retrieval was performed by submerging free-floating tissue sections in citrate buffer (pH 6.0), then boiled at 100 °C for 10 min, followed by 5 min of cooling down. Antigen retrieval for collagen IV was performed by submerging sections in a 0.5 M acetic acid and 10 mg/ml pepsin solution at 37 °C for 8 min. After antigen retrieval, sections were transferred to 12-well plates using a glass hook and washed three times. Sections were then incubated with 0.3% hydrogen peroxide and 0.3% Triton X-100 in buffer for 30 min at room temperature, to inhibit endogenous peroxidase activity. The sections were then washed four times and blocked in buffer with 5% normal donkey serum (017000121, JacksonImmuno, West Grove, PA, USA) or normal goat serum (for TDP43 and collagen IV. 1002635372, Sigma, St. Louis, MD, USA) for 1 h, followed by an incubation in primary antibody (Iba1, GFAP, CP13, AT8, PHF1, MOAB-2, TDP43, SMI-312, dMBP, or collagen IV) in buffer with 5% normal donkey serum and 0.3% Triton X-100, at 4 °C overnight (Iba1, GFAP, AT8, MOAB-2, dMBP, SMI-312, collagen IV) or for 64 h (CP13, PHF1, TDP43); the control sections were incubated in buffer. After primary antibody incubation, sections were washed three times in buffer and 0.3% Triton X-100, then incubated with the appropriate secondary antibody (biotinylated donkey anti-rabbit, secondary antibody, 1:1000, 715065152, JacksonImmuno; biotinylated donkey anti-mouse, secondary antibody, 1:1000, 715065150, JacksonImmuno, biotinylated goat anti-rat secondary antibody 1:1000, BA9400, Vector laboratories, Burlingame, CA, USA) in a 5% normal donkey or goat serum and 0.3% Triton X-100 solution in buffer at room temperature for one hour (Iba1, GFAP, collagen IV, SMI-312) or two hours (AT8, CP13, PHF1, TDP43, MOAB-2, dMBP). After incubation with the secondary antibody, the sections were washed in buffer four times, then incubated with avidin–biotin solution (according to the manufacturer’s instructions, Vectastain ABC kit, Vector Laboratories, Burlingame, CA, USA) for one hour (Iba1, GFAP, MOAB-2, dMBP, collagen IV, SMI-312) or two hours (AT8, CP13, PHF1, TDP43) at room temperature. The sections were then washed four times, followed by an incubation with DAB peroxidase substrate (Vector Laboratories, according to the manufacturer’s instructions, DAB kit SK-4100) to reveal immunostaining. Samples were then mounted on gelatin-coated slides and left to dry overnight. After drying, the samples were counterstained with cresyl violet, dehydrated through an ethanol gradient, and coverslipped with DPX mounting medium. Luxol Fast-Blue was applied to observe demyelination as follows. Mounted sections were incubated in 0.1% Luxol Fast-Blue solution for one hour at 56 °C, then rinsed in distilled water. Sections were then differentiated in a 0.05% lithium carbonate solution for 30 s, followed by 70% ethanol for 30 s. Sections were rinsed in distilled water and checked microscopically for differentiation. Finally, slides were rinsed and differentiated in 95% ethanol then coverslipped as above. The blocking solution was increased from 5 to 10% normal donkey serum and primary antibody incubation was performed as in the immunohistochemistry protocol for both the CP13 and GFAP antibodies. After primary antibody incubation, sections were washed four times, protected from light, and incubated in biotinylated donkey anti-mouse antibody (1:1000, 715065150, JacksonImmuno) and anti-rabbit-AlexaFluor 488 (1:1000, A31570, ThermoFisher Scientific), followed by streptavidin-AlexaFluor 555 (1:500, A21206, ThermoFisher Scientific) in 10% normal donkey serum and 0.1% Triton X-100 in TBS at room temperature for two hours. The sections were washed four times and then incubated in 10% normal donkey serum and 0.1% Triton X-100 in TBS at room temperature for two hours. Sections were then washed four more times, mounted on SuperFrost slides and dried for one hour at 50 °C. Wells were drawn around the sections with an ImmEdge pen and sections were washed for 10 min. Sections were then incubated with TrueBlack (diluted 20 × in 70% ethanol) for 30 s each to reduce autofluorescence, then washed four times. The sections were then incubated with 4′,6-diamino-2-phenylindole dihydrochloride (DAPI, 250 ng/ml) for ten minutes in a humid chamber to stain cell nuclei and then washed a final time, after which they were mounted under Vectashield (H1000, Vector Laboratories) and coverslipped (24 × 50 mm No.1.5 ThermoFisher Scientific). Brightfield microscopy images were taken on an Axiophot brightfield microscope (Carl Zeiss Microscopy, Jena, Germany), with a 10×/0.32 Plan-Apochromat objective using StereoInvestigator (version 11.03, MBF Bioscience, Williston, VT, USA). Fluorescence images were taken on a CLSM 780 confocal microscope (Carl Zeiss Microscopy), using a 20×/0.8 DICII objective and DPSS 561-10 diode and Argon lasers at excitation wavelengths of 461, 555, and 488 nm. Confocal stacks in layers II and III of the cerebral cortex were imaged at 512 × 512 pixel resolution with a Z-step of 1 µm and a pinhole setting of 1 Airy unit for the red wavelength and optimal settings for gain and contrast. Images are presented as maximum intensity projections of the Z-stack using ZenBlue (version 3.3, Carl Zeiss Microscopy). Stereological quantification of tau immunostaining was performed using the optical fractionator workflow probe in StereoInvestigator (magnification × 10, counting frame size 700 × 700 µm, SRS grid layout at 100% of the region of interest, optical dissector height 11 µm with 2 µm of top and bottom guard zones, manual focus), on each muskox specimen in a series of 10 sections, each separated by 500 µm, and counted exhaustively. Cortical layers were manually contoured into layers I, II, III, IV–VI, and white matter. In each layer, different markers were placed for tau-immunoreactive neuropil threads (axonal or dendritic filaments composed of abnormally phosphorylated microtubule-associated tau protein), neuritic thread clusters (circular dense cluster of neuritic threads), and neurons (neuronal cells with tau-immunoreactive inclusions in the cytoplasm). Section contours were aligned manually and the coordinates were exported to create individual and combined heatmaps of tau-immunoreactive neuropil density distribution in Rstudio using the ggplot2 package . Annotated R code and raw data is available on GitHub ( https://github.com/NLAckermans/Ackermans2022BovidTBI.git ). To quantify tau-immunoreactive pathology accumulation in the sulci as opposed to gyri in the muskox specimens, sulcal depths were delimitated as 1/3 of the sulcus as in and pSer202 tau-immunoreactive markers in each region were quantified in StereoInvestigator using the same specifications as above. To quantify pSer202 tau-immunoreactive pathology around blood vessels in the muskoxen as compared to the CTE human control, one section from each individual was subjected to exhaustive counting of vessels larger than 30 µm in diameter at 2.5 × magnification with StereoInvestigator’s Optical Fractionator probe. The percentage of vessels immunostained with pSer202 tau located within 100 µm of the edge of the vessel and the average distance from the edge of the vessel were calculated and reported. None of the specimens showed any external brain trauma and the MRI scans were reviewed by a neuroradiologist for internal signs of TBI pathology such as shrinkage, acute trauma, or microhemorrhages in any of the bovid brains (Fig. ). Such alterations were not observed in our specimens. The male muskox brain only showed ex vivo artifacts of deformation and damage, due to field conditions and storage, but none were related to TBI (Fig. A), in comparison to a control specimen of a human brain with severe TBI (Fig. B). BLAST protein analysis for human MAP-tau against the estimated domestic sheep genome indicated the presence of protein with a comparable amino acid sequence in this species. The homology degree between the human (accession number P10636) and domestic sheep X1 (XP_042112063.1) and X4 (XP_027830174.1) isoforms was 80% and 87%, respectively. All three p-tau antibodies (anti-pSer202 tau, anti-pSer202/Thr205 tau, and anti-pSer396/Ser404 tau) were applied to all three species (Table ). All showed detectable signal in human and muskox, with anti-pSer202 tau specifically providing the most robust signal in muskox (Figs. A, A-I, controls in Fig. S1C); however, these antibodies rarely showed detectable signal in the bighorn sheep (Figs. B, J–L, controls in Fig. S1G–I). In the muskoxen, estimated population counts corrected for volume (Table ) and exhaustive counts (Table S1) revealed pSer202 tau-immunoreactive structures in all the muskox specimens (Figs. , ). pSer202-immunoreactive tau was present in neuropil threads, neuritic thread clusters, and neurons of the prefrontal and parietal cortex. The coefficient of error for tau-immunoreactive population estimates was in an acceptable range (< 15%) for all counts except neurons containing tau, in most cases due to their overall rarity in the sample (Table S2). In the bighorn sheep samples, tau-immunoreactive lesions mostly presented as pSer396/Ser404 tau-immunoreactive neuropil threads in the male bighorn, which were all in the grey matter and showed one grouping at the bottom of a sulcus (Fig. B, J). The positive control human CTE case showed a high number of tau-immunoreactive structures, especially neurons, heterogeneously clustered throughout the sample, especially around sulci and in the superficial cortical layers (Figs. D–F, B, E). The human AD case showed a much higher density of homogeneous tau immunoreactivity (Figs. A–C, A, D). Exhaustive stereological counting was, therefore, not performed in either the bighorn sheep or the human control samples. Tau-immunoreactive neuropil thread quantification Density of pSer202 tau-immunoreactive structures was calculated by dividing exhaustive counts by sample volume. Estimated population counts in muskoxen prefrontal cortex and parietal cortex indicated that neuropil threads were the most common structure by far in all individuals. At their highest density, neuropil threads were twice as numerous as neuritic thread clusters in the prefrontal cortex and 60 times more numerous than pSer202 tau-immunoreactive neurons. In the prefrontal cortex of the old female muskox pSer202 tau-immunoreactive neuropil thread density was over 500 times more numerous than in the old male and around 16 times more numerous than in the middle-aged female muskox (Table ). In the parietal cortex, the old female muskox still showed the highest density of pSer202 tau-immunoreactive neuropil threads with five times more than the old male and three times more than the middle-aged female muskox (Table ). In the prefrontal cortex, neuropil thread density decreased with layer depth in all three specimens (Fig. A). Whereas in the parietal cortex, the old female showed the highest density in layer II, the old male and middle-aged female showed decreasing densities with layer depth (Fig. D). Depending on the individual and cortical layer, the prefrontal cortex showed about a sixfold density increase as compared to the parietal cortex. Distribution heatmaps based on neuropil thread density show the highest concentrations in the superficial layers and at the bottom of sulci for all specimens and both brain regions (Fig. ), the latter of which was confirmed by stereological quantification (Table ). Despite the highest density of pSer202 tau-immunoreactive neuropil threads being in the old female muskox, the other two specimens showed higher neuropil densities in the sulci, especially in the prefrontal cortex, indicating a more numerous but also more even distribution of tau-immunoreactive neuropils in the old female (Fig. ). Overall, there were very few pSer202 tau-immunoreactive neuropil threads in the white matter (Figs. , ). Tau-immunoreactive neuritic thread cluster quantification The density of pSer202 tau-immunoreactive neuritic thread clusters was eight times higher in the old female than in the middle-aged female muskox and about 20 times higher in the middle-aged female muskox than in the male (Table ). The clusters were 5 to 100 times more frequent in the prefrontalthan in the parietal cortex depending on the individual. The middle-aged female muskox showed a pattern of decreasing density of neuritic thread clusters from the cortical surface to the deeper layers in both brain regions, while the male showed a similar pattern in the prefrontal cortex but a higher density in layer II of the parietal cortex. The old female muskox had the highest density of neuritic thread clusters in layer II in both brain regions with a more drastic pattern than the other two individuals (Fig. B, E). Overall, apart from being less numerous, neuritic thread clusters followed a similar distribution to the neuropil threads, especially in the parietal cortex, and were most concentrated in the superficial layers and at the base of the sulci (Fig. , Table ). Tau-immunoreactive neuron quantification Estimated and exhaustive counts both indicated pSer202 tau-immunoreactive neurons as relatively rare in all muskox specimens. However, they were five times more numerous in the old female than the middle-aged female and about 12 times denser in the middle-aged female than the old male muskox in the prefrontal cortex (Table ). Although the low counts make estimates somewhat unreliable, tau-containing neurons were at their highest densities in layers IV–VI for the old female and layer II for the middle-aged female muskox, with very few pSer202 tau-immunoreactive neurons in the parietal cortex (Fig. C, F). Some grouping is visible at the base of a sulcus in the prefrontal cortex (Fig. ) and slightly more tau-immunoreactive neurons clustered at the bottom of the sulci then elsewhere in the sample (Table ). Overall, pSer202 tau-immunoreactive neurons were more numerous in the prefrontal than the parietal cortex, especially for female muskoxen (Tables , S1). Non-neuronal cell types were not immunostained by any of the three tau antibodies applied in the bovid specimens, as opposed to what is commonly seen in human CTE cases. Tangle-like structures were observed in the parietal cortex of the muskox brains and are shown in Fig. . Blood vessel association with p-tau pathology All three types of pSer202 tau-immunoreactive structures were found around blood vessels (Figs. I, P). When quantified in muskoxen, 4–20% of blood vessels were associated with these structures (Table ). The old female muskox showed the highest percentage of blood vessels associated with pSer202 tau-immunoreactive structures, with an average distance from the blood vessel that was also smaller than in the other two cases at around 40 µm. As a comparison, the same quantification applied to a human CTE case revealed around 60% of blood vessels to be associated with these structures, at an average distance of around 30 µm. Glia Microglia and astrocytes were labeled using antibodies Iba1 and GFAP in all species. In the AD and CTE human specimens microglia with abnormal morphology and astrocytosis were observed in association with neurodegenerative changes, but such changes were rarely visible in either bovid species in our study (Figs. L, ). One microglial cluster was observed in the parietal cortex of the middle-aged female muskox (Fig. F). Furthermore, none of the muskoxen showed neurons or neurites immunoreactive for pSer202 tau in combination with activated astrocytes detected by GFAP (Fig. L, P), although this combination was apparent in the human AD specimen (Fig. D). Additional immunohistochemistry and staining Luxol Fast-Blue was used to investigate demyelination, as dMBP did not produce immunoreactivity (Supplementary Fig. S3G-I). Luxol Fast-Blue staining showed no pathological changes (Fig. S4). In addition, anti-pSer409/410, which was present in the human sample did not reveal any pathology in the bovids (Supplementary Fig. S3J-L). Similarly, anti-Aβ revealed Aβ plaques in the CTE human specimen and anti-collagen IV revealed immunostained blood vessels in the human specimens but not in the bovids (Supplementary Fig. S3). The pNFP-immunostained axons were well visible in the muskox specimens, but no axonal damage was observed (Fig. S5). Density of pSer202 tau-immunoreactive structures was calculated by dividing exhaustive counts by sample volume. Estimated population counts in muskoxen prefrontal cortex and parietal cortex indicated that neuropil threads were the most common structure by far in all individuals. At their highest density, neuropil threads were twice as numerous as neuritic thread clusters in the prefrontal cortex and 60 times more numerous than pSer202 tau-immunoreactive neurons. In the prefrontal cortex of the old female muskox pSer202 tau-immunoreactive neuropil thread density was over 500 times more numerous than in the old male and around 16 times more numerous than in the middle-aged female muskox (Table ). In the parietal cortex, the old female muskox still showed the highest density of pSer202 tau-immunoreactive neuropil threads with five times more than the old male and three times more than the middle-aged female muskox (Table ). In the prefrontal cortex, neuropil thread density decreased with layer depth in all three specimens (Fig. A). Whereas in the parietal cortex, the old female showed the highest density in layer II, the old male and middle-aged female showed decreasing densities with layer depth (Fig. D). Depending on the individual and cortical layer, the prefrontal cortex showed about a sixfold density increase as compared to the parietal cortex. Distribution heatmaps based on neuropil thread density show the highest concentrations in the superficial layers and at the bottom of sulci for all specimens and both brain regions (Fig. ), the latter of which was confirmed by stereological quantification (Table ). Despite the highest density of pSer202 tau-immunoreactive neuropil threads being in the old female muskox, the other two specimens showed higher neuropil densities in the sulci, especially in the prefrontal cortex, indicating a more numerous but also more even distribution of tau-immunoreactive neuropils in the old female (Fig. ). Overall, there were very few pSer202 tau-immunoreactive neuropil threads in the white matter (Figs. , ). The density of pSer202 tau-immunoreactive neuritic thread clusters was eight times higher in the old female than in the middle-aged female muskox and about 20 times higher in the middle-aged female muskox than in the male (Table ). The clusters were 5 to 100 times more frequent in the prefrontalthan in the parietal cortex depending on the individual. The middle-aged female muskox showed a pattern of decreasing density of neuritic thread clusters from the cortical surface to the deeper layers in both brain regions, while the male showed a similar pattern in the prefrontal cortex but a higher density in layer II of the parietal cortex. The old female muskox had the highest density of neuritic thread clusters in layer II in both brain regions with a more drastic pattern than the other two individuals (Fig. B, E). Overall, apart from being less numerous, neuritic thread clusters followed a similar distribution to the neuropil threads, especially in the parietal cortex, and were most concentrated in the superficial layers and at the base of the sulci (Fig. , Table ). Estimated and exhaustive counts both indicated pSer202 tau-immunoreactive neurons as relatively rare in all muskox specimens. However, they were five times more numerous in the old female than the middle-aged female and about 12 times denser in the middle-aged female than the old male muskox in the prefrontal cortex (Table ). Although the low counts make estimates somewhat unreliable, tau-containing neurons were at their highest densities in layers IV–VI for the old female and layer II for the middle-aged female muskox, with very few pSer202 tau-immunoreactive neurons in the parietal cortex (Fig. C, F). Some grouping is visible at the base of a sulcus in the prefrontal cortex (Fig. ) and slightly more tau-immunoreactive neurons clustered at the bottom of the sulci then elsewhere in the sample (Table ). Overall, pSer202 tau-immunoreactive neurons were more numerous in the prefrontal than the parietal cortex, especially for female muskoxen (Tables , S1). Non-neuronal cell types were not immunostained by any of the three tau antibodies applied in the bovid specimens, as opposed to what is commonly seen in human CTE cases. Tangle-like structures were observed in the parietal cortex of the muskox brains and are shown in Fig. . All three types of pSer202 tau-immunoreactive structures were found around blood vessels (Figs. I, P). When quantified in muskoxen, 4–20% of blood vessels were associated with these structures (Table ). The old female muskox showed the highest percentage of blood vessels associated with pSer202 tau-immunoreactive structures, with an average distance from the blood vessel that was also smaller than in the other two cases at around 40 µm. As a comparison, the same quantification applied to a human CTE case revealed around 60% of blood vessels to be associated with these structures, at an average distance of around 30 µm. Microglia and astrocytes were labeled using antibodies Iba1 and GFAP in all species. In the AD and CTE human specimens microglia with abnormal morphology and astrocytosis were observed in association with neurodegenerative changes, but such changes were rarely visible in either bovid species in our study (Figs. L, ). One microglial cluster was observed in the parietal cortex of the middle-aged female muskox (Fig. F). Furthermore, none of the muskoxen showed neurons or neurites immunoreactive for pSer202 tau in combination with activated astrocytes detected by GFAP (Fig. L, P), although this combination was apparent in the human AD specimen (Fig. D). Luxol Fast-Blue was used to investigate demyelination, as dMBP did not produce immunoreactivity (Supplementary Fig. S3G-I). Luxol Fast-Blue staining showed no pathological changes (Fig. S4). In addition, anti-pSer409/410, which was present in the human sample did not reveal any pathology in the bovids (Supplementary Fig. S3J-L). Similarly, anti-Aβ revealed Aβ plaques in the CTE human specimen and anti-collagen IV revealed immunostained blood vessels in the human specimens but not in the bovids (Supplementary Fig. S3). The pNFP-immunostained axons were well visible in the muskox specimens, but no axonal damage was observed (Fig. S5). This study assessed whether headbutting behavior in bovids is linked to TBI, mainly using tau-immunoreactive structures. We specifically chose bighorn sheep and muskoxen for this study as the most extreme representatives of their order. They headbutt at the highest forces in the animal kingdom and have evolved thick skulls and large headgear through sexual selection. The MRI scans showed no macroscopic signs of TBI and susceptibility-weighted imaging showed no microhemorrhages in these animals. This was expected, as the skulls were intact in all specimens and brains showed no outer macroscopic signs of brain trauma, such as regional shrinkage, which only becomes evident in the late stages of neurodegeneration in humans . Immunohistochemical p-tau immunostains on the bovid brains revealed a large amount of abnormally phosphorylated tau-immunoreactive neuropil threads, neurons, and neuritic thread clusters in the prefrontal cortex and the parietal cortex to a lesser extent. In the muskoxen, while anti-pSer202/Thr205 tau and anti-pSer396/Ser404 tau showed immunoreactivity, anti-pSer202 tau was most prevalent in all three structures and was, therefore, used for stereological quantification in both brain regions of the muskoxen. pSer202 tau-immunoreactive structures were concentrated in the superficial layers of the cerebral cortex, at the depths of the sulci, and occasionally around blood vessels. This distribution pattern was reminiscent of mild TBI or early-stage CTE cases, in American football players for example , and in other causes of repeated head trauma in humans . This pattern differs from the tau distribution pattern in cases of AD in which the neocortical distribution of tau NFTs exhibits a well-defined bilaminar pattern . Some pretangle-like structures with cytoplasmic tau immunoreactivity were observed in the muskoxen, mainly in the parietal cortex. Pretangles are an early form of NFT, representing a protective state of neuronal defense against abnormal tau in human neurodegeneration and consistent with an early CTE diagnosis. Microglia and astrocyte morphology in our samples appeared relatively normal, apart from a few microglial aggregates observed in the middle-aged muskox. Glial activation and association with tau-immunoreactive neurons is diagnostic of human TBI and potentially also CTE , but were not observed in our specimens. As perivascular tau can be indicative of CTE , we sought to quantify its severity in our specimens. Our results show that the old female muskox had a higher percentage of tau-immunoreactive-associated blood vessels than the other two muskoxen, closer to that measured in the human CTE case. Our original hypothesis stated that if headbutting bovids sustained brain trauma at all, it would be most evident in the males, as they headbutt more often and at higher forces than the females. We also expected the oldest specimens to show the most pathology because of the cumulative effects of brain injury potentially acquired over years of repetitive headbutting. Our quantitative results highlighting immunoreactive structures with anti-pSer202 tau told a somewhat different story, with the old male muskox showing the lowest number of pSer202 tau-immunoreactive structures and the old female showing the highest by a factor of up to 500 in the case of neuropil threads. This seems to indicate a cumulative aspect of chronic TBI in the older individual and surprisingly, a much higher avoidance of TBI in the male than in the female specimens. TBI caused by frequent and forceful headbutting in male muskoxen may be mitigated to a certain extent by their extreme anatomy, as a male muskox skull is on average 300% heavier than those of females . Muskox bulls most likely show additional dimorphism in protective soft tissue structures like neck musculature, forehead fat pads, and meninges, as well as in postcranial and vertebral elements . The discrepancy in tau-immunoreactive densities between the two females could be caused by individual differences, as varying degrees of tau phosphorylation are also present in middle-aged humans with no behavioral differences . Alternatively, social hierarchy may also have come into play, as new herd members of both sexes headbutt to establish dominance (Jamie Luce, The Muskox Farm, personal communication). In combination with these factors, age is likely the main contributing factor to the high tau-positivity in the old female, representative of accumulated chronic brain trauma in combination with age-related neurodegeneration, as indicated by the more evenly distributed neuropil threads and high occurrence of tau-immunoreactive structures in layer II. High numbers of tau-immunoreactive neurites, neurons, and neuritic thread clusters are indicative of repetitive brain trauma in human cases . While early-stage CTE shows no gross abnormalities in most brains, perivascular NFT clusters, neuropil threads, and astrocytic tangles are found in the sulcal depths of the brain at the microscopic level , as these areas are biomechanically vulnerable to trauma forces . CTE is generally characterized by an irregular, patchy distribution of argyrophilic tau-immunoreactive neocortical NFTs . When CTE increases in severity they become more densely distributed and are increasingly found across all brain regions, preferentially in layer II and the upper third of layer III , accompanied by increasing macroanatomical anomalies . Additionally, tau-immunoreactive fibrillar astrocytic tangles and dot-like or spindle-shaped neurites have also been observed in the white matter as well as the basal ganglia and brain stem . CTE originally described in the brains of boxers presented NFTs lacking Aβ plaques . More recently, other athletes presented focal p-tau abnormalities near focal axonal injury, alongside microhemorrhages, astrocytosis, and perivascular microgliosis, indicating a potential link to axonal injury and damage to the blood brain barrier . In parallel, military blast-related TBI and has revealed multiple areas of p-tau and glial immunoreactivity near small blood vessels . The tau-containing neurons and neuritic thread clusters in the muskoxen in the present study are not as neuropathologically advanced as NFTs in human neurodegenerative disorders, and direct comparison of p-tau quantification between the current study and human CTE studies is not possible because of sampling from different regions with various methods. Nevertheless, their presence alone, in addition to the characteristic distribution pattern, is the first evidence of any form of naturally occurring brain trauma in bovids. Abnormally phosphorylated tau has previously been detected using anti-pSer202/Thr205 tau in other aged animals, including bovids. Braak et al. recorded abnormally phosphorylated tau in the allocortical regions of aged sheep and goats, with pSer202/Thr205 tau-immunoreactive neurons that resembled NFTs of early-stage AD. Additionally, Härtig et al. reported two aged female American bison ( Bison bison ) as severely affected with abnormally phosphorylated tau in the prefrontal cortex [further explored in ]. This could present a parallel to the female muskoxen in the present study which were also the most severely affected. Although presence of abnormally phosphorylated tau can be a sign of ageing, the severity of tau pathology in these bison, muskoxen, and bighorn sheep might also be related to headbutting, as the non-headbutting species in the Härtig et al. study, such as reindeer, showed less severe tau-phosphorylation despite similarly advanced age. Other mammalian models present varying tauopathies when subjected to experimentally induced TBI. Histopathological analysis of mouse brains 24 h after lateral impact injury revealed dystrophic axons with hyperphosphorylated neurofilament proteins in proximity to reactive microglia and astrocytes . Another mouse study detected pSer202/Thr205 tau immunoreactivity in the neocortex and hippocampus of animals subjected to mild repetitive TBI after four and ten weeks . A similar experiment in rats showed an increase in p-tau expression in layers II/III of the motor cortex . Using ferrets as a gyrencephalic TBI model, Schwerin et al. demonstrated increased pSer202/Thr205 tau-immunoreactivity in the hippocampus after blast injuries, and a strong pSer202 tau-immunoreactivity in the superficial neocortical layers. In pigs where TBI was induced by rotational acceleration, Aβ and pSer396/Ser404 immunoreactive tau accumulated in axonal bulbs throughout the brains in addition to tau accumulations resembling pretangles in neuronal cytoplasm in the frontal, parietal, and temporal cortices colocalized with Aβ . However, experimentally induced TBI is not directly comparable to naturally occurring TBI, especially for smaller model animals and depending on the methodology, additional damage is created through craniotomy , which can impede cellular investigation due to inflammatory and morphological change independent of TBI . In humans, a single TBI leads to a high chance of developing neurodegenerative diseases later in life , and chronic TBI only increases those chances. Although little empirical data are available, muskoxen and bighorn sheep are known to headbutt every year during the rut. These bovids have a relatively short lifespan (muskoxen ♂: 10–12 years, ♀: 15–23 years approximately ; bighorn sheep ♂:10–12 years, ♀: 12–16 years approximately ) with an active reproductive period of 5–10 years in males. Each year the rut lasts from July to mid-October for muskoxen and is slightly shorter for bighorn sheep . It is uncertain how frequently individual males butt heads; however, single fights have been observed to comprise of 4–20 clashes, lasting multiple hours. A low estimate of three fights per week during the rut, with five clashes each, leads to about 210 clashes per year at around 60 km/h, averaging 2100 clashes in a lifetime. In comparison, studies on professional football players recorded a median number of 250 impacts per season at around 20 km/h, similarly averaging 2000 clashes in a lifetime for an average career of 8 years, furthering headbutting bovids as a model for sports-related TBI. The minimal p-tau immunostaining across bighorn sheep specimens could be attributed to the young age of the animals and the captive status of most of the specimens, impeding natural headbutting behaviors. Tissue degradation may have been an issue, as only one bighorn brain was injected with formalin before being shipped on ice overnight, while the muskox brains were preserved in formalin within hours after death. The limited immunoreactivity to pSer202/Thr205 tau and pSer202 tau in the bighorn sheep specimens could also indicate an absence of Serine 202 and Threonine 205 tau epitopes, as antibodies have less affinity for these epitopes in bovids, highlighting how tau physiopathology differs between species , which is further supported by an overall medium degree of homology between human and sheep tau protein reported in the BLAST protein analysis. Nevertheless, a human specimen with the same tau pathology present in the male bighorn sheep would have likely been diagnosed with mild CTE. Another factor that may have contributed to the individual and species differences reported here is tau isoform expression shifting between different neurodegenerative diseases and stages of NFT development. Anti-pSer202 tau and pSer202/Thr205 tau primarily recognize early and late NFT maturity levels, whereas anti-pSer396/Ser404 tau recognize more mature NFTs . A lower detection by anti-pSer396/Ser404 tau and anti-pSer202/Thr205 tau in combination with the highest tau detection by anti-pSer202 tau in muskoxen suggests that all three individuals were in the early stages of neurodegeneration. Additionally, lack of reaction could be related to species differences in the structures themselves and trauma progression, in addition to different physiological processes of TBI in the early stages of pathology . Additional discrepancies may include different tau sequences, protein folding, phosphorylation regulators, splicing and isoform expression, or other post-translational modifications of tau yet to be investigated in these species . Despite most NFT maturation studies being focused on ageing and AD, many pathologies are shared with CTE and can be used as comparisons. For example, tau shows cellular and molecular changes specific to CTE cases, notably isoform signatures and 4R to 3R tau ratio in relation to CTE severity . The results of the present study offer an avenue of further research on brain trauma in wild animals. Different artiodactyls, including cetaceans , display different headbutting and sparring behaviors , which present potential for further exploration of this phenomenon. As the pathology in our specimens appears to be in its early stages, future research should focus on whether further neurodegeneration is related with cognitive decline, as is the case in humans, where neurofibrillary degeneration can last more than 20 years. However, measuring cognitive function in bovids remains problematic, as no standardized behavioral tests exist for any bovid to date. The muskoxen’s capacity to survive yearly, repetitive, brain trauma and their similar frequency in this behavior to high-risk TBI individuals like football players and war veterans, presents them and likely other bovids including domestic sheep as a model with enough similarities to human CTE to explore the natural development of TBI. This is especially true given the translational difficulties of rodent studies. In conclusion, this study has highlighted that muskoxen, bighorn sheep, and possibly other bovids, exhibit tauopathies in relation to TBI caused by headbutting. Our results indicate that while both sexes headbutt and develop related neuropathology, males may be better protected by their thick skull and horns. The subsistence of these animals despite chronic tauopathies is a distinctive evolutionary adaptation that encourages further investigation into the pathophysiology of TBI and CTE in the bovid model as an avenue for TBI prevention and treatment in humans. Below is the link to the electronic supplementary material. Supplementary file1 (PDF 1632 kb) |
Education and Training: Key Factors in Global Occupational and Environmental Health | 7ae77091-4ecf-4646-88de-04e76b05bdc0 | 6748229 | Preventive Medicine[mh] | Education and training of health workers and clinician partners play an enormous role in the success of any prevention practice effort in the public health arena. Younger generations of professionals are necessary to sustain the gains current practitioners have achieved and this need is acutely felt in the fields of occupational and environmental health particularly in its global perspective. Even as the World Health Organization estimates that 24% of the global burden of disease and death is attributable to occupational and environmental exposures , a severe global shortage exists in the number of health professionals trained in environmental and occupational health (OEH). This shortage is especially acute for occupational physicians and is likely to worsen in future years with a globally growing working population, steady increases in chemical production, and increasing global demand for safer workplaces and healthier environments . Education and training are critically important for building global capacity in OEH. Training is needed to enhance the knowledge, skills, and practice of the current workforces in these fields and to build the future generation of workers. Working professionals, both those trained in OEH and those serving in roles without formal specialty instruction, need training to improve safety and health practices, design safer workplaces and provide cleaner environments. This Special Issue of the Annals of Global Health is the second volume of a series of three international, multi-disciplinary recurring courses on Occupational and Environmental Health: “Global Occupational and Environmental Determinants of Diseases: a Multidisciplinary and Multicultural Approach for Prevention,” University of Brescia, Italy . “Advanced International Training Course in Occupational and Environmental Health,” Chulabhorn Research Institute, Bangkok, Thailand . “Teaching Interventions Crossing Borders,” at the Ludwig Maximillian University of Munich, Germany . In addition to high-quality, hands-on training in occupational and environmental health, these courses provide unique opportunities for networking among students and professionals from both public health and medical specialties, from high-income and low and middle-income countries, and for developing multi-national approaches to continuing education that extend beyond the courses. Students and faculty attending the courses are encouraged to provide information about occupational and environmental health in their countries, including topics such as Occupational Safety and Health (OSH) services coverage and organization; workers’ benefits and compensation; ratification and implementation of policies from the International Labour Office (ILO) and the World Health Organization (WHO); preventive interventions in environmental health; and remediation of hazardous waste sites. The Munich course is additionally focused on another important task of the OSH experts: how to train workers, managers, and the community, as well as healthcare and safety professionals and political stakeholders. Course participants learn how to set up a competency-based training program for adult learners, according to a ‘train the trainer’ approach. Active learning methods are offered by all three courses and thus provide opportunities for networking and bi-directional exchange of ideas and experiences between faculty and participants. This was mainly achieved by the structure of the courses allowing ample time for groupwork, discussion, and presentation by participants. Education and training programs provide highly valuable outputs when participants are actively involved, are encouraged and have opportunities to illustrate data and critical policy aspects from their personal experience or work on a concrete problem (problem-based-learning). This is especially relevant for occupational and environmental health, given the extreme lack of information and research data from the global perspectives, the need for education and capacity building . The LDOH Foundation (Learning and Development of Occupational Health) has the mission to support professionals through developing and promoting quality education and information on Occupational Safety and Health. Influenced by our three courses, LDOH started supporting health professionals through developing and promoting good quality education and information on Occupational Safety and Health. A library of online and blended lessons, modules and courses from all over the world is available at the LDOH website ( https://ldoh.net/ ). LDOH is involved in two EU-funded projects focusing on capacity building for Occupational and Environmental Health. The first takes place in Turkey, where the Ministry of Health has implemented the project Scientific Performance of Public Health Institution of Turkey (ESPrIT – http://esprit-ohs.eu/en/ ), funded by the EU Horizon2020 program. ESPrIT includes a one-week training module in occupational health surveillance to illustrate the principles of adult learning: goal oriented, practical and problem centred. Step by step information (20-minute presentations) is provided on the different phases of a surveillance project and small interdisciplinary groups work on a mutually chosen subject. The other is an Erasmus+ programme coordinated by the University of Milan, Italy, jointly with ten universities in Central Asia and India in which educational materials for a blended learning Master Program in Occupational and Environmental Health are developed. This Special Issue is an example of what participants can produce to fill the knowledge gap about the global context, by providing data on specific situations in the many parts of the world from which they come. This data is absolutely necessary to provide more accurate estimates on the global burden of occupational and environmental diseases and to optimize preventive strategies. Data available today is almost completely related to Higher Income Countries and do not reflect accurately the current trends in most countries of the world. According to a recent survey by the International Occupational Medicine Society Collaborative on Global Trends in Occupational Medicine , the number of occupational physicians is not sufficient to meet demand. This trend is likely to worsen with increased demand and fewer trainees . In addition, legislative and government support for occupational medicine, while present in some countries, is not universal, due to a lack of understanding by employers and government officials of the benefits of an occupational medicine service. These two elements (shortage of professionals and lack of knowledge) clearly indicate the severe need for education and training in Global Occupational and Environmental Health. This discipline can also attract and offer potential for career development for students from Higher Income Countries, who are showing increasing interest in Global Health and may be attracted to the Occupational and Environmental field. Occupational safety and health is a broad field that addresses critical issues to help ensure workers and work places remain safe. No matter the profession, specific job or workplace, there are hazards present that need to be controlled. In the US, there were 4,836 fatal work injuries and approximately 2.9 million nonfatal workplace injuries and illnesses in 2015 . Globally, it is estimated that 2.3 million workers are killed each year (including 352,769 fatal work injuries and almost 2 million from occupational diseases) and there are over 313 million non-fatal occupational accidents . The burden is particularly heavy in low and middle-income countries, where manufacturing is mostly concentrated and health and safety law and its application is often not implemented properly . In the United States, the Occupational Safety and Health Act was enacted to “assure safe and healthful working conditions for working men and women .” The European Union Strategic Framework on Health and Safety at Work 2014–2020 lists the goal of “ensuring a safe and healthy work environment for over 217 million workers in the EU .” Other countries have enacted safety and health legislation to enhance safety and protect workers, with much of the legislation included in country profiles on the ILO website . There is a critical need for education and training in occupational safety and health to address basic and more advanced threats to worker health. Training is needed for the current and future OSH workforce. Working professionals, trained in OSH, together with those professionals working in roles potentially related to OSH, may gain capacity through short-courses and collaborate with OSH trained professionals to improve safety and health practices and design safer workplaces. Training the future occupational health workforce will enable the practices and actions taken to date to be sustained and to be built upon for the benefit of future generations. Educators need to look at the methods and practices used to provide occupational safety and health training. OSHA standards provide minimal guidance for training content and delivery methods . Developing and implementing training programs is difficult, as OSH is a broad field and encompasses many disciplines including, among others, occupational medicine and nursing, industrial/occupational hygiene, occupational safety, ergonomics, and occupational psychology. The focus of OSH practice is the worker or workplace, and each of these disciplines focus on a portion of the overall field of OSH. Industrial/occupational hygiene looks at worker exposures, occupational safety identifies workplace injury hazards, and occupational medicine examines health threats to workers and clinical care when they are injured or ill. Training must include how each of these distinct disciplines overlap and intersect. Interdisciplinary Learning OSH training should be interdisciplinary and include experiential learning as the basis for the development and implementation of training programs. This interdisciplinary approach will bridge the gap between the OSH disciplines, and refine the focus of OSH practice on improving the health and safety of workers and reducing hazards in the work place. Interdisciplinary training programs also provide opportunities for trainees to present their professional and country-based experience of a topic, thus improving their skills and knowledge base across disciplines. Erickson writes about “interdisciplinarity” as a means to increase safety performance. Interdisciplinarity “integrates knowledge from different disciplines. It blends the assumptions and practices of each into an integrative relationship to accomplish a larger purpose such as improving safety performance.” Recognizing the importance of this approach, the National Institute for Occupational Safety and Health (NIOSH) provides extramural funding to support Education and Research Centers (ERCs) to provide high-quality, interdisciplinary graduate training, research training, continuing education, and outreach in the core occupational safety and health disciplines . The ERCs are model training programs that are based on the understanding and awareness of the interdisciplinary nature of OSH professional practice. Experiential Learning Incorporating experiential learning into training programs provides valuable opportunities for trainees to visualize occupational hazards. The experiential learning theory (ELT) defines learning as “the process whereby knowledge is created through the transformation of experience. Knowledge results from the combination of grasping and transforming experience .” Experiential learning is a cycle that includes experiencing, reflecting, thinking, and acting. Training developed with the ELT identifies why what students are learning is important, it allows participants to practically navigate through content, and it uses real life examples and scenarios to anchor instruction. Experiential learning connects prior knowledge with new knowledge , allowing trainees to reflect on their personal experiences to transform the way they understand and act on what they learned. The New York and New Jersey ERC (NYNJ ERC) developed their Historical Perspectives on Occupational Safety and Health course in 2006 to provide trainees with interdisciplinary, experiential learning opportunities. In the Historical Perspectives course, trainees visit workplaces to experience how workers work and experience the occupational hazards and controls at work sites. These work place visits are a highlight of the academic training provided by the NYNJ ERC. Trainees’ comments on the course exemplify the value of interdisciplinary and experiential learning. Several direct quotes from the evaluation from the Historical Perspectives course include: “How crucial inter-professional education is to become a good, competent occupational medicine physician. Seeing workers at their own sites, helps you learn and appreciate their work conditions, their challenges and hazards on a whole new level that can never be learned as well from a classroom or clinic.” “This experience opens my eyes to question conditions of fellow workers, to ask more informed questions about what is being done to prevent not only injury in the work place but what can we do to systematically improve and prevent injury and promote well-being in the first place.” “Workers may not feel empowered to advocate for themselves if an employer asks them to complete tasks that are not compliant with the restrictions. This is something that I will keep in mind. With the knowledge that I am gaining through experiences like this tour I will be able to advocate for workers.” “The more powerful aspect of learning comes from experiencing things first hand, and from relevant ‘people,’ not just teachers. That kind of learning goes deeper and lasts longer.” Active Learning Based on the constructivism learning theory, when developing (adult) learning programs in OSH, as in any other discipline, learning is considered as an active process controlled by the learner and based on his/her previous experiences . Such experiences always differ from one person to another but are even more pronounced among participants from high, middle and low-income countries like those attending these courses. Therefore, in the setting of the described summer courses it was essential to take this previous experience into account. It also has to be acknowledged that rather than just listening to the lecturer, learners remember more when they interact with each other and work on real and relevant topics . Furthermore, thanks to current technology, every learner has all information right at hand. Therefore, the most important task of the teacher is to act as a facilitator and accompany the learner to find, understand and apply current evidence . Based on theoretical considerations, the Munich summer course uses a very interactive problem and employed a project-based learning approach during which participant develop a teaching intervention for their workers, managers, and community, targeting a concrete OSH problem in their home country. After successfully formulating “smart” (Specific, Measurable, Attainable, Relevant, Timely) learning objectives for the training, participants approach a teaching scheme structured in five phases of learning: 1) Adjusting and initial setting of the ‘tone’ and best mood for learning; 2) Reactivation of learners’ previous knowledge; 3) Information about the new knowledge; 4) Processing of the new information; 5) Evaluation, according to the ‘ARIPE’ (Adjust, Reactivate, Inform, Process, Evaluate) model . The ARIPE steps follow and support the learning process . For each ARIPE-step, participants become familiar with several interactive methods that can be applied within the different social forms of learning and that ensure an active role of the target group. Participants in the Munich course directly apply the ARIPE structure with the interactive methods to the teaching intervention they develop. In order to encourage exchange, participants work in small groups of up to four students, supported by tutors. Back home, participants apply their teaching intervention to their target groups and evaluate the outcomes. Models for Education and Training Based on the models of interdisciplinary, experiential, and active learning, all three training courses incorporate site visits, including visits to a marble quarry and milling operations, a steel plant and an automobile parts manufacturing facility in Italy, to the Ramathibodi Hospital, the largest health care center in Bangkok, and to a large car manufacturing facility in Germany. The addition of these workplace visits enable the participants to understand the complexities of work, experience how work impacts worker safety, and identify ways to control workplace hazards. Based on trainee feedback, the inclusion of these work place visits is a valuable learning experience. One trainee commented that he “gained practical workplace insight and experience” from the visits to the workplaces. Another commented, “The interdisciplinary approach helps us to resolve various issues and come up with common solutions that will benefit the workforce.” The model used by the NYNJ ERC and the Summer Programs is not new. Over 300 years ago Bernardino Ramazzini visited workplaces to observe how various types of work were performed and to discuss illnesses with workers . In 1910 Alice Hamilton investigated the “dangerous trades” in Illinois, visiting workplaces to see the conditions faced by workers and to understand the hazards present that were making workers ill . The concept of experiential learning in OSH, that you need to see the hazards and experience what workers are facing, is the core concept that is incorporated in the design of the programs described. Utilizing interdisciplinary and experiential learning is a powerful way to provide information, because the learners experience the conditions faced by workers. The programs offered by the NYNJ ERC, the Brescia Summer Program, the Bangkok International training course and the Munich Summer School are examples and models that can be utilized by others across the globe. By experiencing the work places, trainees are able to understand the value of interdisciplinary OSH teams, and working collaboratively ultimately improve health and safety conditions for workers and work places. OSH training should be interdisciplinary and include experiential learning as the basis for the development and implementation of training programs. This interdisciplinary approach will bridge the gap between the OSH disciplines, and refine the focus of OSH practice on improving the health and safety of workers and reducing hazards in the work place. Interdisciplinary training programs also provide opportunities for trainees to present their professional and country-based experience of a topic, thus improving their skills and knowledge base across disciplines. Erickson writes about “interdisciplinarity” as a means to increase safety performance. Interdisciplinarity “integrates knowledge from different disciplines. It blends the assumptions and practices of each into an integrative relationship to accomplish a larger purpose such as improving safety performance.” Recognizing the importance of this approach, the National Institute for Occupational Safety and Health (NIOSH) provides extramural funding to support Education and Research Centers (ERCs) to provide high-quality, interdisciplinary graduate training, research training, continuing education, and outreach in the core occupational safety and health disciplines . The ERCs are model training programs that are based on the understanding and awareness of the interdisciplinary nature of OSH professional practice. Incorporating experiential learning into training programs provides valuable opportunities for trainees to visualize occupational hazards. The experiential learning theory (ELT) defines learning as “the process whereby knowledge is created through the transformation of experience. Knowledge results from the combination of grasping and transforming experience .” Experiential learning is a cycle that includes experiencing, reflecting, thinking, and acting. Training developed with the ELT identifies why what students are learning is important, it allows participants to practically navigate through content, and it uses real life examples and scenarios to anchor instruction. Experiential learning connects prior knowledge with new knowledge , allowing trainees to reflect on their personal experiences to transform the way they understand and act on what they learned. The New York and New Jersey ERC (NYNJ ERC) developed their Historical Perspectives on Occupational Safety and Health course in 2006 to provide trainees with interdisciplinary, experiential learning opportunities. In the Historical Perspectives course, trainees visit workplaces to experience how workers work and experience the occupational hazards and controls at work sites. These work place visits are a highlight of the academic training provided by the NYNJ ERC. Trainees’ comments on the course exemplify the value of interdisciplinary and experiential learning. Several direct quotes from the evaluation from the Historical Perspectives course include: “How crucial inter-professional education is to become a good, competent occupational medicine physician. Seeing workers at their own sites, helps you learn and appreciate their work conditions, their challenges and hazards on a whole new level that can never be learned as well from a classroom or clinic.” “This experience opens my eyes to question conditions of fellow workers, to ask more informed questions about what is being done to prevent not only injury in the work place but what can we do to systematically improve and prevent injury and promote well-being in the first place.” “Workers may not feel empowered to advocate for themselves if an employer asks them to complete tasks that are not compliant with the restrictions. This is something that I will keep in mind. With the knowledge that I am gaining through experiences like this tour I will be able to advocate for workers.” “The more powerful aspect of learning comes from experiencing things first hand, and from relevant ‘people,’ not just teachers. That kind of learning goes deeper and lasts longer.” Based on the constructivism learning theory, when developing (adult) learning programs in OSH, as in any other discipline, learning is considered as an active process controlled by the learner and based on his/her previous experiences . Such experiences always differ from one person to another but are even more pronounced among participants from high, middle and low-income countries like those attending these courses. Therefore, in the setting of the described summer courses it was essential to take this previous experience into account. It also has to be acknowledged that rather than just listening to the lecturer, learners remember more when they interact with each other and work on real and relevant topics . Furthermore, thanks to current technology, every learner has all information right at hand. Therefore, the most important task of the teacher is to act as a facilitator and accompany the learner to find, understand and apply current evidence . Based on theoretical considerations, the Munich summer course uses a very interactive problem and employed a project-based learning approach during which participant develop a teaching intervention for their workers, managers, and community, targeting a concrete OSH problem in their home country. After successfully formulating “smart” (Specific, Measurable, Attainable, Relevant, Timely) learning objectives for the training, participants approach a teaching scheme structured in five phases of learning: 1) Adjusting and initial setting of the ‘tone’ and best mood for learning; 2) Reactivation of learners’ previous knowledge; 3) Information about the new knowledge; 4) Processing of the new information; 5) Evaluation, according to the ‘ARIPE’ (Adjust, Reactivate, Inform, Process, Evaluate) model . The ARIPE steps follow and support the learning process . For each ARIPE-step, participants become familiar with several interactive methods that can be applied within the different social forms of learning and that ensure an active role of the target group. Participants in the Munich course directly apply the ARIPE structure with the interactive methods to the teaching intervention they develop. In order to encourage exchange, participants work in small groups of up to four students, supported by tutors. Back home, participants apply their teaching intervention to their target groups and evaluate the outcomes. Based on the models of interdisciplinary, experiential, and active learning, all three training courses incorporate site visits, including visits to a marble quarry and milling operations, a steel plant and an automobile parts manufacturing facility in Italy, to the Ramathibodi Hospital, the largest health care center in Bangkok, and to a large car manufacturing facility in Germany. The addition of these workplace visits enable the participants to understand the complexities of work, experience how work impacts worker safety, and identify ways to control workplace hazards. Based on trainee feedback, the inclusion of these work place visits is a valuable learning experience. One trainee commented that he “gained practical workplace insight and experience” from the visits to the workplaces. Another commented, “The interdisciplinary approach helps us to resolve various issues and come up with common solutions that will benefit the workforce.” The model used by the NYNJ ERC and the Summer Programs is not new. Over 300 years ago Bernardino Ramazzini visited workplaces to observe how various types of work were performed and to discuss illnesses with workers . In 1910 Alice Hamilton investigated the “dangerous trades” in Illinois, visiting workplaces to see the conditions faced by workers and to understand the hazards present that were making workers ill . The concept of experiential learning in OSH, that you need to see the hazards and experience what workers are facing, is the core concept that is incorporated in the design of the programs described. Utilizing interdisciplinary and experiential learning is a powerful way to provide information, because the learners experience the conditions faced by workers. The programs offered by the NYNJ ERC, the Brescia Summer Program, the Bangkok International training course and the Munich Summer School are examples and models that can be utilized by others across the globe. By experiencing the work places, trainees are able to understand the value of interdisciplinary OSH teams, and working collaboratively ultimately improve health and safety conditions for workers and work places. Given the deficit of specialist expertise in the field globally, education and training in Occupational and Environmental health must be given a greater priority at the national, regional and global level. Capacity building as a priority may be achieved through a ‘Train the trainer’ approach, where the target groups plays an active role. Courses should be intensified and offered more extensively also to generalists in clinical medicine and public health, covering all geographical areas and with special focus on sectors with high occupational and environmental hazards. Training should be fully interdisciplinary using an active, experiential learning approach. Aspects of Occupational and Environmental health should be integrated into standard public health and medical curricula when possible, to optimize human resources and capacity building. This Special Issue of the Annals of Global Health will show-case the professional efforts of the participants of the three international courses described above. While some participants had occupational health training and are authors here, the majority were generalists in clinical medicine or public health, who were engaged in occupational/environmental health activity in their home countries, as a ‘collateral’ assignment. Their documentation of local problems and solutions from their countries, described here, are first-hand “reports from the field” and help to close the existing knowledge gap between countries, regarding the occupational health agenda globally. |
Proteomic analysis reveals chromatin remodeling as a potential therapeutical target in neuroblastoma | 71a20fa1-d470-403b-b0c9-5803534ab9cb | 11866834 | Biochemistry[mh] | Neuroblastoma (NB), a cancer that arises from the developing sympathetic nervous system, is the most common extracranial solid tumor in children, accounting for approximately 15% of childhood cancer-related deaths . The genetic, morphological, and clinical heterogeneity of NB limits the effectiveness of current treatment modalities . Prior to 2009, evidence-based treatment for high-risk NB involved a multimodal approach, including surgery, local radiotherapy, and combination chemotherapy regimens, often supplemented by consolidation protocols that utilized autologous stem cell transplantation . The introduction of immunotherapy using anti-disialoganglioside (GD2) monoclonal antibodies (mAbs) in combination with chemotherapy has significantly improved the 5-year survival rate for patients with metastatic NB, increasing it from less than 20% to over 50% . The recently revised Children’s Oncology Group (COG) classifier incorporates factors such as age, INRG stage, and tumor histologic and genetic features, including MYCN amplification status and chromosomal aberrations . Although many potential biomarkers identified through next-generation sequencing techniques have been studied, very few have demonstrated independent prognostic significance in multivariable analyses compared to established classifiers like MYCN and stage . To date, large-scale analyses, including those from the Cancer Cell Line Encyclopedia (CCLE), have primarily focused on genetic information, while a thorough exploration of the proteomic profile remains largely unexplored . The protein expression profile of clinical tissues from NB patients is still unknown. Recent studies utilizing advanced high-throughput “omics” technologies have uncovered numerous genetic and genomic alterations, as well as dysregulated pathways that drive the onset, progression, and treatment resistance of NB . Investigating the molecular changes within these pathways may help identify new therapeutic targets and approaches for NB. Histological features are particularly impactful in predicting outcomes for patients with locoregional disease and for children with metastatic disease . In NB, the degree of differentiation and the mitosis-karyorrhexis index have independent prognostic significance and, alongside age, inform the favorable or unfavorable International Neuroblastoma Pathology Classification (INPC) histology . In this study, we aim to identify key gene models and hub genes associated with pathological histology in NB, based on a proteomic analysis of tumor tissues from NB, GNB, GN, and AG. Furthermore, we plan to conduct computational screening for drug repositioning for NB and validate these potential drugs in NB cell models. Study design and clinical characteristics of the patient groups All subjects were enrolled at the Department of Pediatric Surgery in Hunan Childen’s Hospital and clinically diagnosed as ganglioneuroma (GN), ganglioneuroblastoma (GNB) and NB patients. All patients underwent laboratory and imaging examinations at the hospital. Non-renal NB were excluded. Clinical samples consist of 9 NB samples, 6 GN samples, 4 GNB samples and 5 adrenal gland (AG) samples. Proteins of the tumor tissues were enriched and quantitatively analyzed by data-independent acquisition (DIA) LC–MS/MS. This study was approved by the Medical Ethics Committee of Hunan Children’s Hospital (No: HCHLL-2021-110). The clinical information and protein expression profiles of the patients were collected for prognosis analysis of NB study. The clinical characteristics and histopathological phenotypes are described in the Table S1 of Additional file . In this study there is another NB cohort for verification experiment via western blot, including 6 NB, 2 GNB and 5 GN. Their clinical characteristics and histopathological classification are described in the Table S2 of Additional file . Sample preparation and LC–MS/MS analysis The sample was grinded with liquid nitrogen into cell powder and then transferred to four volumes of lysis buffer (8 M urea, 1% protease inhibitor cocktail), followed by sonication three times on ice. The remaining debris was removed by centrifugation at 12,000 g at 4 °C for 10 min. For digestion, the protein solution was reduced with 5 mM dithiothreitol for 30 min at 56 °C and alkylated with 11 mM iodoacetamide for 15 min at room temperature in darkness. The protein sample was then diluted by adding 100 mM TEAB to urea concentration less than 2 M. Protein samples were digested by trypsin in a trypsin-to-protein mass ratio of 1:50 for the first digestion overnight and 1:100 trypsin-to-protein mass ratio for a second 4 h-digestion. The digested peptides were desalted by Strata X C18 SPE column (Phenomenex) and vacuum-dried. Peptides are fractionated using high pH reverse-phase HPLC. The chromatography column used is an Agilent 300 Extend C18 (5 μm particle size, 4.6 mm inner diameter, 250 mm length). The gradient for peptide fractionation is set at 8%–32% acetonitrile, pH 9, over 60 min. The peptides are then combined into 12 fractions. Following the instructions in the iRT reagent manual, an appropriate amount of iRT reagent is added to each combined fraction, and the samples are vacuum freeze-dried for subsequent processing. The peptides are dissolved in mobile phase A of the liquid chromatography and then separated using a NanoElute ultra-high-performance liquid chromatography system. Mobile phase A consists of 0.1% formic acid and 2% acetonitrile in water, and mobile phase B is acetonitrile with 0.1% formic acid. The LC gradient is set to 90 min with a flow rate maintained at 450 nL/min. After separation by the UHPLC system, peptides are ionized in the capillary ion source and then injected into the tims-TOF Pro mass spectrometer for data acquisition. The ion source voltage is set to 1.7 kV, and both the precursor ions and their secondary fragments are detected and analyzed using TOF. Data acquisition is performed in data-independent parallel accumulation and serial fragmentation (dia-PASEF) mode. The first-stage mass spectrum scan range is set from 100–1700 m/z, with 16 PASEF acquisitions following each primary mass spectrum. The second-stage mass spectrum scan range is set from 400–1200 m/z, with a window of 25 m/z per scan. DDA data is searched using MSFragger (v 2.3). The database used is SwissProt_Human (20,380 sequences), with a decoy database added to calculate the false discovery rate (FDR) due to random matching. The enzyme setting is Trypsin/P, allowing up to 2 missed cleavage sites. The precursor ion mass tolerance is set at 20 ppm, and the fragment ion mass tolerance is 0.02 Da. Cysteine alkylation is set as a fixed modification, with variable modifications set for methionine oxidation and protein N-terminal acetylation. The FDR for protein and PSM identification is set at 1%. DIA data is processed using Skyline (v 20.1.0) software, importing the corresponding spectral library and adding the iRT parameters under Prediction. Transition precursor ion charges are set to 2, 3, 4, and 5, and fragment ion charges to 1 and 2. The top six ions with the highest intensity in the spectral library are extracted for peptide quantification. After generating the decoy library, DIA data is imported and FDR filtering is applied using the mProphet algorithm. The MSstats R package is used to obtain relative quantification results for proteins. Proteomic data analysis Enrichment of biological function of DEGs was performed via online platform metascape ( https://metascape.org/gp/index.html ). In the analysis the functional set (GO molecular functions) and pathways (GO Biological Processes, Reactome Gene Set, KEGG Pathway, WikiPathways, Canonical Pathways, Hallmark Gene Sets and BioCarta Gene Sets) were involved and applied for enrichment of biological processes. WGCNA for histopathology-related prognostic gene modules of NB To find out core clinicopathological prognostic gene modules, weighted gene co-expression network analysis (WGCNA) is used to cluster the genes with similar co-expression mode into a gene module. Then the correlation between gene modules and clinicopathological phenotypes was analyzed for the core prognostic gene modules of NB. From a pathological perspective, there is a Shimada classification method that divides NB into two types: Favorable Histology Group (FH) and Unfavorable Histology Group (UH) . The correlation between gene expression modules and clinical prognosis in the clinical datasets of GN, GNB and NB was analyzed using Spearman correlation analysis. The criteria for selecting significant DEGs for WGCNA were described as following. Firstly, the DEGs(|FC|≥ 2, P < 0.05) of each group NB/AG, GNB/AG, GN/AG, NB + GNB/AG, NB + GNB + GN/AG were integrated together. Then the DEGs must be identified and quantified in at least 21 samples among total 24 proteomic samples. Finally, it leads to 3313 DEGs for further WGCNA analysis. Co-expression gene modules are constructed from the DEGs and WGCNA was performed using Sangerbox ( http://sangerbox.com/home.html ) online platform. Heatmaps depicting the correlation of each module with clinical prognosis were generated. We used the "plot" function to generate scatter plots between gene significance (GS) and module membership (MM) within each module in order to understand the importance of highly connected genes within the module. The differentially expressed proteins in NB were subjected to protein–protein interaction (PPI) network analysis using the STRING database (version 11.5, http://string-db.org/ ) and Cytoscape software (version 3.8.2, https://cytoscape.org/ ) to identify hub genes. Basic settings of STRING were described as following. Active interaction sources include textmining, experiments, databases, co-expression, neighborhood, gene fusion and co-occurrence. Minimum required interaction score was set as medium confidence (0.400). Max number of interactors to show in the 1st shell was set as “none/query proteins only” and 2nd shell was “none”. The plugin, cytoHubba (version 0.1), was used to identify hub genes, and the top 15 hubba nodes’ scores were calculated by using the Maximum Clique Centrality (MCC) method and ranked. GO pathway enrichment analysis was conducted on hub genes. P < 0.05 in pathway enrichment analysis was considered statistically significant. Transcriptome proximity analysis in L1000FWD and CMap The reversed expression profile of core prognostic modules was used to query L1000FWD online database ( http://amp.pharm.mssm.edu/L1000FWD ). The similarity score of predicted drugs were given to quantify the similarity between the reversed protein expression of the core prognostic regulated genes and their drug-perturbed gene expression profiles in L1000FWD. Based on the similarity score, p-value, comprehensive score and other factors, the relevant candidate drug molecules were selected, so fa their similarity score absolute value greater than 0.1. The CMap ( https://clue.io ) is another online database for large-scale drug query. As a complementary query in L1000FWD, the core prognostic gene modules were used to query CMap. The similarity of the query to each CMap signature was computed and yielded a rank-ordered list of the signatures. A tau ≥ 90 was considered as convinced strong score. Bioassays CCK8 assay The human neuroblastoma cell line SH-N-AS was cultured in EMEM (with NEAA) containing 10% fetal bovine serum. Cells were seeded into 96-well plates (1000 per well), treated for 24 h with different concentration of PI-828, mocetinostat, clofarabine, ethacrynic acid and mafenide (10 –8 –10 –5 mol/L, all purchased from MCE). The cells were incubated at 37℃ for 0.5 h with CCK-8 Kit (Abiowell, China). The experiment is performed according to the manual of the kit. RNA extraction and quantitative real-time PCR (qRT-PCR) Total RNA from clinical tumor tissues were extracted with trizol. Reverse transcription was performed using the mRNA reverse transcription kit (CW2569, CWBIO, China). qRT-PCR was carried out on the CFX Connect Real-Time PCR system (1855201, Bio-Rad Laboratories, Hercules, CA, USA) using UltraSYBR Mixture (CW2601, CWBIO, China). The expression levels of all target genes were normalized to β-actin. The quantification of mRNA was carried out using the 2 −ΔΔCt method. The primers used are shown in Table . Western blot (WB) The proteins from the samples were separated by SDS-PAGE and then transferred to a polyvinylidene difluoride (PVDF) membrane. The blocked membrane was incubated with the indicated primary antibodies. The antibodies used were ALYREF (1:1000, 16690-1-AP, Proteintech, USA), SMARCA4/BRG1 (1:1000, ab110641, Abcam) and β-actin (1:5000, AWA80002, Abiowell, China). Next, the membrane was incubated with HRP-conjugated goat anti-rabbit IgG (SA00001-2, Proteintech, USA). The immunoreactive bands were visualized using a SuperECL Plus ultrasensitive luminescent liquid (awb0005, Abiowell, China). Data analysis The survival analysis and immune filtration analysis of the hub genes were performed by using Sangerbox online platform. The data was presented as mean ± standard deviation. All data were collected from at least three independent experiments. Student's t-test or Wilcoxon test was used to analyze the differences between the two groups. One-way analysis of variance (ANOVA) was used for comparisons among multiple groups. Survival curves were depicted using Kaplan–Meier plots, and the log-rank test was used for comparisons. The software used were R and GraphPad Prism (v8.0.1, GraphPad Software, USA). P < 0.05 was considered statistically significant. All subjects were enrolled at the Department of Pediatric Surgery in Hunan Childen’s Hospital and clinically diagnosed as ganglioneuroma (GN), ganglioneuroblastoma (GNB) and NB patients. All patients underwent laboratory and imaging examinations at the hospital. Non-renal NB were excluded. Clinical samples consist of 9 NB samples, 6 GN samples, 4 GNB samples and 5 adrenal gland (AG) samples. Proteins of the tumor tissues were enriched and quantitatively analyzed by data-independent acquisition (DIA) LC–MS/MS. This study was approved by the Medical Ethics Committee of Hunan Children’s Hospital (No: HCHLL-2021-110). The clinical information and protein expression profiles of the patients were collected for prognosis analysis of NB study. The clinical characteristics and histopathological phenotypes are described in the Table S1 of Additional file . In this study there is another NB cohort for verification experiment via western blot, including 6 NB, 2 GNB and 5 GN. Their clinical characteristics and histopathological classification are described in the Table S2 of Additional file . The sample was grinded with liquid nitrogen into cell powder and then transferred to four volumes of lysis buffer (8 M urea, 1% protease inhibitor cocktail), followed by sonication three times on ice. The remaining debris was removed by centrifugation at 12,000 g at 4 °C for 10 min. For digestion, the protein solution was reduced with 5 mM dithiothreitol for 30 min at 56 °C and alkylated with 11 mM iodoacetamide for 15 min at room temperature in darkness. The protein sample was then diluted by adding 100 mM TEAB to urea concentration less than 2 M. Protein samples were digested by trypsin in a trypsin-to-protein mass ratio of 1:50 for the first digestion overnight and 1:100 trypsin-to-protein mass ratio for a second 4 h-digestion. The digested peptides were desalted by Strata X C18 SPE column (Phenomenex) and vacuum-dried. Peptides are fractionated using high pH reverse-phase HPLC. The chromatography column used is an Agilent 300 Extend C18 (5 μm particle size, 4.6 mm inner diameter, 250 mm length). The gradient for peptide fractionation is set at 8%–32% acetonitrile, pH 9, over 60 min. The peptides are then combined into 12 fractions. Following the instructions in the iRT reagent manual, an appropriate amount of iRT reagent is added to each combined fraction, and the samples are vacuum freeze-dried for subsequent processing. The peptides are dissolved in mobile phase A of the liquid chromatography and then separated using a NanoElute ultra-high-performance liquid chromatography system. Mobile phase A consists of 0.1% formic acid and 2% acetonitrile in water, and mobile phase B is acetonitrile with 0.1% formic acid. The LC gradient is set to 90 min with a flow rate maintained at 450 nL/min. After separation by the UHPLC system, peptides are ionized in the capillary ion source and then injected into the tims-TOF Pro mass spectrometer for data acquisition. The ion source voltage is set to 1.7 kV, and both the precursor ions and their secondary fragments are detected and analyzed using TOF. Data acquisition is performed in data-independent parallel accumulation and serial fragmentation (dia-PASEF) mode. The first-stage mass spectrum scan range is set from 100–1700 m/z, with 16 PASEF acquisitions following each primary mass spectrum. The second-stage mass spectrum scan range is set from 400–1200 m/z, with a window of 25 m/z per scan. DDA data is searched using MSFragger (v 2.3). The database used is SwissProt_Human (20,380 sequences), with a decoy database added to calculate the false discovery rate (FDR) due to random matching. The enzyme setting is Trypsin/P, allowing up to 2 missed cleavage sites. The precursor ion mass tolerance is set at 20 ppm, and the fragment ion mass tolerance is 0.02 Da. Cysteine alkylation is set as a fixed modification, with variable modifications set for methionine oxidation and protein N-terminal acetylation. The FDR for protein and PSM identification is set at 1%. DIA data is processed using Skyline (v 20.1.0) software, importing the corresponding spectral library and adding the iRT parameters under Prediction. Transition precursor ion charges are set to 2, 3, 4, and 5, and fragment ion charges to 1 and 2. The top six ions with the highest intensity in the spectral library are extracted for peptide quantification. After generating the decoy library, DIA data is imported and FDR filtering is applied using the mProphet algorithm. The MSstats R package is used to obtain relative quantification results for proteins. Enrichment of biological function of DEGs was performed via online platform metascape ( https://metascape.org/gp/index.html ). In the analysis the functional set (GO molecular functions) and pathways (GO Biological Processes, Reactome Gene Set, KEGG Pathway, WikiPathways, Canonical Pathways, Hallmark Gene Sets and BioCarta Gene Sets) were involved and applied for enrichment of biological processes. To find out core clinicopathological prognostic gene modules, weighted gene co-expression network analysis (WGCNA) is used to cluster the genes with similar co-expression mode into a gene module. Then the correlation between gene modules and clinicopathological phenotypes was analyzed for the core prognostic gene modules of NB. From a pathological perspective, there is a Shimada classification method that divides NB into two types: Favorable Histology Group (FH) and Unfavorable Histology Group (UH) . The correlation between gene expression modules and clinical prognosis in the clinical datasets of GN, GNB and NB was analyzed using Spearman correlation analysis. The criteria for selecting significant DEGs for WGCNA were described as following. Firstly, the DEGs(|FC|≥ 2, P < 0.05) of each group NB/AG, GNB/AG, GN/AG, NB + GNB/AG, NB + GNB + GN/AG were integrated together. Then the DEGs must be identified and quantified in at least 21 samples among total 24 proteomic samples. Finally, it leads to 3313 DEGs for further WGCNA analysis. Co-expression gene modules are constructed from the DEGs and WGCNA was performed using Sangerbox ( http://sangerbox.com/home.html ) online platform. Heatmaps depicting the correlation of each module with clinical prognosis were generated. We used the "plot" function to generate scatter plots between gene significance (GS) and module membership (MM) within each module in order to understand the importance of highly connected genes within the module. The differentially expressed proteins in NB were subjected to protein–protein interaction (PPI) network analysis using the STRING database (version 11.5, http://string-db.org/ ) and Cytoscape software (version 3.8.2, https://cytoscape.org/ ) to identify hub genes. Basic settings of STRING were described as following. Active interaction sources include textmining, experiments, databases, co-expression, neighborhood, gene fusion and co-occurrence. Minimum required interaction score was set as medium confidence (0.400). Max number of interactors to show in the 1st shell was set as “none/query proteins only” and 2nd shell was “none”. The plugin, cytoHubba (version 0.1), was used to identify hub genes, and the top 15 hubba nodes’ scores were calculated by using the Maximum Clique Centrality (MCC) method and ranked. GO pathway enrichment analysis was conducted on hub genes. P < 0.05 in pathway enrichment analysis was considered statistically significant. The reversed expression profile of core prognostic modules was used to query L1000FWD online database ( http://amp.pharm.mssm.edu/L1000FWD ). The similarity score of predicted drugs were given to quantify the similarity between the reversed protein expression of the core prognostic regulated genes and their drug-perturbed gene expression profiles in L1000FWD. Based on the similarity score, p-value, comprehensive score and other factors, the relevant candidate drug molecules were selected, so fa their similarity score absolute value greater than 0.1. The CMap ( https://clue.io ) is another online database for large-scale drug query. As a complementary query in L1000FWD, the core prognostic gene modules were used to query CMap. The similarity of the query to each CMap signature was computed and yielded a rank-ordered list of the signatures. A tau ≥ 90 was considered as convinced strong score. CCK8 assay The human neuroblastoma cell line SH-N-AS was cultured in EMEM (with NEAA) containing 10% fetal bovine serum. Cells were seeded into 96-well plates (1000 per well), treated for 24 h with different concentration of PI-828, mocetinostat, clofarabine, ethacrynic acid and mafenide (10 –8 –10 –5 mol/L, all purchased from MCE). The cells were incubated at 37℃ for 0.5 h with CCK-8 Kit (Abiowell, China). The experiment is performed according to the manual of the kit. RNA extraction and quantitative real-time PCR (qRT-PCR) Total RNA from clinical tumor tissues were extracted with trizol. Reverse transcription was performed using the mRNA reverse transcription kit (CW2569, CWBIO, China). qRT-PCR was carried out on the CFX Connect Real-Time PCR system (1855201, Bio-Rad Laboratories, Hercules, CA, USA) using UltraSYBR Mixture (CW2601, CWBIO, China). The expression levels of all target genes were normalized to β-actin. The quantification of mRNA was carried out using the 2 −ΔΔCt method. The primers used are shown in Table . Western blot (WB) The proteins from the samples were separated by SDS-PAGE and then transferred to a polyvinylidene difluoride (PVDF) membrane. The blocked membrane was incubated with the indicated primary antibodies. The antibodies used were ALYREF (1:1000, 16690-1-AP, Proteintech, USA), SMARCA4/BRG1 (1:1000, ab110641, Abcam) and β-actin (1:5000, AWA80002, Abiowell, China). Next, the membrane was incubated with HRP-conjugated goat anti-rabbit IgG (SA00001-2, Proteintech, USA). The immunoreactive bands were visualized using a SuperECL Plus ultrasensitive luminescent liquid (awb0005, Abiowell, China). The human neuroblastoma cell line SH-N-AS was cultured in EMEM (with NEAA) containing 10% fetal bovine serum. Cells were seeded into 96-well plates (1000 per well), treated for 24 h with different concentration of PI-828, mocetinostat, clofarabine, ethacrynic acid and mafenide (10 –8 –10 –5 mol/L, all purchased from MCE). The cells were incubated at 37℃ for 0.5 h with CCK-8 Kit (Abiowell, China). The experiment is performed according to the manual of the kit. Total RNA from clinical tumor tissues were extracted with trizol. Reverse transcription was performed using the mRNA reverse transcription kit (CW2569, CWBIO, China). qRT-PCR was carried out on the CFX Connect Real-Time PCR system (1855201, Bio-Rad Laboratories, Hercules, CA, USA) using UltraSYBR Mixture (CW2601, CWBIO, China). The expression levels of all target genes were normalized to β-actin. The quantification of mRNA was carried out using the 2 −ΔΔCt method. The primers used are shown in Table . The proteins from the samples were separated by SDS-PAGE and then transferred to a polyvinylidene difluoride (PVDF) membrane. The blocked membrane was incubated with the indicated primary antibodies. The antibodies used were ALYREF (1:1000, 16690-1-AP, Proteintech, USA), SMARCA4/BRG1 (1:1000, ab110641, Abcam) and β-actin (1:5000, AWA80002, Abiowell, China). Next, the membrane was incubated with HRP-conjugated goat anti-rabbit IgG (SA00001-2, Proteintech, USA). The immunoreactive bands were visualized using a SuperECL Plus ultrasensitive luminescent liquid (awb0005, Abiowell, China). The survival analysis and immune filtration analysis of the hub genes were performed by using Sangerbox online platform. The data was presented as mean ± standard deviation. All data were collected from at least three independent experiments. Student's t-test or Wilcoxon test was used to analyze the differences between the two groups. One-way analysis of variance (ANOVA) was used for comparisons among multiple groups. Survival curves were depicted using Kaplan–Meier plots, and the log-rank test was used for comparisons. The software used were R and GraphPad Prism (v8.0.1, GraphPad Software, USA). P < 0.05 was considered statistically significant. Identification of differentially regulated proteins based on proteomic analysis of NB, GN, GNB, and AG tumor tissues To identify neuroblastoma (NB)-specific biological pathways distinct from other pathological categories, we compared tumor tissues from NB, GNB, and GN with those from AG using proteomic analysis. Principal Component Analysis (PCA) of protein levels in patient samples revealed clear distinctions between the groups and similar protein profiles within each group (Fig. A). A total of 8525 proteins were identified and quantified (Fig. B). Differentially expressed genes (DEGs) were defined as |Fold Change|≥ 2 and P < 0.05, leading to the identification of 872 upregulated and 1168 downregulated proteins in NB/AG, 2945 upregulated and 124 downregulated proteins in GN/AG, and 1258 upregulated and 688 downregulated proteins in GNB/AG (Fig. C). Notably, cell cycle and DNA replication pathways were highly upregulated in NB/AG, while oxidative phosphorylation, pyruvate metabolism, and the TCA cycle were significantly downregulated compared to GNB/AG and GN/AG (Fig. D). Identification of clinical prognosis-related gene modules based on histopathology of NB NB is classified into FH, and UH phenotypes based on its pathological histology, which are closely associated with NB outcomes. Compared to survival analysis, pathological histology evaluation is more precise and can reflect the severity of NB in real-time. To construct a more precise prognostic gene model for NB, we compared the proteome profiles of NB, GN, GNB, NB + GNB, NB + GNB + GN with adrenal glands (AG) respectively. A total of 3,313 differentially regulated proteins were identified. Using WGCNA, these differentially expressed proteins were clustered into 6 gene modules (Fig. A, B, E). Correlation analysis with histopathology classification revealed that high expression of the yellow module was positively associated with the UH subtype and high MKI, while the black module was negatively associated with the UH subtype and high MKI, and the red module was negatively associated with the high MKI phenotype (Fig. D, P < 0.05). Therefore, the yellow module is a key upregulated gene expression module associated with poor NB outcomes, enriched in functional pathways such as chromatin binding and mRNA metabolic processes (Fig. F). PPI network analysis identified 15 hub genes in the yellow module: SMARCA4, SMARCA5, SMARCC2, SMARCC1, PBRM1, BRD3, ARID1A, BRD2, ARID1B, KDM1A, TP53BP1, ALYREF, CBX1, SF3B1, and ADNP, which is mainly associated with chromatin remodeling (Fig. C). Verification in the TARGET NB database, which includes transcriptomic data and clinical survival information from 159 NB patients, confirmed that high expression of ALYREF and SMARCA4 was associated with high mortality risk, consistent with our proteomic results (Fig. A-B). Additionally, experimental validation in another NB cohort showed that RNA expression of TP53BP1, ALYREF, SMARCC2, and SMARCA4 was also upregulated in NB tumor tissues compared to GN tissues (Fig. C) and protein expression of ALYREF and SMARCA4 were much higher in NB tumor tissues compared to GN tissues (Fig. D, E, Figs. S1, S2 in the Additional file ).. Furthermore, it is also demonstrated that the protein expression level of ALYREF and SMARCA4 in tumor samples are correlated with their histopathological phenotypes (Table S3 in the Additional file ), e.g. INPC (International Neuroblastoma Pathology Classification) and INRG (International Neuroblastoma Risk Group) stage. Especially, SMARCA4 is strongly associated with INPC phenotype (Spearman’s ρ correlation coefficient is 0.803, p = 0.001). Therefore, ALYREF and SMARCA4 are potential prognostic biomarkers for NB patients with poor outcomes. Furthermore, high expression of these genes is associated with low immune infiltration (Fig. F). Computational screening of drug repositioning candidates for NB To discover effective and safe therapeutic drugs against NB, we conducted a drug repurposing search based on the proximity of our core prognostic gene module profile to data in L1000FWD and Cmap (Figs. F, and Fig. S4). We queried the opposed expression profile of the core prognostic gene set in these databases. In L1000FWD, we identified 50 drugs with a similarity score > 0.1, targeting dopamine receptors, cyclooxygenase, topoisomerase, calcium channels, and histamine receptors. In CMap, we identified 310 potential anti-NB drugs with tau > 90, targeting HDAC, mTOR, mitochondrial oxidative phosphorylation, PI3K, and other pathways. There were 31 overlapping potential drugs in both databases. From these, we selected the top 5 drugs for further analysis: PI-828, mocetinostat, clofarabine, ethacrynic acid, and mafenide. These drugs target HDAC, PI3K, ribonucleotide reductase V, GST, and bacterial pathways. Experimental validation of repurposing drugs in neuroblastoma cells To verify the anti-tumor activities of the predicted drugs, we tested the top 5 overlapping drugs from L1000FWD and CMap in neuroblastoma cell lines SH-N-AS and SH-SY-5Y.. Among them, only mocetinostat and clofarabine significantly increased cell death of both NB cell lines (Fig. A–D). Meanwhile, mocetinostat and clofarabine have been demonstrated to induce apoptosis of NB cell line (SH-SY-5Y) significantly, which indicate that these two drugs may induce cell death of NB cells by the way of apoptosis (Fig. E, F). Furthermore, treatment with mocetinostat and clofarabine suppressed the expression of SMARCA4, suggesting that these two drugs may target our core prognostic gene SMARCA4 and affect yellow module expression (Fig. G, Fig. S3 in Additional file ). These two drugs had not been previously reported as anti-NB agents, warranting further validation. To identify neuroblastoma (NB)-specific biological pathways distinct from other pathological categories, we compared tumor tissues from NB, GNB, and GN with those from AG using proteomic analysis. Principal Component Analysis (PCA) of protein levels in patient samples revealed clear distinctions between the groups and similar protein profiles within each group (Fig. A). A total of 8525 proteins were identified and quantified (Fig. B). Differentially expressed genes (DEGs) were defined as |Fold Change|≥ 2 and P < 0.05, leading to the identification of 872 upregulated and 1168 downregulated proteins in NB/AG, 2945 upregulated and 124 downregulated proteins in GN/AG, and 1258 upregulated and 688 downregulated proteins in GNB/AG (Fig. C). Notably, cell cycle and DNA replication pathways were highly upregulated in NB/AG, while oxidative phosphorylation, pyruvate metabolism, and the TCA cycle were significantly downregulated compared to GNB/AG and GN/AG (Fig. D). NB is classified into FH, and UH phenotypes based on its pathological histology, which are closely associated with NB outcomes. Compared to survival analysis, pathological histology evaluation is more precise and can reflect the severity of NB in real-time. To construct a more precise prognostic gene model for NB, we compared the proteome profiles of NB, GN, GNB, NB + GNB, NB + GNB + GN with adrenal glands (AG) respectively. A total of 3,313 differentially regulated proteins were identified. Using WGCNA, these differentially expressed proteins were clustered into 6 gene modules (Fig. A, B, E). Correlation analysis with histopathology classification revealed that high expression of the yellow module was positively associated with the UH subtype and high MKI, while the black module was negatively associated with the UH subtype and high MKI, and the red module was negatively associated with the high MKI phenotype (Fig. D, P < 0.05). Therefore, the yellow module is a key upregulated gene expression module associated with poor NB outcomes, enriched in functional pathways such as chromatin binding and mRNA metabolic processes (Fig. F). PPI network analysis identified 15 hub genes in the yellow module: SMARCA4, SMARCA5, SMARCC2, SMARCC1, PBRM1, BRD3, ARID1A, BRD2, ARID1B, KDM1A, TP53BP1, ALYREF, CBX1, SF3B1, and ADNP, which is mainly associated with chromatin remodeling (Fig. C). Verification in the TARGET NB database, which includes transcriptomic data and clinical survival information from 159 NB patients, confirmed that high expression of ALYREF and SMARCA4 was associated with high mortality risk, consistent with our proteomic results (Fig. A-B). Additionally, experimental validation in another NB cohort showed that RNA expression of TP53BP1, ALYREF, SMARCC2, and SMARCA4 was also upregulated in NB tumor tissues compared to GN tissues (Fig. C) and protein expression of ALYREF and SMARCA4 were much higher in NB tumor tissues compared to GN tissues (Fig. D, E, Figs. S1, S2 in the Additional file ).. Furthermore, it is also demonstrated that the protein expression level of ALYREF and SMARCA4 in tumor samples are correlated with their histopathological phenotypes (Table S3 in the Additional file ), e.g. INPC (International Neuroblastoma Pathology Classification) and INRG (International Neuroblastoma Risk Group) stage. Especially, SMARCA4 is strongly associated with INPC phenotype (Spearman’s ρ correlation coefficient is 0.803, p = 0.001). Therefore, ALYREF and SMARCA4 are potential prognostic biomarkers for NB patients with poor outcomes. Furthermore, high expression of these genes is associated with low immune infiltration (Fig. F). To discover effective and safe therapeutic drugs against NB, we conducted a drug repurposing search based on the proximity of our core prognostic gene module profile to data in L1000FWD and Cmap (Figs. F, and Fig. S4). We queried the opposed expression profile of the core prognostic gene set in these databases. In L1000FWD, we identified 50 drugs with a similarity score > 0.1, targeting dopamine receptors, cyclooxygenase, topoisomerase, calcium channels, and histamine receptors. In CMap, we identified 310 potential anti-NB drugs with tau > 90, targeting HDAC, mTOR, mitochondrial oxidative phosphorylation, PI3K, and other pathways. There were 31 overlapping potential drugs in both databases. From these, we selected the top 5 drugs for further analysis: PI-828, mocetinostat, clofarabine, ethacrynic acid, and mafenide. These drugs target HDAC, PI3K, ribonucleotide reductase V, GST, and bacterial pathways. To verify the anti-tumor activities of the predicted drugs, we tested the top 5 overlapping drugs from L1000FWD and CMap in neuroblastoma cell lines SH-N-AS and SH-SY-5Y.. Among them, only mocetinostat and clofarabine significantly increased cell death of both NB cell lines (Fig. A–D). Meanwhile, mocetinostat and clofarabine have been demonstrated to induce apoptosis of NB cell line (SH-SY-5Y) significantly, which indicate that these two drugs may induce cell death of NB cells by the way of apoptosis (Fig. E, F). Furthermore, treatment with mocetinostat and clofarabine suppressed the expression of SMARCA4, suggesting that these two drugs may target our core prognostic gene SMARCA4 and affect yellow module expression (Fig. G, Fig. S3 in Additional file ). These two drugs had not been previously reported as anti-NB agents, warranting further validation. The treatment of neuroblastoma (NB) remains a significant challenge, with a lack of highly effective prognostic biomarkers and therapeutic targets. To our knowledge, this study is the first to analyze NB tissue proteomes and correlate them with their histopathological phenotypes. Through Weighted Gene Co-expression Network Analysis (WGCNA), we identified a yellow gene module enriched in pathways related to chromatin binding and mRNA metabolism, which correlates strongly with unfavorable histology (UH) and high mitosis-karyorrhexis index (MKI) phenotypes. Among the hub genes within this module, ALYREF and SMARCA4 are associated with a high mortality risk in NB, as shown by survival analysis using the TARGET-NB dataset. RNA and protein expression levels of these genes were validated in NB and ganglioneuroma (GN) tissues, supporting ALYREF and SMARCA4 as potential prognostic biomarkers for NB. ALYREF, a nuclear protein with known roles in tumor growth and progression, is frequently expressed due to chromosome 17q21-ter gain in NB, where it stabilizes the MYCN protein . SMARCA4 encodes the transcription activator BRG1, a frequently altered ATP-dependent catalytic subunit of SWI/SNF chromatin-remodeling complexes involved in transcriptional regulation across cancer types . SMARCA4 expression is consistently upregulated in advanced NB stages, with high levels correlating with reduced event-free and overall survival . Our findings confirm that these two proteins are linked not only to a high mortality risk but also to UH phenotype in NB. Given the high heterogeneity of NB, high-risk patients often require intensive chemotherapy, yet only less than half experience a cure. Chemoresistance in NB is a pressing issue, highlighting the urgent need for effective and safe therapeutic agents. Our computational drug screening identified mocetinostat and clofarabine, two compounds previously unreported in NB, as promising therapeutic candidates based on their proximity to our core prognostic gene module profile. Mocetinostat, an isoform-selective histone deacetylase (HDAC) inhibitor, has demonstrated anticancer potential. By inhibiting HDAC, mocetinostat promotes histone acetylation, activating gene expression and impacting cancer cell proliferation, apoptosis, angiogenesis, and inflammation . Currently in clinical trials for various cancers, such as follicular lymphoma, Hodgkin’s lymphoma and acute myelogenous leukemia, mocetinostat’s diverse mechanisms make it a promising cancer therapy candidate . Clofarabine, a nucleotide analog, disrupts DNA replication and repair by inhibiting DNA polymerase and ribonucleotide reductase . It interferes with nucleotide metabolism, blocks DNA synthesis, and triggers apoptosis through mitochondrial pathway disruption . Clofarabine has shown high therapeutic efficacy in pediatric patients with refractory or relapsed acute lymphoblastic leukemia (ALL) . Although our study demonstrates clofarabine’s inhibitory effect on NB cell viability and proliferation, further exploration is required to determine its role in high-risk NB therapy. Notably, both mocetinostat and clofarabine involve chromatin remodeling and epigenetic regulation, suggesting that targeting chromatin remodeling could be an effective strategy for inhibiting NB cell proliferation and survival. In NB studies, genetic alterations in chromatin remodeling related genes have been documented . Our proteomic analysis of differentially expressed proteins in NB tissues, combined with public database insights, points to chromatin remodeling as critical therapeutic target for NB. Moving forward, we aim to investigate the specific regulatory mechanisms of these identified hub genes in NB, with the objective of identifying more direct therapeutic targets. This study identified a core prognostic gene module in neuroblastoma (NB), primarily associated with RNA processing and chromatin remodeling. SMARCA4 and ALYREF were validated as key prognostic biomarkers for NB patients. Through drug repurposing, we identified five top candidates for NB treatment, with mocetinostat and clofarabine showing efficacy in two NB cell lines. Further validation of these drugs in a murine NB model is strongly recommended, offering valuable insights for developing novel molecular targeted therapies for NB. Additional file 1. Additional file 2. |
Pharmacogenomics of Dementia: Personalizing the Treatment of Cognitive and Neuropsychiatric Symptoms | 15df2396-6ab9-467d-9e15-b873de3fd085 | 10671071 | Pharmacology[mh] | Pharmacotherapy of dementia is partially effective in only some individuals, with side effects, drug interactions, intolerance, and non-compliance occurring in the majority of dementia patients. Interindividual variability in drug response among dementia patients is largely due to genetic variations, which could influence the activity or availability of drug-metabolizing enzymes, receptors, channels, transporters, and other proteins involved in drug pharmacokinetics and pharmacodynamics . Pharmacokinetics refers to the variability in the drug’s absorption, distribution, metabolism, and elimination (ADME) that modulates the delivery or removal of drugs and their metabolites at their action sites. On the other hand, pharmacodynamics refers to variability in the drug action dependent on the interaction of the active drug with its target molecules, such as receptors, ion channels, and enzymes, and can also affect therapeutic response and drug side effects. Different studies aimed to identify genetic variants that could predict patients who may optimally benefit from specific, individually tailored treatment. Both pharmacogenetics and pharmacogenomics, as rapidly growing fields with huge potential in drug discovery and personalized medicine, address interindividual variations in DNA sequence affecting drug efficacy and toxicity in order to optimize the pharmacotherapy based on the patient individual genetic signature. Whereas pharmacogenetics generally refers to the variations in a single or several genes influencing the response to drugs, pharmacogenomics addresses genome-wide alterations and the mutual interaction of many genes affecting drug efficacy and safety. The development of pharmacogenomics as an interdisciplinary large-scale systematic approach has been reinforced by the introduction of genomic techniques, such as genotyping, gene sequencing, gene expression, genetic epidemiology, transcriptomics, proteomics, metabolomics and bioinformatics, and other multiplex assay technologies, which allow deeper assessment of disease mechanisms, potential drug targets and metabolism, or associated pathway components . However, the application of pharmacogenomics in dementia patients is very challenging since dementia is a complex disorder represented not only by cognitive decline but also by behavioral and neuropsychiatric symptoms, as well as progressive functional deterioration, in which more than 200 different genes associated with the dementia pathogenesis, drug mechanism of action, phase I and phase II metabolism reactions, transporters, and concomitant pathologies might be involved . A complex clinical picture of dementia usually requires simultaneous therapy with several different drugs, targeting both cognitive and neuropsychiatric symptoms. Specifically, patients with dementia typically receive 6–10 different drugs per day, including conventional anti-dementia drugs, antidepressants, antipsychotics, anxiolytics, anticonvulsants, and also other types of drugs (antihypertensive drugs, diuretics, statins, anti-histaminics, anti-inflammatory, and antidiabetic drugs), for treating various comorbid and somatic disorders in the elderly . Therefore, optimization of therapy in dementia patients is a major goal to which pharmacogenomics could contribute by improving patient stratification, resulting in more effective therapy and reduced drug adverse effects. In this review, we summarize current knowledge on molecular mechanisms of dementia, the most relevant associated genes, as well as genes involved in the activity or availability of drugs commonly used for the management of both cognitive and neuropsychiatric symptoms of dementia. We also discuss the importance of pharmacogenomics studies in the search for predictive strategies and new and effective medications for dementia. Dementia is a complex condition that involves the interaction of various factors such as genetics, epigenetics, metabolic and vascular health abnormalities, as well as various environmental influences, ultimately resulting in the death of brain cells. This multifaceted syndrome represents a significant healthcare problem around the world and ranks as the foremost cause of disability among the elderly. The weight of dementia extends its impact not only on individuals but also on their caregivers and healthcare systems, given the profound cognitive and functional impairments it entails. A concise medical history, as well as neurological and cognitive examinations, are necessary in order to evaluate possible dementia, with the patient’s history gathered from both the individual and a family member or friend playing a crucial role in this process. The cognitive assessment aims to determine the presence and characteristics of cognitive deficits and often utilizes screening tools such as the Montreal Cognitive Assessment (MoCA) or the Mini-Mental State Examination (MMSE), while the neurologic examination assesses neurocognitive problems (agnosia, aphasia, and apraxia) and unusual behaviors associated with specific types of dementia . The standard assessment includes blood tests and neuroimaging to identify potential causes of dementia, with specialized neuropsychological testing being required in specific instances. Advanced diagnostic tools, such as positron emission tomography (PET) scans and cerebrospinal fluid testing, can provide valuable information in atypical or diagnostically challenging cases, while genetic testing might be suitable for younger patients with a family history of dementia . There are over 100 diseases that can cause dementia, although the four main types include Alzheimer’s disease (AD) (50–75%), vascular dementia (VaD) (15–20%), Lewy body dementia (LBD) (10–15%), and frontotemporal dementia (FTD) (2%). Cognitive impairments, including dementia, are often present in other proteinopathies such as Parkinson’s disease (PD), Huntington’s disease (HD), amyotrophic lateral sclerosis (ALS), and prion diseases . Currently, dementia affects more than 55 million people worldwide, and it is estimated that over the next 20–25 years, the number of individuals at risk may exceed 153 million . The prevalence of dementia exhibits an exponential increase, starting at around 1–2% in individuals aged 60–65 years and rising to over 30–35% in people aged over 80 years . It is highly probable that among patients aged 75–80 years, most dementia cases result from a combination of degenerative and vascular factors (mixed dementia). In contrast, cases of pure AD have become less common for people aged 80 and older . AD, the most common type of dementia, is primarily characterized by the accumulation of extracellular amyloid β (Aβ) plaques and intracellular tangles of hyperphosphorylated tau protein in the brain, leading to neural degeneration and synaptic dysfunction. Rare forms of dominantly inherited early-onset AD can result from mutations in the amyloid precursor protein (APP) and presenilin (PSEN1 and PSEN2) genes, collectively accounting for less than 1% of all AD cases . Late-onset AD, which is more prevalent, is typically categorized as sporadic, although there are identified genetic risk factors, with the most important gene coding for apolipoprotein E (APOE) . Beyond the primary risk factors like age, family history, and the APOE4 genotype , late-onset AD is also influenced by additional risk factors, such as triggering receptor expressed on myeloid cells-2 (TREM2), a disintegrin and metalloproteinase 10 (ADAM10), and phospholipase D3 (PLD3), that not only impact APP and tau but also play a role in cholesterol metabolism and immune response . In addition, other risk factors associated with AD include environmental and metabolic factors, such as cerebrovascular disease, diabetes, inadequate dietary habits, stress, and head injuries . Progressive decline in memory, particularly episodic memory, as well as problems with executive functions, usually appear in the earlier stage of the disease, whereas challenges related to perceptual motor skills, social cognition, and language abilities tend to become noticeable at a later dementia stage . Additionally, non-memory-related symptoms can also be manifested, including mood disturbances, such as anxiety, depression, and apathy, which may persist throughout the course of the disease . Moreover, during the middle to later stages of dementia, individuals may exhibit behavioral symptoms like aggression, irritability, restlessness, and wandering . VaD, also known as multi-infarct dementia, arteriosclerotic dementia, or vascular cognitive impairment, is the second most common type of dementia resulting from cerebrovascular disease and leading to impaired blood flow to the brain. It can be caused by both large and small vessel diseases, and the critical factor in its development is the location of the lesions rather than the extent of tissue damage . Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) is the most prevalent hereditary stroke condition attributed to mutations in the neurogenic locus notch homolog protein 3 (Notch-3) gene and serves as a significant contributor to VaD . It is important to note that risk factors for stroke align with risk factors for VaD, given that stroke represents a significant pathway that connects cardiac and cerebrovascular diseases to vascular brain injury and, ultimately, cognitive impairment. Furthermore, age, diabetes, hypertension, and smoking are some of the other important risk factors for VaD . Although cognitive impairment varies depending on the location and extent of vascular damage, it typically involves deficits in attention, executive function, and processing speed, alongside common symptoms such as alterations in mood and personality . Depression linked to VaD, a condition known as vascular depression, may become apparent in later life, often accompanied by difficulties in executive functions . LBD, the third most common type of dementia, predominantly involves the misfolding and aggregation of α-synuclein, leading to the formation of Lewy bodies, which is a characteristic feature also observed in Parkinson’s disease (PD). This leads to cognitive deficits, which result in impaired attention, executive functions, and visuospatial abilities , accompanied by fluctuations in cognitive performance, persistent visual hallucinations, and the presence of parkinsonism . The primary difference between LBD and dementia in PD lies in the chronological order of the occurrence of cognitive and movement symptoms . While in LBD, cognitive impairment occurs before the onset of parkinsonism, in PD, cognitive problems develop after the appearance of motor symptoms . LBD can also manifest with additional features, such as increased sensitivity to specific medications and rapid eye movement (REM) sleep behavior disorder. Clinical indicators that support LBD diagnosis include loss of consciousness, frequent falls, hallucinations, delusions, and depression . The cause of LBD remains elusive, with genetics, age-related changes, and environmental factors playing a role in the etiology; however, further research is needed for its comprehensive understanding . FTD is marked by significant frontal and temporal lobe atrophy, typically containing abnormal tau or ubiquitin protein inclusions. It primarily represents a sporadic condition, although genetics play a significant role in approximately 40% of cases having a familial origin and a quarter of cases showing autosomal dominant inheritance. Key genes implicated in FTD pathogenesis include genes coding for microtubule-associated protein (MAPT), granulin (GRN), and chromosome 9 open reading frame 72 (C9ORF72) . Additionally, it is assumed that mutations in the C9ORF72 gene may serve as a link between FTD and amyotrophic lateral sclerosis (ALS), contributing to the incidence of both conditions . Furthermore, thyroid disease and head trauma have been associated with an increased risk of developing FTD . FTD includes clinical subtypes, such as behavioral and language variants, which align with specific regions of brain atrophy. In the behavioral variant, there are notable changes in behavior and personality, encompassing a lack of interest in personal responsibilities, neglect of personal hygiene, isolation from social interactions, and displays of socially inappropriate behavior . Some patients may exhibit repetitive or compulsive motor actions or develop unconventional eating habits , leading to potential misdiagnoses, such as major depressive or bipolar disorder. Apart from the behavioral variant, there are three language variants in FTD: the semantic variant, characterized by difficulties in naming and comprehending words; nonfluent aphasia, characterized by the challenges related to speech and/or grammar apraxia; and the logopenic subtype, characterized by issues with word retrieval . The primary obstacles to the effective diagnosis and treatment of dementia revolve around the absence of specific early detection markers and the limited availability of effective therapies. However, recent advancements in genomic medicine have significantly improved our understanding of the underlying causes of dementia. These breakthroughs have led to significant improvements in diagnostic accuracy via the introduction of new biomarkers. Additionally, these advancements made it possible to customize treatment approaches by incorporating pharmacogenetic and pharmacogenomic methods into both drug development and clinical practice . Genetic factors can play a role in the development of dementia through Mendelian inheritance patterns, leading to high heritability in families, or act as contributing factors in complex heterogeneous multifactorial types of dementia, usually with small effect sizes . As demonstrated in , Mendelian forms of dementia are usually rare and are characterized by mutations in disease-causing genes, and they are usually inherited through an autosomal dominant pattern . Sporadic forms of different dementias are partly explained by single nucleotide polymorphisms (SNPs), which represent the common type of genetic variation that occurs in a population, and they represent single-letter differences in the DNA sequence at a particular position in the genome and with structural variants (SV), defined as DNA segments of minimum 50 bp, that include duplications, deletions, and insertions of specific genes, as well as their inversions or translocations . Besides genomic variations, epigenetic alternations, such as DNA methylation and hidroxymethylation, histone modifications, non-coding RNA (ncRNA) regulation, and mitochondrial epigenetics, have been included in the pathogenesis of many diseases, including AD . Structural and functional genomics can help identify risk factors associated with dementia and aid in early detection, diagnosis, and drug development. With the development of high-throughput methods for the detection of genetic variants and epigenetic marks on a genome-wide scale, many genes and genomic regions have been implicated in the pathogenesis of AD and other dementias . In contrast to candidate-gene association studies, genome-wide association studies (GWAS), whole genome/exome sequencing (WGS/WES), and next-generation sequencing (NGS) provide hypothesis-free approaches to identify novel genes or genomic regions associated with the development or pathology of dementia. These methods lead to the identification of many potential risk variants, which could pinpoint novel biological pathways included in the pathogenesis of dementia . However, most variants do not exhibit a direct effect on the protein function; moreover, their individual effect on the total polygenic risk score is usually low, so targeting most of these variants might have little or no therapeutic effects. Hence, the important step is to implement the functional genomics approaches by integrating the signals obtained by GWAS with other multiomic datasets in order to identify the possible role of these variants and affected biological pathways in the pathogenesis of dementia . The application of CRISPR-Cas gene editing technology significantly enhances the feasibility of large-scale genetic screenings, allowing the usage of precise modifications of the human genome to investigate functional outcomes in human cells, including neurons, microglia, and astrocytes . Integration of functional genomics with genetic studies and single-cell profiling of patient tissues will, therefore, significantly contribute to the uncovering of the complex mechanisms underlying dementia, as well as potential therapeutic targets. 3.1. Alzheimer’s Disease The heritability of neurodegenerative dementias can vary widely between individuals and families, with some genetic overlaps indicating shared biological pathways involved in their development. For example, the overall heritability of AD is estimated to be between 60 and 80%, with significant differences between early-onset AD (EOAD), which develops under the age of 60, represents 5–10% of all AD cases, and has a heritability of 92% to 100% , and late-onset AD (LOAD), which develops after 60 years of age, represents the majority of AD cases, is more heterogeneous, and has heritability between 58 and 70% . EOAD is mainly caused by mutations in three genes: gene coding for amyloid β precursor protein ( APP ), as well as genes coding for presenilin 1 ( PSEN1 ) and presenilin 2 ( PSEN2 ), components of γ-secretase, an enzyme involved in the proteolytic cleavage of APP . These mutations, which mostly follow an autosomal dominant pattern, directly result in the overproduction, aggregation, and impaired degradation of Aβ peptides, leading to neurodegeneration; however, they do not show a clear association with LOAD . Tri-allelic polymorphism in the apolipoprotein E gene ( APOE) was the first identified susceptibility gene for LOAD and has been characterized by missense mutations resulting in the structural and functional differences of the ApoE protein . APOE*4 allele is considered the highest-risk allele with adverse effects on lipid metabolism, cardiovascular diseases, and different proteinopathies, including AD, FTD, LBD, and ALS, while APOE*2 is considered a protective allele . Besides the prevalent APOE polymorphism, which accounts for approximately 25% of genetic variation in AD, rare coding and noncoding alterations within the APOE gene have also been associated with the susceptibility to AD . With the development of GWAS, more than 100 additional risk loci for LOAD have been identified, of which 16 lead SNPs are located in the coding exons or in the 3′UTR and 5′UTR, and many others harboring them, as reviewed in . Most of the associated genes are involved in the metabolism of Aβ, in the immune response (especially microglial activation), or in the lipid and endocytosis pathways . However, most of the identified variants are non-coding and do not have a direct effect on the protein function. It has been suggested that they could act as regulators of gene expression by altering DNA methylation and affecting the binding of transcriptional factors . Although links between SVs and AD were not distinctively found in GWAS , many variants showed association with glucuronosyltransferase activity, neuron projection, histone modifications, gene expression, RNA splicing, or protein abundance in post-mortem AD brains, thus providing valuable material for studying their function . For instance, several GWAS and functional studies identified variants in the SORL1 gene, coding for sortilin-related receptor 1, which has been involved in the modulation of Aβ peptide production in the brain, to be associated with the risk of LOAD, and possibly familial EOAD . Moreover, SNPs and gene duplication within complement receptor gene 1 ( CR1 ), highly expressed in astrocytes and microglia, are considered one of the most important risk variants in AD in several GWAS and are shown to significantly affect the Aβ accumulation in the brain . In addition, mutations in the ADAM10 gene, which codes for the component of α-secretase, were shown to attenuate its activity, resulting in the accumulation of Aβ plaques and reactive gliosis in transgenic mice . Significantly altered micro RNA associated with sirtuin 1 ( SIRT1 ), β-secretase 1 ( BACE1 ), and α-secretase ( ADAM10 ) transcripts have also been reported, suggesting potential epigenetic regulation of expression of genes associated with APP metabolism . BIN1 is the second most important AD susceptibility gene after APOE , which encodes for bridging integrator 1 protein, involved in endocytosis, intracellular trafficking, and synaptic plasticity . AD-associated BIN1 variants are non-coding, but they could act as gene expression modulators by facilitating the binding of transcriptional factor MEF2C in primary microglia and induced pluripotent stem cell-derived macrophages . Moreover, variants in the triggering receptor expressed on the myeloid cells 2 ( TREM2) gene have been associated with an increased risk of AD and other neurodegenerative disorders . Functional studies have shown that TREM2 plays a crucial role in regulating microglial activity and that innate immune response may be involved in the Aβ clearance and regulation of tau pathology . Salivary α-amylase AMY1A is an enzyme that degrades polysaccharides such as glycogen and could be responsible for glycogen degradation in astrocytes and neurons that is necessary for neurotransmitter production and memory formation . The high copy number of the AMY1A gene possibly leads to higher production of brain α-amylase, which showed an association with lower AD risk and more preserved episodic memory . In addition, a strong association of SNPs, variable number of tandem repeats (VNTRs), and variants generating premature termination codon in the ABCA7 gene with AD were observed in several GWAS and genetic studies . The ABCA7 gene codes for the ATP-binding cassette (ABC) transporter involved in lipid homeostasis, cholesterol metabolism, and phagocytosis . Several SNPs and retrotransposon Alu insertion into the intron of the TOMM40 gene , which is adjacent to and usually in haplotype with APOE locus, have also been associated with AD . TOM40 protein is crucial for mitochondrial function, including cell metabolism, apoptosis, and lipid synthesis , whereas poly T extension in introne 6 of the TOMM40 gene is shown to be protective against Aβ toxicity . The targeted analysis demonstrated that duplication of SULT1A3/4 genes coding for sulfotransferases, which are involved in the metabolism of catecholamines, are associated with the risk of AD and earlier onset of the disease ; however, this result was not replicated in the GWAS. 3.2. Vascular Dementia The genetic background of VaD is poorly understood, as it is considered a mostly sporadic disease . However, there is supporting evidence that VaD can develop due to single gene mutations, resulting in the development of monogenic disorders, such as CADASIL, which is considered the most common heritable form of VaD . CADASIL is caused by mutations in the NOTCH3 gene, coding for the Notch 3 receptor, which results in impaired function of vascular smooth muscle cells . Much less frequent is cerebral autosomal recessive arteriopathy with subcortical infarcts and leukoencephalopathy (CARASIL), which is developed due to various mutations in the HTRA1 gene, coding for HTRA1 serine peptidase/protease 1 . Other disorders include Fabry disease (FD), an X-linked lysosomal disease caused by a mutation of the GLA gene, resulting in impaired α-galactosidase activity and accumulation of glycosphingolipids ; retinal vasculopathy with cerebral leukodystrophy (RVCL) due to frame-shift TREX1 mutations that result in a DNase III exonuclease impairment ; cerebral amyloid angiopathy (CAA), characterized by defective protein deposits, including Aβ and highly affected by mutations in APP , PSEN1, and PSEN2 genes but also in transthyretin ( TTR ), cystatin C ( CST3 ), gelsolin ( GSN ), and integral membrane protein 2B ( ITM2B ) genes ; and disorders related to mutations in collagen type IV α1 chain ( COL4A1 ) gene, such as small vessel arteriopathy and cerebral small vessel disease (CSVD) . The genetic basis of VaD sporadic forms is mostly based on candidate-gene studies, and it overlaps with the genetic background of the vascular risk factors such as hypertension, dyslipidemia, and smoking, as well as of other diseases, such as AD and stroke . APOE*E4 allele was associated with a higher risk of VaD in several meta-analyses, irrespective of patient ethnicity . Genetic variants in the methylenetetrahydrofolate reductase ( MTHFR ) gene, which affects the level of homocysteine ; polymorphisms in the paraoxonase 1 ( PON1 ) gene ; insertion–deletion variant in the intron 16 of ACE gene, coding for the angiotensin-converting enzyme, associated with vascular reactivity, have also been implicated in the moderating the risk of development of sporadic VaD . SNPs in the genes related to the inflammation, such as interleukin ( IL-1α , IL-1β , IL-6 ), and tumour necrosis factor (TNF-α , TGF-β1) genes, could also possibly influence VaD development; however, these findings were not replicated in all ethnic groups . GWAS also detected associations of VaD with polymorphisms in the androgen receptor ( AR ) gene on the X-chromosome and RPGRIP1L gene, whose product regulates thromboxane A2 and consequently vasoconstriction and platelet aggregation , while functional studies confirmed the association of spleen associated tyrosine kinase ( SYK ) and pleckstrin homology like domain family B member 2 ( PHLDB2 ) genes with VaD . The challenge in determining the genetic basis of sporadic VaD is due to the small effects of many genetic variants, as well as the heterogeneity of VaD phenotypes. Therefore, it is necessary to confirm these findings in large replication cohorts and to further explore the biological mechanisms involved in both AD and stroke . 3.3. Frontotemporal Dementia The prevalence of familial FTD represents 30% of the total FTD cases . It develops mostly due to autosomal dominant mutations in chromosome 9 open reading frame 72 ( C9ORF72 ), microtubule-associated protein tau ( MAPT) , and progranulin (GRN) genes, which are responsible for 60% of familial FTD cases . The sporadic form, which represents 70% of FTD cases, is more complex, and its heritability ranges from 26 to 31% and mostly includes SNPs . Pathogenic expansion of GGGGCC hexanucleotide repeats in the intron region of the C9ORF72 gene is the most common genetic cause of FTD and ALS and a rare cause of PD. It accounts for the 20–30% genetic susceptibility of familial and about 6% of sporadic FTD . The exact function of the protein encoded by the C9ORF72 gene is not well known, but it appears to be involved in the regulation of autophagy and inflammation . Mutations in the C9ORF72 gene can lead to both loss-of-function and gain-of-function effects by forming RNA foci in the nucleus, which can be translated into dipeptide repeat proteins and TAR DNA binding protein (TDP-43) inclusions in neurons and oligodendrocytes . Complex inversion of the 673 bp region in the MAPT gene (H2 haplotype) has been associated with FTD/ALS but also with AD and LBD risk. Mutations in exonic and intronic regions of the MAPT gene primarily affect the mRNA splicing, which can lead to disruption of the tau protein structure, resulting in impaired microtubule assembly and aggregation of tau filaments . In addition, complex inversion 673 bp region of MAPT H2 haplotype can reduce the risk of FTD/ALS but also AD, LBD, and PD , while several identified deleterious SVs encompassing the MAPT gene region and H1/H2 haplotype could be implicated in the gene expression . GRN gene mutations are mostly non-sense and deleterious mutations, which generate a premature termination codon that leads to reduced expression of progranulin and, consequently, in lysosomal impairment and accumulation of pathological forms of ubiquitinated TDP-43, characteristic for some types of FTD and ALS . In addition, more rare mutations were associated with FTD with cumulative risk <5%, of which the strongest effect was loss-of-function mutations in tank-binding kinase ( TBK1) gene, coding for serine/threonine kinase, which are estimated as the fourth and second most common genetic cause of FTD and ALS, respectively . TBK1 mutations result in a dysfunctional vesicular transport system, which could lead to deregulated autophagy and neurodegeneration . Other associated genes are mostly involved in the regulation of transcription and RNA splicing, protein degradation, membrane fusion, autophagy, and apoptosis and include genes coding for valosin-containing protein ( VCP ), optineurin ( OPTN ), TAR DNA binding protein ( TARDP ), charged multivesicular body protein 2B ( CHMP2B ), triggering receptor expressed on myeloid cells 2 ( TREM2 ), ubiquilin 2 ( UBQLN2 ), sequestosome 1 ( SQSTM1 ), fused in sarcoma ( FUS ), coiled-coil-helix-coiled-coil-helix domain containing 10 ( CHCHD10) , sigma non-opioid intracellular receptor 1 (SIGMAR1) , cyclin F ( CCNF), and TIA1 cytotoxic granule associated RNA binding protein ( TIA1) . Additional high-risk loci containing common genetic variants (SNPs) were identified and replicated in the recent study , such as several variants located in the introns of LOC730100 gene, coding for long ncRNA, which upregulation has been shown to enhance proliferation and invasion of glioma cells ; CEP131 gene, coding for centrosomal complex involved in the stabilization of genome ; ENTHD gene 2 , involved in the trans-Golgi network vesicular processes ; and C17orf89 gene . 3.4. Lewy Body Dementia The majority of LBD are sporadic cases (>80%), and genetic influence on its development was previously considered small; however, it is now clear that the genetic component of LBD is estimated to be 36–59.9%, based on SNPs only . Moreover, there is increasing evidence of hereditary components in the development of LBD, which is also found in related dementia, such as AD- and PD-associated dementia . Not only does LBD share similar clinical and neuropathological features with PD and in a subset of AD cases, but also similar genetic factors have been implicated in the development of these diseases, suggesting similar molecular pathways underlying their pathogenesis . However, recent findings have shown genetic variants specific to LBD . Well-established risk genes for LBD include the APOE gene, also associated with AD, as well as α-synuclein ( SNCA ) and β-glucosylceramidase ( GBA ), which also represent risk genes for PD . APOE risk alleles have been implicated in the pathology of AD and LBD but not PD , which could explain the presence of AD-related neuropathological hallmarks in numerous LBD cases . Point mutations in the SNCA gene are possibly affecting the membrane binding activity and synuclein aggregation, while locus multiplications of SNCA , leading to the overproduction of synuclein, can result in the formation of Lewy bodies . Besides potential disease-causing mutations, there are several SNPs in the SNCA locus that could modulate the risk of developing LBD and PD, with differential prevalence between these diseases . Moreover, SNCA gene methylation was suggested to be significantly decreased in LBD, leading to higher gene expression . Mutations in the GBA gene, which codes for lysosomal enzyme β-glucocerebrosidase, lead to reduced enzyme activity, resulting in impaired degradation of α-synuclein and its accumulation , and are linked with the higher risk of PD, with variations associated with earlier onset and shorter life-span in PD and LBD . The latest GWAS identified 13 genomic risk loci significantly associated with LBD, contributing to 6.24% of total LBD heritability . They include variations in BIN1 gene (also associated with AD); transmembrane protein 175 and lysosomal K + channel TMEM175 gene (implicated in PD), which deficiency leads to decreased lysosomal catalytic activity due to pH imbalance ; CLU gene, coding for clusterin, a protein that possibly binds α-synuclei aggregated species ; FBXL19 gene, which encodes for the type of ubiquitin ligases involved in the regulation of ubiquitination and degradation of inflammatory cytokines with potential neuroprotective effect ; and the MAPT gene, which is also involved in the pathogenesis of FTD and AD . Functional enrichment analysis showed that many variants associated with LBD were found in regions associated with the regulation of gene transcription and translation, such as exone regions, enhancers, and regions linked to histone modifications, especially H3K36me3 . A common structural variant (309 bp deletion) in the intron region of the two-pore calcium channel ( TPCN1) gene that encodes a two-pore calcium channel has been associated with the risk of LBD and AD . The functional implications of this gene were confirmed in Tpcn1 knockout mice, who have shown impaired memory and spatial learning . Moreover, deletion in the OPTN gene was associated with an increased risk for LBD . Accumulation of optineurin in Lewy bodies and previous involvement of OPTN mutation in the development of FTD confirm the importance of this gene in the pathogenesis of neurodegenerative dementias. These results showing genetic overlap and potentially shared biological mechanisms involved in AD, FTD, PD, and LBD could provide insight into both the prevention and treatment of these diseases. The heritability of neurodegenerative dementias can vary widely between individuals and families, with some genetic overlaps indicating shared biological pathways involved in their development. For example, the overall heritability of AD is estimated to be between 60 and 80%, with significant differences between early-onset AD (EOAD), which develops under the age of 60, represents 5–10% of all AD cases, and has a heritability of 92% to 100% , and late-onset AD (LOAD), which develops after 60 years of age, represents the majority of AD cases, is more heterogeneous, and has heritability between 58 and 70% . EOAD is mainly caused by mutations in three genes: gene coding for amyloid β precursor protein ( APP ), as well as genes coding for presenilin 1 ( PSEN1 ) and presenilin 2 ( PSEN2 ), components of γ-secretase, an enzyme involved in the proteolytic cleavage of APP . These mutations, which mostly follow an autosomal dominant pattern, directly result in the overproduction, aggregation, and impaired degradation of Aβ peptides, leading to neurodegeneration; however, they do not show a clear association with LOAD . Tri-allelic polymorphism in the apolipoprotein E gene ( APOE) was the first identified susceptibility gene for LOAD and has been characterized by missense mutations resulting in the structural and functional differences of the ApoE protein . APOE*4 allele is considered the highest-risk allele with adverse effects on lipid metabolism, cardiovascular diseases, and different proteinopathies, including AD, FTD, LBD, and ALS, while APOE*2 is considered a protective allele . Besides the prevalent APOE polymorphism, which accounts for approximately 25% of genetic variation in AD, rare coding and noncoding alterations within the APOE gene have also been associated with the susceptibility to AD . With the development of GWAS, more than 100 additional risk loci for LOAD have been identified, of which 16 lead SNPs are located in the coding exons or in the 3′UTR and 5′UTR, and many others harboring them, as reviewed in . Most of the associated genes are involved in the metabolism of Aβ, in the immune response (especially microglial activation), or in the lipid and endocytosis pathways . However, most of the identified variants are non-coding and do not have a direct effect on the protein function. It has been suggested that they could act as regulators of gene expression by altering DNA methylation and affecting the binding of transcriptional factors . Although links between SVs and AD were not distinctively found in GWAS , many variants showed association with glucuronosyltransferase activity, neuron projection, histone modifications, gene expression, RNA splicing, or protein abundance in post-mortem AD brains, thus providing valuable material for studying their function . For instance, several GWAS and functional studies identified variants in the SORL1 gene, coding for sortilin-related receptor 1, which has been involved in the modulation of Aβ peptide production in the brain, to be associated with the risk of LOAD, and possibly familial EOAD . Moreover, SNPs and gene duplication within complement receptor gene 1 ( CR1 ), highly expressed in astrocytes and microglia, are considered one of the most important risk variants in AD in several GWAS and are shown to significantly affect the Aβ accumulation in the brain . In addition, mutations in the ADAM10 gene, which codes for the component of α-secretase, were shown to attenuate its activity, resulting in the accumulation of Aβ plaques and reactive gliosis in transgenic mice . Significantly altered micro RNA associated with sirtuin 1 ( SIRT1 ), β-secretase 1 ( BACE1 ), and α-secretase ( ADAM10 ) transcripts have also been reported, suggesting potential epigenetic regulation of expression of genes associated with APP metabolism . BIN1 is the second most important AD susceptibility gene after APOE , which encodes for bridging integrator 1 protein, involved in endocytosis, intracellular trafficking, and synaptic plasticity . AD-associated BIN1 variants are non-coding, but they could act as gene expression modulators by facilitating the binding of transcriptional factor MEF2C in primary microglia and induced pluripotent stem cell-derived macrophages . Moreover, variants in the triggering receptor expressed on the myeloid cells 2 ( TREM2) gene have been associated with an increased risk of AD and other neurodegenerative disorders . Functional studies have shown that TREM2 plays a crucial role in regulating microglial activity and that innate immune response may be involved in the Aβ clearance and regulation of tau pathology . Salivary α-amylase AMY1A is an enzyme that degrades polysaccharides such as glycogen and could be responsible for glycogen degradation in astrocytes and neurons that is necessary for neurotransmitter production and memory formation . The high copy number of the AMY1A gene possibly leads to higher production of brain α-amylase, which showed an association with lower AD risk and more preserved episodic memory . In addition, a strong association of SNPs, variable number of tandem repeats (VNTRs), and variants generating premature termination codon in the ABCA7 gene with AD were observed in several GWAS and genetic studies . The ABCA7 gene codes for the ATP-binding cassette (ABC) transporter involved in lipid homeostasis, cholesterol metabolism, and phagocytosis . Several SNPs and retrotransposon Alu insertion into the intron of the TOMM40 gene , which is adjacent to and usually in haplotype with APOE locus, have also been associated with AD . TOM40 protein is crucial for mitochondrial function, including cell metabolism, apoptosis, and lipid synthesis , whereas poly T extension in introne 6 of the TOMM40 gene is shown to be protective against Aβ toxicity . The targeted analysis demonstrated that duplication of SULT1A3/4 genes coding for sulfotransferases, which are involved in the metabolism of catecholamines, are associated with the risk of AD and earlier onset of the disease ; however, this result was not replicated in the GWAS. The genetic background of VaD is poorly understood, as it is considered a mostly sporadic disease . However, there is supporting evidence that VaD can develop due to single gene mutations, resulting in the development of monogenic disorders, such as CADASIL, which is considered the most common heritable form of VaD . CADASIL is caused by mutations in the NOTCH3 gene, coding for the Notch 3 receptor, which results in impaired function of vascular smooth muscle cells . Much less frequent is cerebral autosomal recessive arteriopathy with subcortical infarcts and leukoencephalopathy (CARASIL), which is developed due to various mutations in the HTRA1 gene, coding for HTRA1 serine peptidase/protease 1 . Other disorders include Fabry disease (FD), an X-linked lysosomal disease caused by a mutation of the GLA gene, resulting in impaired α-galactosidase activity and accumulation of glycosphingolipids ; retinal vasculopathy with cerebral leukodystrophy (RVCL) due to frame-shift TREX1 mutations that result in a DNase III exonuclease impairment ; cerebral amyloid angiopathy (CAA), characterized by defective protein deposits, including Aβ and highly affected by mutations in APP , PSEN1, and PSEN2 genes but also in transthyretin ( TTR ), cystatin C ( CST3 ), gelsolin ( GSN ), and integral membrane protein 2B ( ITM2B ) genes ; and disorders related to mutations in collagen type IV α1 chain ( COL4A1 ) gene, such as small vessel arteriopathy and cerebral small vessel disease (CSVD) . The genetic basis of VaD sporadic forms is mostly based on candidate-gene studies, and it overlaps with the genetic background of the vascular risk factors such as hypertension, dyslipidemia, and smoking, as well as of other diseases, such as AD and stroke . APOE*E4 allele was associated with a higher risk of VaD in several meta-analyses, irrespective of patient ethnicity . Genetic variants in the methylenetetrahydrofolate reductase ( MTHFR ) gene, which affects the level of homocysteine ; polymorphisms in the paraoxonase 1 ( PON1 ) gene ; insertion–deletion variant in the intron 16 of ACE gene, coding for the angiotensin-converting enzyme, associated with vascular reactivity, have also been implicated in the moderating the risk of development of sporadic VaD . SNPs in the genes related to the inflammation, such as interleukin ( IL-1α , IL-1β , IL-6 ), and tumour necrosis factor (TNF-α , TGF-β1) genes, could also possibly influence VaD development; however, these findings were not replicated in all ethnic groups . GWAS also detected associations of VaD with polymorphisms in the androgen receptor ( AR ) gene on the X-chromosome and RPGRIP1L gene, whose product regulates thromboxane A2 and consequently vasoconstriction and platelet aggregation , while functional studies confirmed the association of spleen associated tyrosine kinase ( SYK ) and pleckstrin homology like domain family B member 2 ( PHLDB2 ) genes with VaD . The challenge in determining the genetic basis of sporadic VaD is due to the small effects of many genetic variants, as well as the heterogeneity of VaD phenotypes. Therefore, it is necessary to confirm these findings in large replication cohorts and to further explore the biological mechanisms involved in both AD and stroke . The prevalence of familial FTD represents 30% of the total FTD cases . It develops mostly due to autosomal dominant mutations in chromosome 9 open reading frame 72 ( C9ORF72 ), microtubule-associated protein tau ( MAPT) , and progranulin (GRN) genes, which are responsible for 60% of familial FTD cases . The sporadic form, which represents 70% of FTD cases, is more complex, and its heritability ranges from 26 to 31% and mostly includes SNPs . Pathogenic expansion of GGGGCC hexanucleotide repeats in the intron region of the C9ORF72 gene is the most common genetic cause of FTD and ALS and a rare cause of PD. It accounts for the 20–30% genetic susceptibility of familial and about 6% of sporadic FTD . The exact function of the protein encoded by the C9ORF72 gene is not well known, but it appears to be involved in the regulation of autophagy and inflammation . Mutations in the C9ORF72 gene can lead to both loss-of-function and gain-of-function effects by forming RNA foci in the nucleus, which can be translated into dipeptide repeat proteins and TAR DNA binding protein (TDP-43) inclusions in neurons and oligodendrocytes . Complex inversion of the 673 bp region in the MAPT gene (H2 haplotype) has been associated with FTD/ALS but also with AD and LBD risk. Mutations in exonic and intronic regions of the MAPT gene primarily affect the mRNA splicing, which can lead to disruption of the tau protein structure, resulting in impaired microtubule assembly and aggregation of tau filaments . In addition, complex inversion 673 bp region of MAPT H2 haplotype can reduce the risk of FTD/ALS but also AD, LBD, and PD , while several identified deleterious SVs encompassing the MAPT gene region and H1/H2 haplotype could be implicated in the gene expression . GRN gene mutations are mostly non-sense and deleterious mutations, which generate a premature termination codon that leads to reduced expression of progranulin and, consequently, in lysosomal impairment and accumulation of pathological forms of ubiquitinated TDP-43, characteristic for some types of FTD and ALS . In addition, more rare mutations were associated with FTD with cumulative risk <5%, of which the strongest effect was loss-of-function mutations in tank-binding kinase ( TBK1) gene, coding for serine/threonine kinase, which are estimated as the fourth and second most common genetic cause of FTD and ALS, respectively . TBK1 mutations result in a dysfunctional vesicular transport system, which could lead to deregulated autophagy and neurodegeneration . Other associated genes are mostly involved in the regulation of transcription and RNA splicing, protein degradation, membrane fusion, autophagy, and apoptosis and include genes coding for valosin-containing protein ( VCP ), optineurin ( OPTN ), TAR DNA binding protein ( TARDP ), charged multivesicular body protein 2B ( CHMP2B ), triggering receptor expressed on myeloid cells 2 ( TREM2 ), ubiquilin 2 ( UBQLN2 ), sequestosome 1 ( SQSTM1 ), fused in sarcoma ( FUS ), coiled-coil-helix-coiled-coil-helix domain containing 10 ( CHCHD10) , sigma non-opioid intracellular receptor 1 (SIGMAR1) , cyclin F ( CCNF), and TIA1 cytotoxic granule associated RNA binding protein ( TIA1) . Additional high-risk loci containing common genetic variants (SNPs) were identified and replicated in the recent study , such as several variants located in the introns of LOC730100 gene, coding for long ncRNA, which upregulation has been shown to enhance proliferation and invasion of glioma cells ; CEP131 gene, coding for centrosomal complex involved in the stabilization of genome ; ENTHD gene 2 , involved in the trans-Golgi network vesicular processes ; and C17orf89 gene . The majority of LBD are sporadic cases (>80%), and genetic influence on its development was previously considered small; however, it is now clear that the genetic component of LBD is estimated to be 36–59.9%, based on SNPs only . Moreover, there is increasing evidence of hereditary components in the development of LBD, which is also found in related dementia, such as AD- and PD-associated dementia . Not only does LBD share similar clinical and neuropathological features with PD and in a subset of AD cases, but also similar genetic factors have been implicated in the development of these diseases, suggesting similar molecular pathways underlying their pathogenesis . However, recent findings have shown genetic variants specific to LBD . Well-established risk genes for LBD include the APOE gene, also associated with AD, as well as α-synuclein ( SNCA ) and β-glucosylceramidase ( GBA ), which also represent risk genes for PD . APOE risk alleles have been implicated in the pathology of AD and LBD but not PD , which could explain the presence of AD-related neuropathological hallmarks in numerous LBD cases . Point mutations in the SNCA gene are possibly affecting the membrane binding activity and synuclein aggregation, while locus multiplications of SNCA , leading to the overproduction of synuclein, can result in the formation of Lewy bodies . Besides potential disease-causing mutations, there are several SNPs in the SNCA locus that could modulate the risk of developing LBD and PD, with differential prevalence between these diseases . Moreover, SNCA gene methylation was suggested to be significantly decreased in LBD, leading to higher gene expression . Mutations in the GBA gene, which codes for lysosomal enzyme β-glucocerebrosidase, lead to reduced enzyme activity, resulting in impaired degradation of α-synuclein and its accumulation , and are linked with the higher risk of PD, with variations associated with earlier onset and shorter life-span in PD and LBD . The latest GWAS identified 13 genomic risk loci significantly associated with LBD, contributing to 6.24% of total LBD heritability . They include variations in BIN1 gene (also associated with AD); transmembrane protein 175 and lysosomal K + channel TMEM175 gene (implicated in PD), which deficiency leads to decreased lysosomal catalytic activity due to pH imbalance ; CLU gene, coding for clusterin, a protein that possibly binds α-synuclei aggregated species ; FBXL19 gene, which encodes for the type of ubiquitin ligases involved in the regulation of ubiquitination and degradation of inflammatory cytokines with potential neuroprotective effect ; and the MAPT gene, which is also involved in the pathogenesis of FTD and AD . Functional enrichment analysis showed that many variants associated with LBD were found in regions associated with the regulation of gene transcription and translation, such as exone regions, enhancers, and regions linked to histone modifications, especially H3K36me3 . A common structural variant (309 bp deletion) in the intron region of the two-pore calcium channel ( TPCN1) gene that encodes a two-pore calcium channel has been associated with the risk of LBD and AD . The functional implications of this gene were confirmed in Tpcn1 knockout mice, who have shown impaired memory and spatial learning . Moreover, deletion in the OPTN gene was associated with an increased risk for LBD . Accumulation of optineurin in Lewy bodies and previous involvement of OPTN mutation in the development of FTD confirm the importance of this gene in the pathogenesis of neurodegenerative dementias. These results showing genetic overlap and potentially shared biological mechanisms involved in AD, FTD, PD, and LBD could provide insight into both the prevention and treatment of these diseases. One of the primary goals in treating various forms of dementia is to decrease cognitive, behavioral, and psychological symptoms while also attempting to slow the progression of the disease. Pharmacotherapy is frequently one of the initial strategies employed to address symptoms or hinder the progression of disease, with a primary focus on targeting the impairment of cholinergic and glutamatergic systems . At present, the Food and Drug Administration (FDA) has approved two classes of pharmacological medications for managing the cognitive symptoms of AD: acetylcholinesterase (AChE) inhibitors and N-Methyl-D-Aspartate (NMDA) receptor antagonists. However, these medications are not effective in slowing down the progression of the disease itself but only provide relief from cognitive symptoms without altering the course of the underlying disease . Both AChE-selective inhibitors, donepezil and galantamine, and dual AChE and butyrylcholinesterase (BuChE) inhibitor, rivastigmine, promote the increase in AChE levels in the synaptic cleft . AChEIs prevent the breakdown of acetylcholine by inhibiting the action of acetylcholinesterase, leading to an increase in cholinergic neurotransmission . Donepezil, rivastigmine, and galantamine are currently approved for treating mild to moderate symptoms of AD and have shown modest positive effects on cognitive symptoms . Since their introduction into clinical practice, these drugs have remained the standard approach to the symptomatic treatment of AD. Various systemic reviews concluded that AChEI treatment of dementia patients shows small but significant improvement in cognitive function . A slow dose titration of these drugs is recommended to reach the optimal dose with minimal adverse effects . Even with a gradual titration process, these medications can still lead to gastrointestinal and neurological issues, including symptoms like nausea, vomiting, diarrhea, abdominal pain, dizziness, weight loss, tremor, and fatigue . In such cases, the medication dosage may need to be reduced, or an alternative drug can be considered . In addition to AD, cholinergic deficiencies are also observed in other forms of dementia, like dementia associated with PD and LBD . While AChEIs are not officially approved for these dementia types, there is growing evidence supporting their use in alleviating neuropsychiatric symptoms of patients diagnosed with LBD and PD . However, these drugs failed to show benefits among individuals with MCI . Memantine is an NMDA receptor antagonist that reduces the impact of glutamate-induced excitotoxicity . It prevents the over-activation of glutamate receptors by slowing down the flow through the NMDA-receptor subtype of glutamate receptors . This way, memantine prevents the overactivation of the glutamatergic system, still maintaining its normal function. It is used as monotherapy to manage symptoms of moderate and severe AD. Additionally, research has demonstrated its beneficial effects in slowing down the progression of cognitive decline in individuals with AD . Memantine treatment of patients with VaD showed minimal improvement in cognitive status ; however, in patients with LBD, there were no significant effects on cognitive or behavioral symptoms . In addition, when AChEI is no longer effective, memantine is an alternative drug for patients with moderate and severe AD . Memantine is usually better tolerated than AChEI, but in some cases, it can cause headache, fatigue, and gastric pain . Moreover, the combination of donepezil and memantine has been well tolerated, with positive effects on cognition and performing daily activities . Anti-amyloid drug therapy based on monoclonal antibodies is one of the novel approaches for slowing down the progression of AD . Aducanumab is a monoclonal antibody reported to be effective in identifying Aβ aggregates and selectively binding to both oligomeric and fibrillary states rather than amyloid monomers . It was reported that aducanumab has beneficial effects in reducing Aβ plaques in patients with mild AD or MCI and was approved in 2021 by the FDA . However, its approval was met with controversy due to mixed results in clinical trials, with some experts questioning the drug’s efficacy and long-term benefits . Specifically, although some patients experienced a reduction in Aβ levels, it remains uncertain whether these results have a clinical impact on cognitive functions . Another monoclonal antibody, lecanemab, received approval from the FDA in 2023 for the treatment of people with MCI or mild dementia due to AD who have elevated Aβ levels in the brain. Various clinical trials suggested that lecanemab reduces Aβ in the brain . Other drugs targeting Aβ plaques, such as donanemab, are currently in clinical trials . However, the full extent of clinical efficacy, long-term benefits, and safety of these drugs is yet to be investigated. Due to the high complexity of neurodegenerative diseases, single-target therapy approaches have been mainly ineffective in preventing or slowing the progression of these diseases. Additionally, a significant challenge lies in the occurrence of adverse effects and the development of drug tolerance . As a result, multi-target strategies are increasingly being considered, particularly in the case of AD, and a great number of structures based on this polypharmacology concept have been proposed . The main focus of this approach involves the design of a single ligand with pleiotropic effects capable of simultaneously interacting with at least two therapeutic targets. There are three types of polypharmacological ligands, which are classified as conjugate, fused, and merged ligands . Conjugate ligands are composed of pharmacophoric structures linked by a stable or cleavable molecule, allowing them to be released and interact with multiple targets. Fused ligands have pharmacophoric structures that are joined but do not overlap, whereas merged ligands have extensive overlap in their pharmacophoric structures, resulting in smaller and more straightforward molecules . Over the past decades, multi-target therapeutic compounds targeting cholinesterase inhibition, anti-inflammatory and antiapoptotic activity, monoamine oxidase (MAO) inhibition, and neuroprotection have been investigated , particularly focusing on AChE , BuChE , β-secretase 1 (BACE-1) , cannabinoid receptor subtype 2 (CB 2 R) , serotonin (5-HT) receptors , serotonin transporter (SERT) , cyclooxygenase-2 (COX-2) , 5-lipoxygenase (5-LOX) , and nuclear factor erythroid 2-related factor 2 (Nrf2) . The majority of multi-target compounds currently undergoing investigation for AD treatment are specifically designed to moderate cholinesterase and monoamine activity; inhibit Aβ aggregation; and exert metal-chelating, anti-neuroinflammatory and antioxidant activity . The development of novel drugs is inspired by established and approved medications, like donepezil and rivastigmine , as well as by various natural bioactive derivatives, such as resveratrol, flavonoids, or curcumin . An example of a multi-targeted drug candidate is ladostigil, which functions as both AChEI and a brain-selective inhibitor of MAO-A and MAO-B . It is primarily intended for the treatment of dementia, especially AD, PD, and depression . This compound is developed by combining the carbamate rivastigmine with the N-propargyl scaffold from an anti-parkinsonian drug and the irreversible selective MAO-B inhibitor, rasagiline . It was demonstrated that ladostigil is safe and well-tolerated, but it did not show significant effectiveness in delaying the progression of dementia. However, it did display the potential to reduce the brain and hippocampus volume loss, indicating a possible impact on atrophy . Despite encouraging preclinical results, so far, no multi-targeted drug has received approval for dementia treatment. However, as research into the underlying mechanisms of the disease continues, and advances in multi-target drug discovery for AD unfold, multi-targeted ligands hold substantial promise as a potential pharmacotherapeutic strategy for dementia. Numerous research studies have reported a wide range of non-cognitive symptoms in dementia patients, including behaviors such as aggression, agitation, and psychosis, as well as issues related to eating and mood disorders . Collectively, these non-cognitive symptoms are referred to as behavioral and psychological symptoms of dementia (BPSD) . In addition to anti-dementia drugs, pharmacological treatment of BPSD comprises antidepressants, antipsychotics, benzodiazepines, and mood stabilizers . While tricyclic antidepressants and paroxetine are not recommended due to certain anticholinergic side effects, selective serotonin reuptake inhibitors (SSRIs), such as sertraline and citalopram, as well as tradozone, have shown good tolerability and effects in reducing agitation, tension, aggression, psychosis and sleep disturbances . Due to adverse effects, the administration of antipsychotics and mood stabilizers is also not recommended for BPSD therapy, with the exception of atypical antipsychotics risperidone, olanzapine, and aripiprazole, as well as valproic acid . More recently, FDA-approved two drugs for treatment of BPSD: suvorexant, an orexin receptor antagonist, approved in 2020 for the treatment of insomnia in individuals with mild to moderate AD, and brexpiprazole, atypical antipsychotic, approved in 2023 for the treatment of agitation associated with AD. Aside from pharmacological treatments, there is a recommendation to consider non-pharmacological approaches for BPSD treatment, as well as to increase the quality of life for both patients and their caregivers . The aim of non-pharmacological interventions is to enhance or, at the very least, maintain cognitive function, enabling individuals to carry out their regular daily activities while effectively managing the behavioral symptoms associated with cognitive impairment. Non-pharmacological interventions include various disciplines, each of them attempting to have a positive effect on cognition, mood, and other behavioral and psychological symptoms of dementia . Several non-pharmacological treatments have been proposed for targeting cognitive functional aspects of people with dementia. Sensory and multi-sensory stimulation includes visual, olfactory, tactile, taste, and kinaesthetic stimulation in order to reduce agitation and increase awareness . These types of stimulation include art therapy, aromatherapy, light therapy, music, and dance therapy, as well as snoezelen multi-sensory therapy. Cognitive and emotion-oriented care intervention is useful for improving cognitive, emotional, and social functioning . Commonly used treatments include reminiscence therapy, reality orientation therapy, and validation therapy . There is also behavioral management therapy that has been reported effective in suppressing or eliminating stereotypical behavior, such as wandering and incontinence . Other therapies have been applied, such as animal-assisted therapy, home adaptation therapy, and assistive technologies . These types of interventions have been found to be useful in improving outcomes and quality of life in patients with dementia . Non-pharmacological techniques have been reported to be more effective with fewer side effects when compared to pharmacotherapy with antipsychotics . There are several proposed recommendations for reducing responsive behavior, including apathy, hyperactivity, and psychosis , maintaining or improving functional capacity, and reducing comorbid emotional disorders, such as anxiety and/or depression . These symptoms are frequently observed in individuals with dementia, and while medication therapy may be necessary, it is generally recommended that non-pharmacological interventions are used as the primary approach. . Sensory stimulation, such as music and light therapy and validation therapy, has been effective in reducing these types of behavior . Interventions for improving functional capacity, which refers to cognitive function and improving well-being and daily life activity, should include cognitive stimulation, reminiscence for cognitive function, as well as exercise and light therapy for improving daily life activities . Furthermore, exercise, music therapy, reminiscence, validation therapy, and psychological treatments should also be applied to reduce symptoms of depression and anxiety . Therefore, it is important that non-pharmacological treatments become an integral part of the management of dementia symptoms and rehabilitation programs . The field of dementia care continually evolves, with new therapies regularly joining the available options for managing this condition. However, it is essential to recognize that no single method alone provides a comprehensive long-term solution for dementia management and that complementary approaches are needed in order to enhance the long-term care and quality of life for individuals with dementia. When we talk about pharmacogenomics studies related to the treatment of cognitive symptoms in dementia, most of them are focused on AChEIs, memantine, and combined treatments with these four medications . The AChEIs and memantine have different metabolic pathways. Both donepezil and galantamine are metabolized mostly by CYP3A4, CYP2D6, and CYP1A2 enzymes in the liver, while rivastigmine undergoes cholinesterase-mediated hydrolysis and its metabolism minimally relies on major cytochrome P450 isozymes . Memantine is metabolized to a minor extent by the liver and is mainly excreted unchanged by the kidneys . Around 15–20% of patients diagnosed with AD exhibit aberrant AChEI metabolism, with approximately half of them being ultra-rapid metabolizers and the other half slow metabolizers . Donepezil is the most prescribed drug for the treatment of cognitive symptoms in dementia . Different CYP2D6 variants have been studied in order to assess their influence on donepezil efficacy and safety in AD patients . These variants include rs1065852, rs1080985, CYP2D6*3 (rs35742686, 2549delA), CYP2D6*4 (rs3892097, 1846G>A), CYP2D6*6 (rs5030655, 1707delT), CYP3A4*1B (rs2740574, −392A>G), and CYP2D6*10 (rs1065852, 100C>T); however, the results are inconsistent . CYP2D6 rs1080985 (−1584C/G) is one of the most studied polymorphisms in the context of its association with the clinical efficiency of donepezil. The rs1080985 G allele defines the CYP2D6*2A variant, which was found to be potentially associated with a higher drug metabolism rate . CYP2D6 poor metabolizers were found to have a 32% slower clearance rate and a 67% slower metabolism rate of donepezil compared to ultra-rapid metabolizers . Polymorphism rs1065852 (100C>T) appears in CYP2D6*4 and CYP2D6*10 variant. The study in Han Chinese patients with AD found that the CYP2D6*10/*10 allele was associated with better efficacy and higher steady-state plasma concentration of donepezil compared to other CYP2D6 genotypes . The efficiency of donepezil has also been associated with its interaction with CYP3A4/5 . A study by Noetzli and colleagues analyzed the effect of different CYP3A gene variants on donepezil clearance in AD patients. They studied CYP3A4*1B (rs2740574), CYP3A4 (rs4646437), CYP3A4*22 (rs35599367), CYP3A5*3 (rs776746), and CYP3A7*1C (−262T > A and −270T > G) variants and found no connection with donepezil pharmacokinetic parameters . A similar result was reported by Magliulo and colleagues in Italian subjects diagnosed with AD. The study investigated CYP3A4*1B , CYP3A4*3 (rs4986910), CYP3A4*4 , CYP3A5*2 (rs28365083), CYP3A5*3 (rs776746), and CYP3A5*6 (rs10264272) and found no association between these variants and donepezil concentration in plasma samples . The lack of influence of CYP3A4 variants on donepezil efficiency was also reported in Chinese AD patients . The efficiency of donepezil could also be influenced by other genetic factors that are not directly involved in its metabolism . Some of the potential candidates are genes coding for apolipoprotein E ( APOE ), ATP-binding cassette (ABC) transporter ( ABCA1 and ABCB1 ), butyrylcholinesterase ( BCHE ), acetylcholine receptor subunit α7 ( CHRNA7 ), choline acetyltransferase ( ChAT ), estrogen receptor gene ( ESR1 ), or paraoxonase ( PON-1 ). Apolipoprotein E is known to be involved in lipoprotein metabolism and associated with a higher risk of developing AD . Several studies have suggested its association with the efficacy of donepezil treatment. Patients with AD, carriers of the high-risk APOE ε4 allele, were found to have a better response to donepezil treatment and more significant improvement of cognitive symptoms . However, there are also studies reporting opposite results or no association between APOE and treatment efficiency of donepezil . The study by Lu and colleagues suggested that the APO ε3 allele could moderate the efficiency of donepezil treatment by demonstrating better treatment response in subjects who were not APOE ε3 allele carriers . Moreover, it seems that combined APOE and CYP2D6 influence on donepezil treatment efficacy might be explained by their involvement in lipid metabolism and liver function . Two ABC transporters have also been suggested as possible modulators of donepezil efficacy, ABCB1 and ABCA1. The results of different studies reported no association between the efficacy of donepezil and different ABCB1 polymorphisms . Another interesting genetic factor in donepezil pharmacogenetics is the cholesterol transporter ABCA1. Its function is to moderate Aβ aggregation and stimulate the clearance of Aβ peptides . The study by Lu and colleagues suggested that patients who were ABCA1 rs2230806 GG genotype carriers had better responses to treatment with donepezil than AA and GA genotype carriers. The combined effect between APOE and ABCA1 genetic variants was also suggested, indicating that patients who were APOE ε3 non-carriers and ABCA1 rs2230806 GG homozygotes responded better to donepezil . Evidence supporting the role of estrogen in cognitive function has raised the question of potential association between ESR1 gene variants and the therapeutic effects of AChEIs . Two ESR1 polymorphisms, rs2234693 and rs9340799, were examined in AD patients receiving donepezil, rivastigmine, or no treatment . The authors observed a significant effect of ESR1 variants in patients treated with donepezil and reported better treatment response in women than in men . The BCHE is a member of the cholinergic enzyme family. It is mainly synthesized in the liver; however, it is also present in the central and peripheral nervous system . The most researched polymorphism of BCHE is rs1803274, also known as the K-variant, which has been associated with up to 7% reduction in enzyme hydrolytic activity in heterozygotes (AG) and 14% reduction in homozygotes (AA) . This polymorphism has been associated with poor treatment response in patients receiving donepezil . However, two other studies did not confirm this association, showing no significant relationship between the presence of K-variant or rs1355534 polymorphism and donepezil efficacy. A study by De Beaumont and colleagues demonstrated that AD patients, who are carriers of APOE ε4 and BCHE K-variant, have an earlier age of onset, accelerated cognitive decline and better response to donepezil therapy . The chAT gene encodes an enzyme, choline acetyltransferase, responsible for the biosynthesis of acetylcholine. Two genetic variants of the ChAT gene have been associated with response to AChEI treatment, rs2177370 and rs3793790 . These polymorphisms have been associated with impaired synthesis of acetylcholine. Results suggest that the CC haplotype is responsible for the decreased synthesis of acetylcholine, while carriers of the CT haplotype demonstrated a higher acetylcholine synthesis rate . The association of rs2177370 polymorphism with AChEI efficacy was also reported by Harold and colleagues , while other studies did not observe such a connection . Lee and colleagues analyzed the difference in donepezil treatment response between carriers and non-carriers of the rs3810950 (2384G>A) A allele and found that the treatment outcome, after 26 weeks of therapy, is positively influenced by the presence of A allele . Another potential candidate in the pharmacogenetics of donepezil is the CHRNA7 gene, which encodes the α7 subunit of the nicotinic acetylcholine receptor (nAChR). Polymorphisms in CHRNA7 could affect the binding of acetylcholine, which is increased due to donepezil treatment, to nAChRs. A longitudinal study in the Brazilian population demonstrated a significant association between CHRNA7 rs6494223 polymorphism (T allele) and the efficacy of donepezil . The association was present after 6 months of treatment; however, after 2 years of follow-up, the association could no longer be detected . Another study found an association between CHRNA7 rs8024987 (C→G) polymorphism and the outcome of AChEI therapy, but only in female patients . The same SNP was investigated by Clarelli and colleagues, but the results did not confirm the finding reported by Weng et al. . Two SNPs, rs885071 (T→G) and rs8024987 (C→G), were found to be in linkage disequilibrium and associated with treatment response . Arylesterase PON-1 has an important role in protecting cells from injuries caused by oxidative stress. Reduced PON-1 serum levels and activity have been associated with AD . The most studied PON-1 polymorphism is rs662 (Q192R, A>G), glutamine to arginine substitution at amino acid residue 192 . Pola and colleagues were able to associate this polymorphism with AChI treatment (donepezil and rivastigmine) response, showing a higher frequency of the R allele, which exhibits higher enzyme activity, in patients who had good response to therapy . Since PON-1 acts as an endogenous cholinesterase inhibitor, it is possible that it synergistically interacts with other AChEIs and improves their efficacy . As in the case of treatment response to donepezil, variability in rivastigmine efficiency could be explained by the effect of different gene variants . Some of the potential candidates are APOE , BCHE , presenilin ( PSEN ), and UDP glucuronosyltransferase 2B7 ( UGT2B7 ) genes. Better efficacy of combined rivastigmine and memantine therapy has been reported in APOE ε4 carriers . A multicenter study by Blesa and colleagues reported no association between APOE ε4 allele and the response to treatment with rivastigmine . The retrospective analysis by Farlow et al. investigated the efficacy of rivastigmine on cognitive performance in AD patients, taking into consideration the APOE genotype. The study reported more pronounced symptom improvement in subjects who were not APOE ε4 carriers in both rivastigmine and placebo groups . Similar to donepezil treatment efficacy, the BCHE K-variant affects the response to rivastigmine treatment, especially in the presence of the APOE ε4 allele . Presenilin is a subunit of γ-secretase, an enzyme that is crucial in processing APP, thus producing small peptides, including Aβ. Different mutations in the PSEN2 gene can lead to increased production of Aβ, including a common single adenine (A) nucleotide deletion polymorphism, which is located in the upstream promoter region of this gene . Zamani and colleagues reported the best treatment response to rivastigmine in AD patients with PSEN2 +A/−A genotype, alone or in combination with APOE ε3/ε3 or APOE ε4/ε4 genotype, while individuals with combined PSEN2 +A/+A and APOE ε3/ε4 genotypes had the worst response to treatment . UDP glucuronosyltransferase 2B7 is a metabolic enzyme important in the elimination of endogenous compounds and potentially toxic xenobiotics . Different polymorphisms in the UGT2B7 gene could alter the enzyme activity and, thus, affect the biotransformation of its substrates . The study by Sonali and colleagues investigated the effect of UGT2B7 (802C>T, UGT2B7*2 , rs7439366) polymorphism on rivastigmine efficiency, alone and in combination with memantine . Results suggested that carriers of the UGT2B7 variant, who were poor metabolizers, had poor clinical response to rivastigmine therapy . However, the study had a limited sample size, and further research is necessary to confirm or dispute these results. As already mentioned, galantamine is metabolized mainly by CYP3A4 and CYP2D6 enzymes, which is why CYP2D6 genetic variants have been associated with the outcome and side effects of galantamine treatment . A study by Ma and colleagues detected better treatment response in AD patients who were CYP2D6*10 rs1065852 carriers and reported fewer adverse side effects . Genetic variants of CHRNA7 are also interesting targets in pharmacogenetic studies focused on galantamine efficacy. Better treatment response to galantamine was reported in patients carrying minor allele variants of rs8024987 (C/G) or rs6494223 (C/T) polymorphism . Unlike in the case of AChEIs, there are not many studies that focus on the pharmacogenetics of memantine efficacy . From in vitro studies, we know that cytochrome P450 isozymes are not involved in the metabolism of memantine. Memantine is a substrate of the human organic cation transporter 2 (OCT2) , but its clearance is probably also related to other transporters, including organic cation/carnitine transporters (OCTN 1-3), the multidrug and toxin extrusion proteins (MATE1-2), and P-glycoprotein (P-gp) . Some studies also suggest the involvement of nuclear receptors in the regulation of cation transporters, including pregnane X receptor (PXR), constitutive androstane receptor (CAR), and peroxisome proliferator-activated receptor (PPAR) . Genetic variations in different membrane transporters could be associated with variability in memantine pharmacokinetics . Pregnane X regulates the expression of metabolic enzymes and transporters, which are involved in drug metabolism . The polymorphism rs1523130, located in the NR1I2 gene, which encodes pregnane X, was shown to modulate memantine elimination . Memantine clearance was found to be 16% slower in patients carrying at least one T allele (CT and TT genotype) . Ovejero-Benito and colleagues investigated the association of 67 polymorphisms in 21 genes, including CYP2D6 , CYP2C9 , CYP2A6 , ABCB1 , and genes coding for different neurotransmitter receptors, with donepezil or memantine pharmacokinetics and safety. The authors reported no significant association of analyzed SNPs with both memantine and donepezil pharmacokinetics or adverse drug reactions . In the early 2000s, more extensive research began on the effect of a combination of drugs on patients with different variations of genes essential for the onset of dementia. The effects of multifactorial therapy based on pharmacogenomics are most thoroughly described through the therapeutic response related to APOE and CYP2D6 variants in AD. Among all the genetic factors that affect the success of AD therapy, APOE is certainly the most important and affects over 50% of AD cases . In order to investigate the effects of APOE variants on multifactorial treatment, a two-year study with three drugs was conducted. APOE 3/4 carriers emerged as the best responders, and APOE 4/4 carriers as the worst. The response of APOE 2/3, APOE 4/4, and 4/5 was similar, where patients, after initial improvement, showed rapid deterioration . Genetic polymorphisms in CYP2D6 significantly affect drug metabolism and the interindividual response to therapy . The study that investigated the influence of CYP2D6 variants on the therapeutic response in AD patients used a four-drug therapy protocol for 1 year . The results showed that CYP2D6 - extensive metabolizers and CYP2D6 - intermediate metabolizers were the best responders to multifactorial therapy with cognition improvement after 1 year period, while in CYP2D6 - poor metabolizers and CYP2D6- ultra-rapid metabolizers, there was no therapeutic effect and cognitive functions continuously decreased during the mentioned period . Other polymorphic variants, like those of PS1 and PS2 genes, can influence the outcome of AD therapy in general and, therefore, multifactorial therapy, as well . Patients with different PS1 variants did not show significant differences in response to therapy, while regarding the PS2 gene, depending on the exon five variants, responses to therapy differed significantly, and PS2 − patients responded much better to therapy than those with PS2 + . One of the challenges of multifactorial therapy is the existing comorbidities of patients. Due to the relatively late onset of dementia, comorbidities are common and are mostly related to older age. In a study that included 2618 patients with AD, the average age was 76.1 years, and the most common comorbidities were hypertension, osteoarthritis, depression, diabetes mellitus, and cerebrovascular disease . The first problem that comes out of the above is the drug–drug interaction. Medications that a patient is receiving for existing conditions can reverse or modify the effects of dementia therapy and, thus, represent a major obstacle to its effectiveness. Another problem is the side effects of dementia therapy itself. The AChEIs side effects are associated with enhanced cholinergic tone , memantine causes off-target effects in other neurotransmitter systems, and its side effects are related to its anti-glutamatergic activity and even the first disease-modifying treatments for AD, anti-amyloid antibodies such as aducanumab and lecanemab are associated with amyloid-related imaging abnormalities (ARIA), which come in two forms: ARIA-E characterized by edema and ARIA-H characterized by hemorrhage . Carriers of the APOE ε4 allele showed an increased risk for ARIA, with APOE ε4 homozygotes being more prone to severe ARIA . Although there are no FDA-approved tests to determine individual genetic status prior to anti-amyloid treatment, the recommendation is that AD patients should be pre-screened for APOE genotypes due to the risk for ARIA . Non-cognitive symptoms, i.e., BPSD, are a major contributor to the heterogeneity of dementia. BPSD varies in different stages of the disease and includes symptoms such as depression, anxiety, apathy, agitation, delusions, and hallucinations. Due to their variability and prevalence, corresponding therapeutic approaches for these symptoms are an important part of the treatment of patients with dementia . As demonstrated in , pharmacological approaches to dementia treatment include, among others, psychotropics (e.g., antipsychotics, antidepressants, anticonvulsants) . The majority of psychotropic drugs used for treating neuropsychiatric diseases are metabolized by CYP1A2, CYP2B6, CYP2C8/9, CYP219, CYP2D6, and CYP3A4 enzymes . Studies have shown that antidepressant drugs, including tricyclic antidepressants, SSRIs, and norepinephrine/serotonin-reuptake inhibitors, are major substrates of CYP1A2, CYP2C9, CYP2C19, CYP2D6, CYP3A4, UGT1A4, and UGT1A3 enzymes, while typical and atypical antipsychotics are the main substrate of CYP1A2, CYP2C19, CYP2D6, CYP3A4, and UGT1A4 enzymes . Enzymes that are mostly involved in the metabolism of antidepressants and antipsychotics are CYP2D6 (86% and 72%, respectively) and CYP3A4 (72% and 75%, respectively) . Different classes of anticonvulsants, i.e., antiepileptics (benzodiazepines, barbiturates, miscellaneous antiepileptics, fatty acid derivatives, succinimides, oxazolidines, and hydantoin derivatives), are also mostly metabolized by CYP enzymes. For example, over 65% of antiepileptic drugs are major substrates for CYP (CYP3A4, CYP3A5, CYP2E1, CYP2C8, CYP2B6, CYP2D6, CYP2C19, CYP1A2, CYP2C9, CYP1A1, CYP1A6, CYP3A7, CYP2C18, CYP4B1) or UGT enzymes (UGT1A1, UGT1A3, UFT1A9, UGT2B7, UGT1A4, UGT1A6, UGT1A10, UGT2B15) . CYP3A4 is involved in the drug metabolism of most psychotropic drugs, compared to other isoforms . CYP2D6 enzyme plays an important role in oxidase reactions for a large number of commonly prescribed antidepressants, antipsychotics, and antiepileptics, which may act as substrates, inducers, or inhibitors . However, it has been reported that certain enzymatic activities of CYP2D6 and CYP2C19 are associated with treatment discontinuation . More than 100 different CYP2D6 alleles might show deficient (poor metabolizer, PM), normal (extensive metabolizer, EM), intermediate (intermediate metabolizer, IM), or increased (ultra-rapid metabolizer, UM) enzymatic activity, meaning that different patients will require different dosages . While the majority of the general population shows normal enzymatic activity , the proportion of extensive metabolizers and ultra-rapid metabolizers is slightly higher in the general population compared to AD cases . On the other hand, the proportion of intermediate metabolizers and poor metabolizers is vaguely lower in the general population compared to AD cases. It has been shown that between 10 and 20% of Caucasians carry defective CYP2D6 variants that influence drug metabolism, especially the metabolism of psychotropics. For example, it is shown for several antidepressants (amitriptyline, clomipramine, citalopram, doxepin, escitalopram, fluvoxamine, imipramine, paroxetine, sertraline, and trimipramine) and antipsychotics (aripiprazole, brexiprazole, haloperidol, pimozide, risperidone, and zuclopenthixol) that discontinuation of their treatment is associated with deficient and/or increased enzymatic activity of CYP2C19 and/or CYP2D6 . Likewise, a large number of individuals with altered responsiveness to benzodiazepines and neuroleptics show deficient or increased enzymatic activity, i.e., carry mutant variants of the CYP3A4 , CYP2D6 , and CYP2C9 genes . Association studies of CYP2D6 variants and genes ( ACE , AGT , APP , MAPT , APOE , PSEN1 , PSEN2 , FOS , and PRNP ) related to dementia demonstrated that in individuals with deficient or increased enzymatic activity, there is an accumulation of the risk variants, which might influence therapeutic response . Though there were no reported differences between females and males in the general population, the proportion of extensive metabolizers was somewhat higher in females than in males with AD, whereas poor metabolizers were more frequent in males than females with AD, suggesting a higher risk for males of developing an adverse drug reaction . In addition, variations in certain genes are associated with geographic and ethnic differences, which affect drug metabolism and, correspondingly, individual responses to a certain therapeutic approach . Associations between certain genetic variants encoding for various enzymes involved in drug metabolism and the effects (positive or adverse) of drug treatment have been extensively studied. The metabolism of antidepressants, antipsychotics, and anticonvulsants also includes, among others, various groups of enzymes (esterases, transferases, reductases, oxidases, histamine methytransferases), receptors (adrenergic, dopamine, and serotonin receptors), transporters (solute carrier family 6, ATP-binding cassettes), and channels (potassium voltage-gated and sodium channels), which are genetically variable . Polymorphic variations in the genes encoding for these proteins may influence drug metabolism . For example, variants in the genes encoding for the transporters of antipsychotics ( ABCB1 , SLC6A2 , SLC6A4 , SCN5A , KCNH2 , KCNE1 , KCNE2 , KCNQ1 ) and antidepressants ( SLC6A4 , SLC6A2 , ABCB1 , and 5-HTTLPR ) influence the metabolism of these drugs . It has also been reported that responsiveness to antipsychotics is higher in Ins/Ins carriers (−141C Ins/ins), as well as in A1 carriers of the Taq 1 A SNP of the DRD2 gene. The Ser allele of the Ser9Gly SNP in the DRD3 gene was associated with a better response to clozapine, while the Gly allele was associated with a higher risk for tardive dyskinesia . Several SNPs in the SLC6A4 and SLC6A3 genes showed an association between clozapine responsiveness and genotype or allele frequencies . In addition, antipsychotic-induced extrapyramidal symptoms ( DRD2 , HTR2A , GRIK3 , SLC6A4 VNTR , COMT Val158Met , ADORA1 , ADORA3 , ADORA2A ), tardive dyskinesia ( HTR2A , HTR2C , DRD2 , DRD3 , DPP6 , SOD2 , CYP2D6 , CNR1 , HSPG2 ), metabolic syndrome ( HTR2C , LEP , LEPR ), and other antipsychotic-induced symptoms have been associated with polymorphisms in several genes ( DRD2 , LEP , BDNF , LPL , TPH , etc.) . Moreover, Met/Met homozygotes of the Val108Met SNP in the COMT gene showed a better response to clozapine . Several polymorphisms in the serotonin receptor gene ( 5HTR2A ) showed that certain variants are associated with a better response to clozapine (A/A, A-1438G; His allele, His452Tyr), olanzapine (A/A genotype of A-1438G), or risperidone (C/C, T102C). Moreover, repeat-length polymorphisms in the serotonin transporter gene have been associated with responses to certain antidepressants and antipsychotics. For example, a long allele is associated with a better response to citalopram, paroxetine, fluoxetine, risperidone, and clozapine , while CYP2D6 and CYP2C19 variants are associated with antidepressant-induced symptoms, such as nightmares, anxiety, and panic attacks . Furthermore, polymorphisms in the genes SCN1A , ABCB1 , UGT2B7 , ABCC2 , CYP1A2 , HNF4A , and CYP3A5 are associated with altered drug metabolism of carbamazepine. Certain variants in the genes SCN1A , CYP2C9 , CYP2C19 , and ABCB1 influence the metabolism of phenytoin . Moreover, pathogenic variants in the SLC2A1 gene might predict the responsiveness and selection of adequate antiepileptics . Clobazam is a substrate for several CYP enzymes (CYP2C19, CYP3A4, CYP2B6, CYP2C18), while individuals with certain genotypes in the CYP2C19 , CYP3A4 , and CYP3A5 genes require adjustments in clobazam dosage due to adverse drug reactions . According to the different genes involved in the pharmacogenomics of AD as well as the response to antipsychotics, antidepressants, and antiepileptics, further studies are necessary for better characterization of the pharmacogenomics profile and determination of drug efficacy and safety in the treatment of non-cognitive symptoms of AD . Excessive anxiety and worry, as well as restlessness, fatigue, concentration problems, irritability, muscle tension, and sleep disturbance, are common symptoms in patients with dementia. According to a recent meta-analysis , prevalence rates of anxiety in dementia are around 40%, with no obvious association with the stages of illness or dementia severity . Moreover, people with dementia often experience sleep problems such as insomnia, impaired nocturnal sleep with increased awakenings, and decreased rapid eye movement (REM) sleep, as well as increased daytime sleep . The prevalence of sleep disorders especially rises in patients with VaD, LBD, or dementia related to PD . Additionally, sleep disruption normally interferes with the maintenance of cognitive health and is associated with the rate of cognitive decline in older adults . Anxiolytics, hypnotics, and sedatives are pharmaceuticals used for a reduction in anxiety, to relieve sleep difficulties, or to induce a calming effect. The primary group of medications within this category includes benzodiazepines. They are one of the most prescribed pharmaceuticals in developed countries, commonly used for the treatment of anxiety, sleep disorders, agitation, and alcohol withdrawal . However, the treatment methods for these non-cognitive disorders are more challenging in the context of dementia because, in dementia, they can manifest differently than in typical early-onset individual disorders. Moreover, benzodiazepines, as first-line anxiolytics and commonly used sedatives, might contribute to cognitive and psychomotor impairment . Due to the extensive list of potential side effects, the use of psychotropic drugs in older patients with dementia must be individually tailored. This means that, in addition to comorbidities and other concomitant medicines, distinctive individual characteristics, including pharmacogenomics factors, should be addressed when estimating the risks and benefits of prospective therapy. Metabolism of most benzodiazepines starts with oxidation, followed by conjugation to glucuronide, which is then eliminated by the urine . Although there are benzodiazepines that are directly conjugated, most of them go through the oxidation stage catalyzed by liver CYP enzymes , whose activity greatly influences drug metabolism and plasma concentration. As already mentioned, genes coding for CYP enzymes are highly polymorphic, influencing the enzyme’s activity and leading to absent, reduced, or increased drug metabolism. Consequently, higher drug concentrations, due to the poor metabolizing ability, can increase side effects or toxicity, while on the other hand, due to extensive drug metabolism, efficient therapeutic doses can be higher than usual. The majority of benzodiazepines are metabolized by CYP2C19 and CYP3A4/5; other enzymes such as CYP1A2, CYP2C9, and CYP2B6 may also play a role in the metabolism of some benzodiazepines . For example, it is known that diazepam is metabolized to nordiazepam by CYP2C19 and CYP3A4 and to temazepam by CYP3A4. Both metabolites undergo hydroxylation to oxazepam, which is catalyzed by CYP3A4 and/or CYP2C19 . However, a recent paper showed that the CYP2B6 phenotype also affects diazepam pharmacokinetic variability . Additionally, it was shown among the elderly population that carriers of CYP2C9*2 and CYP2C9*3 , as poor metabolism alleles, have an increased risk of falls associated with diazepam treatment . There are more than 40 polymorphic variants of the CYP2C19 gene, resulting in around 35 enzyme isoforms with at least 7 alleles ( CYP2C19*2 to CYP2C19*8 ) associated with partial or complete inactivation of the enzyme resulting in poor drug metabolism . On the other hand, the CYP2C19*17 variant is associated with increased activity, and carriers of this allele, especially homozygotes, are considered extensive metabolizers . The presence of poor metabolism alleles raises the chance of diazepam side effects, whereas the presence of CYP2C19*17 minimizes the risk of side effects but possibly decreases its efficacy when administered in a standard dose . Since the clearance of benzodiazepines decreases as the number of low metabolizing CYP2C19 alleles increases , it would be advisable to adjust their dose according to the CYP2C19 genotype. For example, Zubiaur et al. recommend lowering the dose of diazepam for 25–50% in patients whose genotype indicates poor drug metabolism. Enzyme CYP2C19 is also included in the metabolism of clobazam. It was shown that the response to clobazam was higher among carriers of poor metabolism variants, with an evident gene–dose effect . The same trend was noticed in the occurrence of side effects, such as drowsiness and dizziness, which were more prominent in poor metabolizers . Poor CYP2C19-associated metabolism of clobazam in a patient receiving a standard therapeutic dose for seizure disorder caused comatose condition due to the elevated concentration of clobazam active metabolite, norclobazam . Additionally, Riva et al. reported an increased enzymatic activity associated with the CYP2C19*17 allele. They found, however, that the magnitude of observed effects was smaller than the one reported for poor metabolizing alleles, implying that the effects of CYP2C19*17 probably do not have clinical significance, except for medicines with very narrow therapeutic windows . Another benzodiazepine, midazolam, is highly metabolized by CYP3A4 and CYP3A5, and it is also used as a probe substrate in studying the activity of those enzymes . Amino acid sequences for the two enzymes have 83% similarity, and the main differences between them are in their active sites and substrate access channels . There are studies reporting the association between CYP3A5 genotype and rates of midazolam hydroxylation and clearance . However, it seems that CYP3A genetics has only a limited impact on midazolam metabolism in vivo. Specifically, several studies reported a lack of the functional significance of polymorphisms resulting in common variants, including CYP3A4*1B , CYP3A5*3 , CYP3A5*6, and CYP3A5*7 . This could be due to the fact that midazolam is also a highly permeable substrate of P-glycoprotein . Additionally, plasma midazolam concentration and sedation grade were found to be associated with 1236C>T polymorphism of the MDR1 (multidrug resistance 1) gene . Metabolism of lorazepam, as well as structurally related benzodiazepines oxazepam and temazepam, skips phase I catalyzed by the CYP enzymes and is predominantly based on glucuronidation . Enzymes included in pharmaceuticals’ glucuronidation are uridine 5′-diphosphate-glucuronosyltransferases (UGTs) , with UGT1 and UGT2 enzymes mostly involved in drug metabolism processes. Variations in their genes, resulting in changes in their expression and function, are significant contributing factors to interindividual variability in drug disposition . For instance, the UGT2B15 genotype highly affects the pharmacokinetics of lorazepam. A single nucleotide polymorphism (G/T) in UGT2B15 gene coding region can result in UGT2B15*2 variant, which is associated with lower systemic clearance and metabolic activity of lorazepam and significantly higher lorazepam concentrations in homozygotes . Higher lorazepam plasma levels are associated with more pronounced clinical effects. For example, it was shown that UGT2B15*2 homozygotes, especially women, have greater postoperative anxiety reduction after lorazepam premedication when compared with carriers of other genotypes . Structurally different from benzodiazepines but with a similar mechanism of action via GABA signaling are Z-drugs, which have significant hypnotic effects by reducing sleep latency and enhancing sleep quality . The major metabolism pathways of Z-drug zolpidem include hydroxylation followed by oxidation, mediated mostly by CYP3A4; however, CYP2C9, CYP1A2, CYP2D6, and CYP2C19 have also been reported to be included . In a previous study, the CYP3A4*18 variant was associated with increased and CYP2C19*2 with reduced zolpidem metabolism . However, other authors reported no evidence for the impact of the CYP2C19 genotype on the pharmacokinetic parameters of zolpidem . In another study, participants received zolpidem and clarithromycin, a CYP3A4 inhibitor, in order to eliminate the contribution of CYP3A4 to zolpidem metabolism. However, no differences in zolpidem plasma concentrations were found when subjects were divided according to CYP2D6 genotype . Similarly, a lack of association between CYP2C9 genotype and zolpidem metabolism was also reported Pharmacogenomic research resulted in various findings contributing to the improvement in establishing the anxiety treatment and predicting its outcome . It is obvious that both clinicians and patients could benefit from defining the relations between genetic variation and variable drug responses to anxiolytics and sedatives, especially considering the high prescription rates of this group of psychiatric medications. So far, dementia treatment has been directed against only several pharmacological targets, emphasizing the need for the development of novel therapeutic strategies. Since the therapeutic response is a complex trait, it is not likely that a single drug could be effective in the treatment of a variety of cognitive impairments, behavioral disturbances, and functional decline . Therefore, multifactorial treatments with a combination of several drugs represent the most feasible option in dementia. However, current as well as potential novel anti-dementia treatments of both cognitive and neuropsychiatric symptoms require evaluation from a pharmacogenomic perspective on a case-by-case basis in order to obtain optimal therapeutic efficacy, as well as to avoid drug side effects and unnecessary costs. Pharmacogenomics could offer help in detecting safer and more effective medications for each dementia patient, as well as new pharmacotherapeutic targets, whose identification has been complicated by the interplay of numerous genetic factors with only minor, moderate effect on pharmacokinetic or pharmacodynamic variability . Although prediction of drug response with respect to genetic variations affecting ADME has already been established, further studies are needed to better understand the functional consequences of genetic polymorphisms in neurotransmitter receptors, transporters, and signal transduction molecules. In addition to genetic background, drug efficacy, and safety are influenced by many other factors, including mechanisms of drug action, drug-specific adverse reactions, drug–drug interactions, nutritional factors, etc. . Patient characteristics, such as age, gender, and ethnicity, also represent important parameters that might determine individual drug response. Despite the accumulation of genetic information on dementia, the role of epigenetic and environmental factors is still not well known. Hence, in order to better understand such complex multifactorial disorders, both gene–gene and gene–environment models need to be established. Moreover, recent progress in functional genomics, proteomic profiling, high-throughput screening methods, large databases, and bioinformatic tools stimulates the development of pharmacogenomic studies, speeding up clinical trials, improving patient stratification, reducing costs and potential adverse effects and optimizing therapeutic outcomes . Despite the challenge of translation from the research laboratory into clinics, pharmacogenomics holds promise of future cost-effective, safe, and efficacious personalized medicine for patients with dementia. However, future research and strategy advances are needed to overcome scientific, economic, and clinical obstacles and involve pharmacogenomics as a routine intervention in personalized treatment approaches in neuropsychiatry worldwide. |
Pioneering family medicine: A collaborative global health education partnership in Ethiopia | 7ed0bc2c-acc0-46aa-a8f4-924e4cbf712b | 11447600 | Family Medicine[mh] | In 2013, in line with a wave of similar efforts across the African region, Ethiopia launched its first Family Medicine (FM) programme at Addis Ababa University, College of Health Sciences, School of Medicine. , The Toronto Addis Ababa Academic Collaboration in Family Medicine (TAAAC-FM) is a capacity-strengthening partnership established between Addis Ababa University’s Department of Family Medicine (AAU-FM) and the Department of Family and Community Medicine (DFCM) at the University of Toronto (U of T) to support Ethiopia’s inaugural FM residency programme. The TAAAC-FM collaboration was built on a pre-existing model initially established between the departments of psychiatry at AAU and U of T. One of the distinctive features of the model is the presence of two U of T faculty and one senior learner in Addis for 1 month, three times per year. This paper describes this unique institutional collaboration and identifies four key levers of academic engagement and partnership that contributed to the ongoing evolution of Ethiopian family medicine. Lastly, it distils some of the main lessons learned to date. Primary health care (PHC), as a whole-of-system approach to health and wellbeing, has been recognised by the World Health Organization (WHO) as the necessary pathway to achieve universal health coverage (UHC) and other health related Sustainable Development Goals (SDGs). It includes as one of three essential components, primary care and essential public health functions at the core of integrated health services. Family medicine is a medical discipline that inherently drives toward high-quality primary care. Primary care, as a core component of PHC, delivers first-contact, accessible, continued, comprehensive, coordinated and person-focused care. Primary care is pivotal in bolstering PHC-oriented systems and advancing UHC through person-centred care across the lifespan. In 1991, in an effort to address one of the most severe shortages of physicians in the world and orient towards a PHC approach, the Ethiopian Federal Ministry of Health launched ambitious Health Sector Development Plans (1991–2015). Significant emphasis was placed on developing human resources for health (HRH) including increasing the number of physicians and enhancing postgraduate training programmes. The establishment of a FM postgraduate training programme in Ethiopia emerged in response to the prioritisation of the development of HRH and in particular of comprehensive, patient-centred, community-oriented primary care. Philpott et al. detail the conceptualisation, early development, and launch of the country’s first-ever AAU-FM residency in Ethiopia in 2013. The institutional partnership now known as the TAAAC-FM is an invited partnership from the emerging AAU-FM department to share the experience and expertise of Canadian family medicine, expecting that it would be reshaped, adapted, and contextualised to meet the specific needs of Ethiopia. Since 2013, over 40 DFCM faculty, 10 current AAU faculty, and over 70 graduated residents from AAU-FM have participated in 44 virtual teaching sessions, and 31 on-site teaching trips supported with in-kind contributions from both AAU and DFCM. The TAAAC-FM collaboration is grounded in shared contributions from both partners and has been supported by transformative gifts and in-kind supports, which have exponentially grown the capacity of this partnership over the past 10 years. It is grounded in deep commitment, dedication, and trust. provides a snapshot of the key programme features. A first-of-its-kind programme in Ethiopia, TAAAC-FM is guided by principles of collaboration, respect, sustainability, responsiveness, and flexibility. Collaborative activities focus on FM’s scholarly foundations. These included curriculum development and adaptation, learner evaluation, teaching (didactic, case-based, and skills-based training), faculty and leadership development, and scholarship or knowledge sharing as a form of advocacy . Teaching activities delivered during the month-long DFCM faculty visits or virtually are jointly planned by AAU-FM and DFCM faculty using a needs-based approach. This supports strategic decision-making and resource allocation that prioritises the evolving needs and preferences of the programme as identified by the AAU-FM programme director, leadership and faculty. Monthly virtual meetings bring together TAAAC-FM leaders from AAU and the DFCM to debrief, discuss, plan, and problem-solve in ways that are respectful of culture, academic context, and values. In-person requested teaching, during the month-long teaching trips, is delivered jointly by DFCM and AAU-FM faculty at academic teaching days, community health centres, and on the wards. Support for the examination and certification of trainees has included exam development and review with DFCM examiners on invitation by the AAU-FM leadership. The coronavirus disease 2019 (COVID-19) pandemic necessitated a pivot that included the introduction of virtual co-teaching, and virtual faculty development, all of which are now integrated into the collaborative teaching programme as needed. To facilitate and orient DFCM faculty to their roles within the TAAAC-FM collaboration, the DFCM provides detailed FM-specific orientation to the principles of TAAAC-FM alongside consultation with AAU-FM faculty to inform teaching priorities and preparation . Faculty development and leadership development has been a priority focus for TAAAC-FM, with virtual and in-person sessions at AAU-FM, as well as leadership course offerings at U of T focused on capacity-building for FM leaders . Addis Ababa University’s Department of Family Medicine faculty have been sponsored to present scholarly work and attend key family medicine conferences as a mechanism to enhance networking and building a FM community. The TAAAC-FM collaboration is committed to the principles of allyship, including a shared vision and goals, mutual trust and respect, a bidirectional flow of knowledge, flexibility, and navigating challenges together. These principles, further outlined in , are critical to the success of this partnership. Addis Ababa University’s Department of Family Medicine delivers a robust 3.5-year residency programme, with its own autonomous leadership and oversight of the residency programme, operations, curriculum, and functions, while navigating the complexity of a newly emerging specialty in the healthcare system. Toronto Addis Ababa Academic Collaboration in Family Medicine has emphasised building trustful, relational leadership and communication, underpinning the institutional and country-context, to bolster AAU-FM’s growth and evolution. Both DFCM and AAU-FM faculty and learners have learned to recognise and address the operational challenges of communication, co-design of contextually relevant teaching materials, and participating in orienting and debriefing activities. Department of Family and Community Medicine teaching faculty are oriented to the scope of the role as a partner, valuing skills of remaining flexible to changing programme circumstances, upholding AAU-FM’s identified academic expectations, adapting to identified curriculum gaps and learning needs, while teaching in areas of strength for all partners. Addis Ababa University’s Department of Family Medicine faculty are new to the FM landscape but have leveraged the TAAAC-FM partnership to lead and represent the specialty while engaging multiple stakeholders. Addis Ababa University’s Department of Family Medicine faculty are increasingly engaged in FM associations and stakeholder organisations both within Ethiopia, and beyond. Addis Ababa University’s Department of Family Medicine faculty and graduates hold leadership positions at AAU, serving as Department Heads and committee leads. They are founding members of the Ethiopian Society of Family Physicians who actively engage with the Ministry of Health to shape policy and curriculum and collaborate with non-governmental organisations, including Hospice Ethiopia. Addis Ababa University’s Department of Family Medicine faculty are participating in global family medicine communities including the Canadian College of Family Physicians, the Besrour Centre for Global Family Medicine in Canada, the Primary Care and Family Medicine (PRIMAFAMED) network in sub-Saharan Africa, and the World Organization of Family Doctors. These intentional efforts strengthen connections to broader FM organisations, supporting the development of academic leaders, advocates, and scholars at both AAU-FM and the DFCM. The Toronto Addis Ababa Academic Collaboration in Family Medicine is cultivating a Community of Practice (CoP) as a group that is connected together through shared vision, knowledge, and interest, to support an emerging FM community that promotes joint learning to improve healthcare service, practice, and delivery. The Toronto Addis Ababa Academic Collaboration in Family Medicine has placed value on forging strong FM faculty connections that extend beyond AAU-FM. Shared resources, sharing lived experience as practising family doctors and educators, co-teaching at AAU, co-presenting, and co-designing new initiatives have been tangible CoP building opportunities for both AAU-FM and DFCM faculty. Addis Ababa University’s Department of Family Medicine and DFCM continue to build this CoP to address shared practice interests, knowledge, and concerns. Furthermore, DFCM faculty work alongside AAU-FM faculty during clinical supervision of learners and didactic teaching when on site. This further fosters reciprocal learning experiences for both AAU-FM and DFCM faculty on themes of clinical presentations, a culture of feedback and preceptorship, cultural humility, and interactive teaching styles. Strengthening FM has been linked to better health outcomes, lower costs, and improved health equity. With comprehensive training and resources, FM specialists can significantly contribute to patient care in the community, health centres, clinics, and hospitals, thereby reshaping the healthcare landscape towards PHC. The internationalisation of medical education through TAAAC-FM comes with a recognition that all offerings evolve based on the AAU-FM specific needs and capacity identified. The offerings are dynamic rather than standardised, and teaching themes or skills are adapted to address the gaps identified by AAU-FM faculty, recognising AAU-FM growth and internal capacity annually. The next phase of the TAAAC-FM partnership aims to adapt collaborative efforts through consideration of AAU-FM’s strength and capacity, principles of allyship, recognition of AAU-FM’s request for ongoing faculty development, while fostering linkages between the Ethiopian and global healthcare community. Professional development will continue to be scaffolded to meet needs identified by AAU faculty while strengthening the knowledge pool of faculty partners at both institutions. The challenges of the COVID-19 pandemic and human resource limitations in FM in both institutional climates are critical considerations going forward. Bi-directional learning in particular has transformed the lens of partnership work for the DFCM, including emphasis on active listening to evolving AAU priorities, strengthening DFCM’s contextual humility, development of orientation and debrief processes, clinical learning while reflecting on health workforce challenges, and considerations for models of care in the future. The DFCM and AAU have learned many lessons in the development of this educational partnership and some of the main lessons to date are shared in . This report provides an overview crafted by the past FM leadership of experiences and insights gained over the 10 years. The innovative TAAAC-FM partnership has highlighted four key levers in high-low resource setting academic engagement and transformation: strengthening education, orientation through preparatory contextualisation, faculty development and leadership, and knowledge dissemination or scholarship. Through these levers, TAAAC-FM has striven to: (1) strengthen teaching capacity in FM for the enhanced delivery of primary care; (2) establish a knowledge-sharing CoP that encourages collaboration and partnership; and (3) enhance local-to-global leadership opportunities with rich lessons learned. These insights may have value to similar partnerships, and highlight the rich expertise, wisdom, collaboration, and friendship of this impactful allyship between the AAU-FM and the DFCM in their joint commitment to advancing primary care and health system strengthening. |
The impact of specialized pediatric palliative care on advance care planning and healthcare utilization in children and young adults: a retrospective analysis of medical records of in-hospital deaths | 8cda138e-899f-47ea-90a3-1e3bed34a897 | 11110344 | Pediatrics[mh] | Pediatric palliative care is an additional layer of support for infants, children, adolescents, and young adults living with life-limiting conditions, along with their families, aiming to minimize physical, psychosocial, and spiritual suffering and enhance quality of life . As an integrative model of pediatric palliative care, specialized palliative care (SPC) is provided by a multidisciplinary team of professionals to children and families with more complex care needs . Advance care planning (ACP), a key element of palliative care, involves anticipatory discussion and healthcare decision-making. ACP is not limited to life-sustaining treatment decisions at the end-of-life; it is a systematic, family-centered conversation to reflect the patient’s and family’s values and preferences into goals of care . Palliative care and ACP discussion often result in patients using less inpatient and intensive care and more outpatient, community, and home-based services during the end-of-life in the adult population, which has been the focus of previous research . Palliative care is not limited to end-of-life care or intended to control healthcare use and costs. However, not assessing the patients’ and families’ preferences for care may lead to increased hospitalization at the end-of-life, undermining the goal concordant care, one of the important quality indicators of successful ACP . ACP is also crucial for children and young adults, as those with life-limiting conditions require multiple specialized healthcare and complicated symptom management, even at the end-of-life . Several studies on the benefits of pediatric SPC have been published in recent years, reporting mixed results on the impact on the use of acute healthcare and intensive medical treatments at the end-of-life . Regarding impact on ACP, patients with SPC showed a higher percentage of ACP discussion , and they initiated ACP earlier . As for healthcare use, previous studies reported that, in children who died of cancer, SPC was associated with reduced intensive care unit (ICU) admissions and less intensive end-of-life care , fewer invasive procedures, and fewer deaths in the ICU . Other studies examining children’s deaths in a children’s hospital suggested that the acute hospital utilization is more affected by the proximity to death and expected deterioration than by the SPC involvement . In addition, few studies included a wide range of ages or disease groups, making it difficult to understand the overall impact of palliative care for children and adolescents. Therefore, this study aimed to explore the impact of SPC on children and young adults who died while hospitalized in a single institution. Specifically, we examined the impact of SPC on the patterns of decision-making, end-of-life care, and healthcare use during the patient’s last month of life by comparing before and after the implementation of palliative care. Moreover, we compared patients with and without palliative care. Study design A retrospective review of medical records was conducted at Seoul National University Children’s Hospital (SNUCH). The ethics committee of the Seoul National University Hospital (registration number 2010-112-1165) waived the requirement to obtain written informed consent, as this was a retrospective study. This study was reported in accordance with the REporting of studies Conducted using Observational Routinely collected health Data (RECORD) recommendations . Setting SNUCH, the largest pediatric tertiary care center in South Korea (317 beds; approximately 1,000 daily outpatient visits), launched a Dreamseeds Center, an SPC service, in 2014. The center provides consultation services and outpatient clinics by a multidisciplinary team of physicians (a palliative care physician and a psychiatrist), nurses, a social worker, expressive therapists (art, movement, play), and more. SPC begins when a patient’s primary physician makes a referral. The center provides care services for inpatients and outpatients, including pain and symptom management, communication and decision-making support, care coordination, emotional and social support, art therapy, and bereavement care. In addition, telephone counseling and need-based home visits are provided for patients at home during end-of-life. Participants This study included all patients aged < 25 years who were treated and died at SNUCH in two different periods, before or after the implementation of palliative care: (a) pre-period (1 January 2011 to 31 December 2013) and (b) post-period (1 January 2017 to 31 December 2019). These periods were selected to account for the time it takes for acculturation of integrated palliative care at the institutional level. Sources Patients were identified as having the status of “death” at discharge, and then data were retrieved from SNUCH’s clinical data warehouse (SUPREME; Seoul National University Hospital Patient Research Environment, https://supreme.snuh.org ). Information on healthcare use and costs for inpatients, outpatients, and emergency services was obtained from institutional administrative data. Finally, the SNUCH palliative care registry was queried to see if the patient received palliative care. The SNUCH palliative care registry is an independent database containing detailed information on care plans and service provisions for each patient and family. Data were linked using hospital patient identifiers and dates of birth as individual-level indicators. Then, SPC professionals (a pediatric palliative physician, a psychiatrist, two nurses, and a social workers) reviewed the patient’s health records and/or SNUCH palliative registry to identify the ACP variables. Data were collected and linked between October and December 2021. Variables We collected patients’ demographic and clinical characteristics (date of birth, sex, insurance type, residential address, primary diagnoses, treatment duration, date of death, and location of death) and SPC enrollment. Treatment duration was defined as months from diagnosis to death based on the primary diagnosis of the last hospitalization. We defined ACP at three levels; if a preference or plan for future care was recorded in the medical records, we categorized it as “discussed” and collected the date to generate “the days from ACP initiation to death”; if wishes for life-sustaining treatment (LST) (cardiopulmonary resuscitation [CPR], hemodialysis, mechanical ventilation, chemotherapy which are specified by the law) are recorded in the medical records but there was no completed legal document, we categorized it as “medical documentation on LST;” if there was a legal document on wishes for LST, we categorized as “legal documentation on LST.” Furthermore, end-of-life care characteristics during the last month of life were collected, including the number of hospitalizations, length of stay, use of intensive treatments (mechanical ventilation, CPR, ICU admission, and ICU days), inpatient costs, number of outpatient department (OPD) visits, and ED visits. Admission following an ED visit was categorized as admission, as the data were not distinguishable; therefore, ED visits could be underestimated in this study. Variables of pediatric ICU admission rates and days were generated to analyze ICU utilization only with physical deterioration, excluding patients who were born and stayed in the neonatal intensive care unit (NICU) until death. The primary diagnoses were categorized using the complex chronic condition (CCC) classification (version 2) based on the International Classification of Disease, 10th edition , using the “pccc” R package . Once a patient was referred to the SPC service but died outside the hospital, demographic and clinical characteristics were extracted from the SNUCH palliative care registry. Statistical methods Descriptive statistics were generated, including means with standard deviations and medians with interquartile ranges for continuous variables, and frequencies and percentages for categorical variables. Demographic and clinical characteristics of patients who died before the introduction of SPC (pre-period) were compared with those who died after (post-period). To examine the impact of SPC on end-of-life care, we compared patterns of end-of-life care for patients who received palliative care (SPC group) and those who did not receive palliative care (non-SPC group) using Fisher’s exact test and the Wilcoxon rank sum test for categorical or continuous variables, respectively. Multivariable logistic regression analysis of key end-of-life care characteristics (completion of legal ACP document, opioid use, death in the general ward, mechanical ventilation, CPR, and ICU admission) was followed to investigate adjusted odds ratios with post-period or SPC as an independent variable controlling confounding factors. Finally, a multivariable linear regression model was fitted to identify the associations between demographic and clinical characteristics and days from the initial ACP to death. Log transformation of the days from the initial ACP to death was performed to normalize the residuals in the regression analysis. Variables with marginally significant associations ( p < .10) in univariable analysis were included in the multivariable model. Data were analyzed using R version 4.0.2. A retrospective review of medical records was conducted at Seoul National University Children’s Hospital (SNUCH). The ethics committee of the Seoul National University Hospital (registration number 2010-112-1165) waived the requirement to obtain written informed consent, as this was a retrospective study. This study was reported in accordance with the REporting of studies Conducted using Observational Routinely collected health Data (RECORD) recommendations . SNUCH, the largest pediatric tertiary care center in South Korea (317 beds; approximately 1,000 daily outpatient visits), launched a Dreamseeds Center, an SPC service, in 2014. The center provides consultation services and outpatient clinics by a multidisciplinary team of physicians (a palliative care physician and a psychiatrist), nurses, a social worker, expressive therapists (art, movement, play), and more. SPC begins when a patient’s primary physician makes a referral. The center provides care services for inpatients and outpatients, including pain and symptom management, communication and decision-making support, care coordination, emotional and social support, art therapy, and bereavement care. In addition, telephone counseling and need-based home visits are provided for patients at home during end-of-life. This study included all patients aged < 25 years who were treated and died at SNUCH in two different periods, before or after the implementation of palliative care: (a) pre-period (1 January 2011 to 31 December 2013) and (b) post-period (1 January 2017 to 31 December 2019). These periods were selected to account for the time it takes for acculturation of integrated palliative care at the institutional level. Patients were identified as having the status of “death” at discharge, and then data were retrieved from SNUCH’s clinical data warehouse (SUPREME; Seoul National University Hospital Patient Research Environment, https://supreme.snuh.org ). Information on healthcare use and costs for inpatients, outpatients, and emergency services was obtained from institutional administrative data. Finally, the SNUCH palliative care registry was queried to see if the patient received palliative care. The SNUCH palliative care registry is an independent database containing detailed information on care plans and service provisions for each patient and family. Data were linked using hospital patient identifiers and dates of birth as individual-level indicators. Then, SPC professionals (a pediatric palliative physician, a psychiatrist, two nurses, and a social workers) reviewed the patient’s health records and/or SNUCH palliative registry to identify the ACP variables. Data were collected and linked between October and December 2021. We collected patients’ demographic and clinical characteristics (date of birth, sex, insurance type, residential address, primary diagnoses, treatment duration, date of death, and location of death) and SPC enrollment. Treatment duration was defined as months from diagnosis to death based on the primary diagnosis of the last hospitalization. We defined ACP at three levels; if a preference or plan for future care was recorded in the medical records, we categorized it as “discussed” and collected the date to generate “the days from ACP initiation to death”; if wishes for life-sustaining treatment (LST) (cardiopulmonary resuscitation [CPR], hemodialysis, mechanical ventilation, chemotherapy which are specified by the law) are recorded in the medical records but there was no completed legal document, we categorized it as “medical documentation on LST;” if there was a legal document on wishes for LST, we categorized as “legal documentation on LST.” Furthermore, end-of-life care characteristics during the last month of life were collected, including the number of hospitalizations, length of stay, use of intensive treatments (mechanical ventilation, CPR, ICU admission, and ICU days), inpatient costs, number of outpatient department (OPD) visits, and ED visits. Admission following an ED visit was categorized as admission, as the data were not distinguishable; therefore, ED visits could be underestimated in this study. Variables of pediatric ICU admission rates and days were generated to analyze ICU utilization only with physical deterioration, excluding patients who were born and stayed in the neonatal intensive care unit (NICU) until death. The primary diagnoses were categorized using the complex chronic condition (CCC) classification (version 2) based on the International Classification of Disease, 10th edition , using the “pccc” R package . Once a patient was referred to the SPC service but died outside the hospital, demographic and clinical characteristics were extracted from the SNUCH palliative care registry. Descriptive statistics were generated, including means with standard deviations and medians with interquartile ranges for continuous variables, and frequencies and percentages for categorical variables. Demographic and clinical characteristics of patients who died before the introduction of SPC (pre-period) were compared with those who died after (post-period). To examine the impact of SPC on end-of-life care, we compared patterns of end-of-life care for patients who received palliative care (SPC group) and those who did not receive palliative care (non-SPC group) using Fisher’s exact test and the Wilcoxon rank sum test for categorical or continuous variables, respectively. Multivariable logistic regression analysis of key end-of-life care characteristics (completion of legal ACP document, opioid use, death in the general ward, mechanical ventilation, CPR, and ICU admission) was followed to investigate adjusted odds ratios with post-period or SPC as an independent variable controlling confounding factors. Finally, a multivariable linear regression model was fitted to identify the associations between demographic and clinical characteristics and days from the initial ACP to death. Log transformation of the days from the initial ACP to death was performed to normalize the residuals in the regression analysis. Variables with marginally significant associations ( p < .10) in univariable analysis were included in the multivariable model. Data were analyzed using R version 4.0.2. Demographic and clinical characteristics Our analysis identified 479 patients aged < 25 years who died in the hospital; of these, 205 (42.8%) deaths occurred during the post-period, and 123 patients were enrolled in SPC (60%) (Fig. ). No statistically significant difference was found in the demographic and clinical characteristics between pre- and post-periods except the proportion of hematological or immunological condition (Table ). The non-SPC and SPC groups were similar in demographic characteristics, although the SPC group had a higher mean age. A larger proportion of patients with malignancy ( p < .001), congenital or genetic ( p = .007), neurologic and neuromuscular ( p = .044), hematological or immunological condition ( p < .001) received SPC, whereas a larger proportion of patients with premature and neonatal condition did not ( p < .001). Multiple CCCs was similar between the pre- and post-period; however, a larger proportion of patients with two or more CCC received SPC ( p < .001). The SPC group showed longer treatment duration than the non-SPC group, which might be explained by differences in CCC categories, particularly neurological and neuromuscular conditions. During the post-period, 59 patients received SPC but died outside the hospital; therefore, these patients were excluded from the analysis. Patients who died outside the hospital were older than those who died in the hospital (median age of 9 years vs. 4 years, p = .001) and were less likely to have cardiovascular (35.8% vs. 15.3%, p = .007), congenital or genetic (20.3% vs. 6.8%, p = .034), and hematological or immunological condition (38.2% vs. 15.3%, p = .003) (Supplementary Table ). The analysis did not include the location of death for patients who died outside the hospital due to incomplete data, which included other hospitals, home, and unknown places. Comparison of before and after the implementation of specialized palliative care In the post-period, more patients engaged in ACP, and they were less likely to receive intensive care at the end-of-life compared to those in the pre-period (Table ). More patients in the post-period engaged in ACP and completed medical and legal documents than those in the pre-period. Days from initial ACP to death were longer in the post-period, indicating earlier ACP discussion. During the last month of life, patients who died in the post-period were less likely to receive mechanical ventilation, cardiopulmonary resuscitation, and more likely to receive opioids. After controlling sex, being infant, residence, insurance type, CCC categories and number of CCCs, patients who died in the post-period were more likely to complete ACP legal document (adjusted odds ratio [aOR] 1.62, 95% CI 1.08 to 2.46), use opioids (aOR 1.89, 95% CI 1.19 to 3.06) and less likely to receive mechanical ventilation (aOR 0.37, 95% CI 0.19 to 0.68), and CPR (aOR 0.45, 95% CI 0.30 to 0.68) (Supplementary Table ). Comparison between patients who received specialized palliative care and those who did not Patients in SPC group were more likely to engage in ACP and were less likely to receive intensive care at the end-of-life than those in non-SPC group (Table ). Patients enrolled in SPC had a higher proportion of ACP and legal documentation. ACP occurred earlier in the SPC group than in the non-SPC group. Additionally, patients in the SPC group were less likely to be mechanically ventilated, more likely to receive opioids, and less likely to receive CPR. The proportions of patients that received transfusion, antibiotics, or chemotherapy were similar, regardless of SPC involvement. In both the SPC and non-SPC groups, over half of the patients died in the ICU (pediatric intensive care unit [PICU] or NICU) with PICU days was longer in SPC group. However, 49 patients enrolled in SPC (39.8%) died in the general ward, compared with only three patients (3.7%) who were not enrolled. The multivariable regression analysis revealed that the SPC group remained a significant factor in explaining the completion of ACP legal documentation (aOR 5.47, 95% CI 2.53 to 12.31), opioids (aOR 19.18, 95% CI 5.64 to 82.61), and CPR (aOR 0.18, 95% CI 0.08 to 0.41) after controlling for confounding factors, including sex, being an infant, CCC categories, and number of CCCs (Supplementary Table ). Regarding healthcare use in the last month of life, patients enrolled in SPC showed more hospital and ICU days. While the PICU admission rates were comparable between the SPC and non-SPC groups, the former had more PICU days. In addition, the SPC group accrued higher total costs for inpatient services; however, the cost per inpatient each day was lower in the SPC group. Patients enrolled in SPC were more likely to visit the OPD during the last month of life, relative to non-SPC patients; however, the two groups had similar numbers of visits to the OPD or ED. Factors associated with early advance care planning prior to death The days from the initial ACP to death was analyzed to identify factors associated with early engagement in ACP. Among 169 patients who discussed ACP, average days from initial ACP to death was 37.6 days (standard deviation 79.3 days). Supplementary Table depicts the descriptive statistics of days from the initial ACP to death. A multivariable linear regression model controlling for sex, being an infant, residential area, insurance type, malignancy, and neurological and neuromuscular condition was conducted (Table ). SPC involvement was associated with more days from initial ACP to death, indicating earlier ACP (β 1.44, 95% CI 0.89 to 1.99, p < .001) (Fig. ), while being an infant was negatively associated with earlier ACP (β -0.74, 95% CI -1.28 to -0.19, p = .008). Being diagnosed with a neurological and neuromuscular condition was associated with more days from the initial ACP to death (β 0.76, 95% CI 0.17 to 1.35, p = .012). Our analysis identified 479 patients aged < 25 years who died in the hospital; of these, 205 (42.8%) deaths occurred during the post-period, and 123 patients were enrolled in SPC (60%) (Fig. ). No statistically significant difference was found in the demographic and clinical characteristics between pre- and post-periods except the proportion of hematological or immunological condition (Table ). The non-SPC and SPC groups were similar in demographic characteristics, although the SPC group had a higher mean age. A larger proportion of patients with malignancy ( p < .001), congenital or genetic ( p = .007), neurologic and neuromuscular ( p = .044), hematological or immunological condition ( p < .001) received SPC, whereas a larger proportion of patients with premature and neonatal condition did not ( p < .001). Multiple CCCs was similar between the pre- and post-period; however, a larger proportion of patients with two or more CCC received SPC ( p < .001). The SPC group showed longer treatment duration than the non-SPC group, which might be explained by differences in CCC categories, particularly neurological and neuromuscular conditions. During the post-period, 59 patients received SPC but died outside the hospital; therefore, these patients were excluded from the analysis. Patients who died outside the hospital were older than those who died in the hospital (median age of 9 years vs. 4 years, p = .001) and were less likely to have cardiovascular (35.8% vs. 15.3%, p = .007), congenital or genetic (20.3% vs. 6.8%, p = .034), and hematological or immunological condition (38.2% vs. 15.3%, p = .003) (Supplementary Table ). The analysis did not include the location of death for patients who died outside the hospital due to incomplete data, which included other hospitals, home, and unknown places. In the post-period, more patients engaged in ACP, and they were less likely to receive intensive care at the end-of-life compared to those in the pre-period (Table ). More patients in the post-period engaged in ACP and completed medical and legal documents than those in the pre-period. Days from initial ACP to death were longer in the post-period, indicating earlier ACP discussion. During the last month of life, patients who died in the post-period were less likely to receive mechanical ventilation, cardiopulmonary resuscitation, and more likely to receive opioids. After controlling sex, being infant, residence, insurance type, CCC categories and number of CCCs, patients who died in the post-period were more likely to complete ACP legal document (adjusted odds ratio [aOR] 1.62, 95% CI 1.08 to 2.46), use opioids (aOR 1.89, 95% CI 1.19 to 3.06) and less likely to receive mechanical ventilation (aOR 0.37, 95% CI 0.19 to 0.68), and CPR (aOR 0.45, 95% CI 0.30 to 0.68) (Supplementary Table ). Patients in SPC group were more likely to engage in ACP and were less likely to receive intensive care at the end-of-life than those in non-SPC group (Table ). Patients enrolled in SPC had a higher proportion of ACP and legal documentation. ACP occurred earlier in the SPC group than in the non-SPC group. Additionally, patients in the SPC group were less likely to be mechanically ventilated, more likely to receive opioids, and less likely to receive CPR. The proportions of patients that received transfusion, antibiotics, or chemotherapy were similar, regardless of SPC involvement. In both the SPC and non-SPC groups, over half of the patients died in the ICU (pediatric intensive care unit [PICU] or NICU) with PICU days was longer in SPC group. However, 49 patients enrolled in SPC (39.8%) died in the general ward, compared with only three patients (3.7%) who were not enrolled. The multivariable regression analysis revealed that the SPC group remained a significant factor in explaining the completion of ACP legal documentation (aOR 5.47, 95% CI 2.53 to 12.31), opioids (aOR 19.18, 95% CI 5.64 to 82.61), and CPR (aOR 0.18, 95% CI 0.08 to 0.41) after controlling for confounding factors, including sex, being an infant, CCC categories, and number of CCCs (Supplementary Table ). Regarding healthcare use in the last month of life, patients enrolled in SPC showed more hospital and ICU days. While the PICU admission rates were comparable between the SPC and non-SPC groups, the former had more PICU days. In addition, the SPC group accrued higher total costs for inpatient services; however, the cost per inpatient each day was lower in the SPC group. Patients enrolled in SPC were more likely to visit the OPD during the last month of life, relative to non-SPC patients; however, the two groups had similar numbers of visits to the OPD or ED. The days from the initial ACP to death was analyzed to identify factors associated with early engagement in ACP. Among 169 patients who discussed ACP, average days from initial ACP to death was 37.6 days (standard deviation 79.3 days). Supplementary Table depicts the descriptive statistics of days from the initial ACP to death. A multivariable linear regression model controlling for sex, being an infant, residential area, insurance type, malignancy, and neurological and neuromuscular condition was conducted (Table ). SPC involvement was associated with more days from initial ACP to death, indicating earlier ACP (β 1.44, 95% CI 0.89 to 1.99, p < .001) (Fig. ), while being an infant was negatively associated with earlier ACP (β -0.74, 95% CI -1.28 to -0.19, p = .008). Being diagnosed with a neurological and neuromuscular condition was associated with more days from the initial ACP to death (β 0.76, 95% CI 0.17 to 1.35, p = .012). Main findings To our knowledge, this is one of the first studies to examine the impact of SPC on children and young adults who died in a tertiary children’s hospital, with particular attention on ACP. This retrospective analysis compared the periods before and after palliative care implementation and those who received SPC and those who did not. The results demonstrated that patients who received SPC were more likely to have ACP and initiate discussions earlier. Furthermore, patients who received SPC were less likely to receive highly intensive care during the last month of life, including mechanical ventilation, CPR, and dialysis, and more likely to receive opioids. Initiation of advance care planning Facilitating ACP is a crucial role of palliative care, providing treatment and care in line with the values and preferences of patients and their families. Early initiation of discussions about the goals of care and routine revisiting of the care plan may improve patients’ and families’ experiences without increasing distress, strain, or emotional burden . We found that patients who received SPC were more likely to engage in ACP, complete medical and legal documents on life-sustaining treatments, and even engage earlier than those who did not. This is meaningful considering the policy in South Korea that limits legal documentation to patients expected to have imminent death, despite the difficulty in clearly defining “imminent death” . This trend to more frequent and earlier ACP was also observed in the post-period than in the pre-period. Our findings are consistent with a previous retrospective cohort study in the U.S., which investigated changes in ACP among children with cancer at the end of life through historical comparisons . Additionally, our study demonstrated an improvement in ACP during the post-period following the implementation of SPC, even in pediatric patients with conditions other than cancer. A possible explanation for this could be that the implementation of SPC within a healthcare institution not only improves the quality of care for patients and families enrolled in SPC but also enhances the culture within the entire institution, allowing a general palliative approach. Our finding of increased opportunities for early ACP among patients with SPC highlights the role of palliative care in setting the goals of care for patients and their families. Our results showed that ACP would likely be delayed if the patient was an infant. Previous qualitative research of NICU healthcare professionals reported that ACP was challenging owing to the uncertain prognosis of infants and various possible options for advanced medical treatment . Nonetheless, parents found that routine ACP, rather than a startling or desperate event, and standardized psychosocial support helped make end-of-life decisions for high-risk infants . Several factors could facilitate systematic early ACP, including designated personnel, professional awareness, and knowledge of ACP . Further research is needed to address how to standardize ACP for neonates and infants. Use of intensive care at the end-of-life Our results showed that patients with SPC received less intensive care such as less mechanical ventilation and CPR, and more opioids during the last month of life. These results were consistent with previous studies indicating higher inpatient service use among children with cancer, yet among them, patients with SPC received less intensive end-of-life care . However, patients with SPC had longer PICU stays and more PICU deaths in the present study, which is inconsistent with previous studies . In conjunction with our findings, recent studies have also presented mixed results regarding the impact of palliative care on end-of-life ICU utilization . Subsequent investigations are required to examine whether these observations indicate goal-concordant care or are influenced by systemic factors, such as the timing of palliative care referrals . Often, reduced use of acute healthcare services, such as fewer hospitalizations, fewer ICU admissions, and more home deaths, are considered quality indicators of hospice and palliative care . Child- and family-centered quality indicators of palliative care should be adopted, regardless of the location of care and/or death, including systematic care planning, expressive therapies , encouraging normalcy, and independence for adolescents . Furthermore, our results that ICU deaths were consistent regardless of palliative care involvement emphasize the importance of integrating palliative care into ICU settings. Palliative care involvement Although it is difficult to directly compare the referral rate owing to the varying roles of the SPC team in patient care, the SPC referral rate of 60% was relatively high within the wide range reported in the previous literature . Rather, the overall involvement of SPC was probably underestimated because patients who were discharged under hospice care or transferred to another children’s hospital before death were not included in our analysis. This study demonstrated that most children and young adult patients with malignancy received SPC, whereas those with premature and neonatal disease or cardiovascular disease did not. SPC involvement was less common in patients who died in the NICU and in younger patients, which was similar to a previous study . The limited utilization of SPC in the NICU could be attributed to various barriers, including the complex and uncertain nature of the diseases, the lack of education and awareness among healthcare professionals, and the lack of institutional policies . Our findings reveal missed opportunities to integrate palliative care into the NICU, as palliative care for neonates may benefit babies, parents, and healthcare professionals in pain and symptom management, decision-making and collaboration with parents, and psychological support . Strength and weaknesses This study contributes to our understanding of the demographic and clinical characteristics of children and young adults who died in the hospital across different age groups as well as diverse disease groups. Furthermore, by comparing in-hospital deaths before and after the implementation of SPC, our findings address the impact of SPC on the acculturation of the general palliative approach within the children’s hospital. This study has some limitations. First, the retrospective nature depends on complete documentation and accurate data retrieval, which makes it vulnerable, as it may depend on the provider and data collection process. However, this study utilized an automatic data-retrieval process to minimize these weaknesses. Second, patients who were cared for at the children’s hospital and died outside the hospital were excluded from the analyses due to limited access to the dataset. In the post-period, 59 patients died outside the hospital while maintaining SPC involvement. Therefore, our study has the potential to underestimate the impact of the SPC intervention on end-of-life healthcare utilization. Further studies are needed to investigate the effect of the SPC on end-of-life care, including deaths outside the hospital, to investigate the complete nature of SPC’s impacts on end-of-life care. Finally, this study investigated a single tertiary children’s hospital, which limits the generalizability of our findings to other institutions in different healthcare contexts. Given the substantial variation in the operations and structures of SPC programs across hospitals , the results from the robust SPC program at SNUCH may be challenging to generalize to settings with more limited PPC resources. Further multi-institutional prospective studies are required to validate the results of our study. To our knowledge, this is one of the first studies to examine the impact of SPC on children and young adults who died in a tertiary children’s hospital, with particular attention on ACP. This retrospective analysis compared the periods before and after palliative care implementation and those who received SPC and those who did not. The results demonstrated that patients who received SPC were more likely to have ACP and initiate discussions earlier. Furthermore, patients who received SPC were less likely to receive highly intensive care during the last month of life, including mechanical ventilation, CPR, and dialysis, and more likely to receive opioids. Facilitating ACP is a crucial role of palliative care, providing treatment and care in line with the values and preferences of patients and their families. Early initiation of discussions about the goals of care and routine revisiting of the care plan may improve patients’ and families’ experiences without increasing distress, strain, or emotional burden . We found that patients who received SPC were more likely to engage in ACP, complete medical and legal documents on life-sustaining treatments, and even engage earlier than those who did not. This is meaningful considering the policy in South Korea that limits legal documentation to patients expected to have imminent death, despite the difficulty in clearly defining “imminent death” . This trend to more frequent and earlier ACP was also observed in the post-period than in the pre-period. Our findings are consistent with a previous retrospective cohort study in the U.S., which investigated changes in ACP among children with cancer at the end of life through historical comparisons . Additionally, our study demonstrated an improvement in ACP during the post-period following the implementation of SPC, even in pediatric patients with conditions other than cancer. A possible explanation for this could be that the implementation of SPC within a healthcare institution not only improves the quality of care for patients and families enrolled in SPC but also enhances the culture within the entire institution, allowing a general palliative approach. Our finding of increased opportunities for early ACP among patients with SPC highlights the role of palliative care in setting the goals of care for patients and their families. Our results showed that ACP would likely be delayed if the patient was an infant. Previous qualitative research of NICU healthcare professionals reported that ACP was challenging owing to the uncertain prognosis of infants and various possible options for advanced medical treatment . Nonetheless, parents found that routine ACP, rather than a startling or desperate event, and standardized psychosocial support helped make end-of-life decisions for high-risk infants . Several factors could facilitate systematic early ACP, including designated personnel, professional awareness, and knowledge of ACP . Further research is needed to address how to standardize ACP for neonates and infants. Our results showed that patients with SPC received less intensive care such as less mechanical ventilation and CPR, and more opioids during the last month of life. These results were consistent with previous studies indicating higher inpatient service use among children with cancer, yet among them, patients with SPC received less intensive end-of-life care . However, patients with SPC had longer PICU stays and more PICU deaths in the present study, which is inconsistent with previous studies . In conjunction with our findings, recent studies have also presented mixed results regarding the impact of palliative care on end-of-life ICU utilization . Subsequent investigations are required to examine whether these observations indicate goal-concordant care or are influenced by systemic factors, such as the timing of palliative care referrals . Often, reduced use of acute healthcare services, such as fewer hospitalizations, fewer ICU admissions, and more home deaths, are considered quality indicators of hospice and palliative care . Child- and family-centered quality indicators of palliative care should be adopted, regardless of the location of care and/or death, including systematic care planning, expressive therapies , encouraging normalcy, and independence for adolescents . Furthermore, our results that ICU deaths were consistent regardless of palliative care involvement emphasize the importance of integrating palliative care into ICU settings. Although it is difficult to directly compare the referral rate owing to the varying roles of the SPC team in patient care, the SPC referral rate of 60% was relatively high within the wide range reported in the previous literature . Rather, the overall involvement of SPC was probably underestimated because patients who were discharged under hospice care or transferred to another children’s hospital before death were not included in our analysis. This study demonstrated that most children and young adult patients with malignancy received SPC, whereas those with premature and neonatal disease or cardiovascular disease did not. SPC involvement was less common in patients who died in the NICU and in younger patients, which was similar to a previous study . The limited utilization of SPC in the NICU could be attributed to various barriers, including the complex and uncertain nature of the diseases, the lack of education and awareness among healthcare professionals, and the lack of institutional policies . Our findings reveal missed opportunities to integrate palliative care into the NICU, as palliative care for neonates may benefit babies, parents, and healthcare professionals in pain and symptom management, decision-making and collaboration with parents, and psychological support . This study contributes to our understanding of the demographic and clinical characteristics of children and young adults who died in the hospital across different age groups as well as diverse disease groups. Furthermore, by comparing in-hospital deaths before and after the implementation of SPC, our findings address the impact of SPC on the acculturation of the general palliative approach within the children’s hospital. This study has some limitations. First, the retrospective nature depends on complete documentation and accurate data retrieval, which makes it vulnerable, as it may depend on the provider and data collection process. However, this study utilized an automatic data-retrieval process to minimize these weaknesses. Second, patients who were cared for at the children’s hospital and died outside the hospital were excluded from the analyses due to limited access to the dataset. In the post-period, 59 patients died outside the hospital while maintaining SPC involvement. Therefore, our study has the potential to underestimate the impact of the SPC intervention on end-of-life healthcare utilization. Further studies are needed to investigate the effect of the SPC on end-of-life care, including deaths outside the hospital, to investigate the complete nature of SPC’s impacts on end-of-life care. Finally, this study investigated a single tertiary children’s hospital, which limits the generalizability of our findings to other institutions in different healthcare contexts. Given the substantial variation in the operations and structures of SPC programs across hospitals , the results from the robust SPC program at SNUCH may be challenging to generalize to settings with more limited PPC resources. Further multi-institutional prospective studies are required to validate the results of our study. This study demonstrated that patients with SPC tended to discuss care plans more and earlier and received less intensive care and more opioids at the end-of-life. This indicates more preparation for end-of-life care and proactive symptom control that reflects the patient’s and their family’s values and preferences. SPC seems to benefit not only the recipients of palliative care but also the institutional culture, fostering more and earlier ACP for patients with or without palliative care support. More research is warranted to investigate barriers and facilitators to integrating specialized palliative care for infants in the initial stage of palliative care implementation. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Reliability and accuracy of artificial intelligence ChatGPT in providing information on ophthalmic diseases and management to patients | 5c7dd157-c4aa-490f-aee7-0dfdb63a0482 | 11076805 | Ophthalmology[mh] | Artificial intelligence chatbots are programs designed to simulate human conversation using natural language processing (NLP) and machine learning (ML) in order to process data to answer requests of all kinds . In recent years, they have become increasingly useful in various fields: marketing, education, customer service, information, health care, entertainment, and other industries. The use of chatbots, both in industry and for people’s personal use, is expected to continue to expand even further in the coming years . In November 2022, ChatGPT, a chatbot developed by OpenAI (San Francisco, CA, USA), was released for free online use. This software can process large amounts of text and continuously “learns”, iteratively teaching itself to perform natural language processing tasks very effectively. It was trained using Reinforcement Learning from Human Feedback (RLHF), and the current version operates based on an initial data input of 570 gigabytes (GB), or roughly 300 billion words . Since its release, ChatGPT has gained considerable popularity and is becoming an increasingly common way for people to seek information of all kinds online . ChatGPT is capable of performing a multitude of tasks, including responding to questions on a wide range of topics, coding in multiple programming languages, writing essays about virtually any subject, simplifying complex concepts, and writing songs, movie scripts, and poetry. It can produce text that is very difficult to distinguish from human-generated text. In a recent study, even medical researchers had difficulty distinguishing abstracts written by humans from those by ChatGPT . Indeed, reputable scientific journals have started to prohibit the use of ChatGPT to write scientific manuscripts; however, some do allow its use in manuscript editing because of its superb language capabilities . Currently, numerous studies are verifying the capabilities and possible applications of this chatbot, but at the moment, the reliability of the information provided by ChatGPT still needs to be validated. Seeking health information online is an increasing trend . A 2020 survey showed that 55% of Europeans aged 16–74 searched for health-related information online, with a 21% increase since 2010, and in the USA, the percentage went from 62.8% in 2008 to 74.7% in 2017, with an 11.9% increase . Considering this increasing trend and the growing popularity of ChatGPT, it would be reasonable to assume that people could soon utilize ChatGPT to ask health-related questions. For this reason, we sought to verify the validity of the information provided by ChatGPT in the medical field and, specifically, in the field of ophthalmology, with respect to certain topics. Moreover, we aimed to evaluate the extent to which this information might be incorrect, and even potentially dangerous to patients interrogating the chatbot independently. To assess the reliability and adequacy of the information related to medical knowledge provided by ChatGPT version 3.5, we submitted a standardized set of questions relevant to various eye diseases. The diseases were divided into 8 subspecialties: General Anterior segment and Cornea Glaucoma Neuro-ophthalmology Ocular Oncology Paediatric ophthalmology Oculoplastics Retina and Uveitis For each subspecialty, the 5 most common diseases available in the American Academy of Ophthalmology (AAO) section “For public & patients – Eye health A-Z” were selected (Table ). A new chat was initiated for questions related to each disease to limit learned patterns between assessments. The following three questions were asked in sequence: What is “X”? How is “X” diagnosed? How is “X” treated? X = name of the disease The answers provided by ChatGPT were collected and compared to the AAO section “For public & patients – Eye health A-Z” and to the latest AAO guidelines for patients. In the case of information beyond that is codified in the ophthalmological guidelines, or when information was limited on the AAO Eye Health A-Z pages, the graders referred to AAO Eyewiki pages. To confirm the more nebulous treatment options recommended by ChatGPT (and ensure that they were correct and not potentially harmful), the graders sought corroborating information in peer-reviewed publications. Two different authors (FC, KC), experts in the discipline, graded the answers separately using the following grading system: −3 or “Potentially dangerous”: At least one response which is incorrect and has the potential to cause harm to the patient’s health or well-being should the patient ultimately pursue such a recommendation, including invasive procedures or other interventions with potential for adverse sequelae which are not supported by the AAO for the disease in question. An example of this is ChatGPT suggesting biopsy as a diagnostic tool in retinoblastoma when in actuality this is not performed due to high risk of seeding and the procedure’s invasive nature. −2 or “Very poor”: At least two responses which are incorrect but do not have the potential to cause harm to the patient’s health or well-being should the patient ultimately pursue such a recommendation, including incorrect information regarding definition, diagnosis, or treatment which are not supported by the AAO for the disease in question. ChatGPT did not produce any responses which scored a −2. −1 or “Poor”: One response which is incorrect but does not have the potential to cause harm to the patient’s health or well-being should the patient ultimately pursue such a recommendation, including incorrect information regarding definition, diagnosis, or treatment which are not supported by the AAO for the disease in question. An example of this is ChatGPT suggesting x-ray as a potential diagnostic modality for entropion, which is not a diagnostic method supported by the AAO. 0: No response. ChatGPT gave responses for every prompt and therefore no scores of 0 were given. 1 or “Good”: Responses are correct but not complete per the AAO patient guidelines. An example of this is ChatGPT correctly stating that amblyopia treatment may include patching, corrective lenses, and surgical correction, but failing to mention atropine penalization of the stronger eye as a potential treatment method. 2 or “Very good”: All responses are correct and complete as per the AAO guidelines. An example of this is ChatGPT describing glaucoma as a group of conditions which affect the optic nerve with damage commonly related to elevated intraocular pressure, as well as distinguishing between primary open-angle glaucoma and less common types of glaucoma. 2* or “Excellent”: All responses within the AAO section “For public & patients - Eye health A-Z” are present. Additionally, there were supplemental responses which were deemed reasonable and helpful per trained ophthalmologists. An example of this is ChatGPT correctly identifying the diagnostic methods for scleritis given by the AAO patient guidelines (slit lamp exam, corroborating patient history, and systemic workup) but additionally mentioning fluorescein angiography, Schirmers test, and scleral biopsy in extreme cases as potential diagnostic methods. Responses which scored 2* were noted and reported within a separate category, but graded as 2 for statistical purposes as it was deemed appropriate to give all answers that were both correct and complete a maximum score. The rationale behind the design of a scale from -3 to 2 was to clearly delineate the responses from ChatGPT which contained any wrong information (all negative number grades) from responses that contained no incorrect information (all positive number grades). Intergrader variability was determined by assessing how many questions had a difference in score between the two graders (FC, KC). Whenever there was a score disparity between the two graders, an experienced third grader (JSP) assumed the role of arbiter and determined the final grade. The score disparities were deliberated amongst the three graders and the final score was based on unanimous agreement regarding the accuracy and safety scores of the ChatGPT-generated responses. Subjectivity bias was limited by isolating each individual statement within the ChatGPT responses and comparing to each sentence within the AAO guidelines in order to strictly adhere to the predefined score criteria. The assessment of the safety of each response inherently carried the greatest risk of bias, but the “potentially dangerous” scores characterized by the potential to cause adverse sequelae were discussed between the graders and agreed upon unanimously. The experience of each grader is as follows: FC has three years of clinical ophthalmology experience, KC has nine years of experience in ophthalmic clinical research, JSP has 41 years of clinical ophthalmology experience. The scores were tabulated and median and range were calculated within each question type within each and overall using Microsoft Excel Version 16.66.1 (Redmond, Washington, USA). For each individual question, a Kruskal-Wallis test was performed to evaluate the scoring differences within each subspecialty. Additionally, a Kruskal-Wallis test was performed to evaluate all scores given for each of the three question for significant differences. All questions were asked using ChatGPT Jan 9 Version. We considered a score ≥1 as optimal. There were no prompts which yielded no answer from ChatGPT, so there were no questions which were scored 0. Of the 120 questions processed by ChatGPT, 93 (77.5%) scored ≥1 meaning they were graded as “Good” or better, indicating the presence of at least some of the correct information provided by AAO and the absence of any incorrect information or recommendations that would cause harm to the patient’s health or well-being should they pursue such a suggestion. Of these correct answers, 35 were to the question “What is x?”, 31 to the question “How is x treated?” and 25 were to the question “How is x diagnosed?”. The total number of responses with a score of 2 or “Very good” (complete and correct) or 2* or “Excellent” (complete, correct, and providing more information than the AAO patient guidelines which was deemed beneficial), indicating the presence of all of the patient information provided by AAO and without any wrong or harmful information, was 74 (61.7%). Among these, 19 (15.8%) answers obtained a score of 2*, indicating the presence of all the correct information provided by AAO and additional correct information that may be useful to patients. One of the grade 2* answers was in response to “What is x?”, 10 in response to “How is x diagnosed?”, and 7 in response to “How is ‘x’ treated?”. There were 27 (22.5%) answers with a score ≤ −1, meaning that they were graded as “Bad” or worse. Among these, 9 (7.5%) obtained a score of −3 or “Potentially dangerous”, indicating a suggestion that includes unvalidated information that may cause unnecessary harm to a patient. The scores attributed to each answer are reported in Table , divided by subspecialty and the type of question. For each type of question, the median score was also calculated for each specialty (Figs. – ). The answers were all optimal in the General category, with median score of 2 for each question “what is x?”, “How is x diagnosed?”, and “How is x treated?”. The results from the other subspecialty categories were variable. The median scores in the Anterior Segment and Cornea subspecialty for each type of question were 2, −1, and 2, respectively. It is worth noting that the conditions ‘cataracts’ and ‘keratoconus’ obtained maximum scores. In the Glaucoma subspecialty, the median scores were 2, 0.5, and −1. Only the condition ‘glaucoma’ obtained maximum scores for each question. In Neuro-Ophthalmology, the median scores by type of question were 2, 2, and 2, respectively. The conditions ‘microvascular cranial nerve palsy’, ‘optic neuritis’, and ‘myasthenia gravis’ achieved maximum scores for each question. The greatest number of potentially harmful responses was found in the Ocular Oncology subspecialty, with 4 scores of −3 out of the 15 questions asked. In this section, the median scores were 1, 2, and 1, respectively. The condition ‘choroidal nevus’ achieved the maximum score for each question. The median scores in the Paediatric Ophthalmology subspecialty were 2, −1, and 1, respectively. No conditions from this subspecialty achieved the maximum score for all three questions. In the Oculoplastics subspecialty, the median scores were 1, 2, and 2, respectively. No conditions from this subspecialty achieved the maximum score for all three questions. In the subspecialty Retina and Uveitis the median scores were 2, 2, and 1, respectively. The conditions ‘diabetic retinopathy’ and ‘uveitis’ achieved the maximum score for each question. Kruskal-Wallis testing showed no significant difference in the score breakdown by subspecialty for “what is x?” ( p = 0.06), “how is x diagnosed?” ( p = 0.52), or “how is x treated?” ( p = 0.36). Kruskal–Wallis testing also showed no statistically significant difference when comparing all scores for each question ( p = 0.13). The overall median score amongst all subspecialties was highest for the question “What is x?” (2), followed by “How is x diagnosed?” (1.5) and finally, “How is x treated?” (1). Currently, many studies are investigating the potential of ChatGPT by assessing its performance in many different types of standardized and specialized tests. In a recent study conducted by Professor Christian Terwiesch at the Wharton School of the University of Pennsylvania, ChatGPT managed to pass the final MBA course exam . Moreover, in a recent study, ChatGPT also managed to score enough to pass the United States Medical Licensing Examination (USMLE) achieving >50% accuracy in all exams and more than 60% accuracy in most analyses . In our study, ChatGPT answered 77.5% of the questions correctly, and 61.7% were both correct and complete as per the AAO patient guidelines. We found that the median score was highest in the definition question (2), followed by the diagnosis question (1.5), and lastly, the treatment question (1). Furthermore, it is interesting to note how it performed much better in some subspecialties than in others. This median score difference could be attributed to several factors. First, the knowledge from which ChatGPT draws depends on the dataset of information from which it was trained. The definition of a common disease is usually standard and well-known, and thus the information the chatbot has received in its training regarding the definition of a disease should be very straightforward. When prompted about diagnosis and treatment, it is more likely that the inputs contained conflicting information. This hypothesis could also be applied to the difference in the median score we found in the various subspecialties. In the general subspecialty, ChatGPT answered all the questions correctly. We suppose this could be because the conditions from this category are more well-known pathologies; therefore, the chatbot may have had a higher amount and more consistent information from which to draw when it ‘learned’ about them. Supporting this, within other subspecialties, well-known and common pathologies such as cataracts, glaucoma, and diabetic retinopathy also obtained maximum scores. Recently, the landscape of artificial intelligence chatbots has diversified significantly with major players introducing their own innovations. Microsoft, for instance, has incorporated a chatbot utilizing a customized version of OpenAI’s GPT-4 into Bing . On March 21, 2023, Google unveiled Bard, powered by Language Model for Dialogue Applications (LaMDA). Unlike traditional models, LaMDA is a transformer-based neural language model, primarily pre-trained on dialogs from public conversations and web documents . In this rapidly progressing domain of large language models (LLMs) such as ChatGPT, we are witnessing the convergence of broader generic models and niche, domain-specific ones. For instance, the recent debut of MedPaLM hints at a future where artificial intelligence tools are intricately tailored for specific fields, including medicine . The introduction of newer versions of LLMs emphasizes the importance of meticulous and robust evaluation. As seen in our study with ChatGPT in the realm of ophthalmology, models can exhibit variations in accuracy based on disease familiarity and available training data. Robust evaluation is not merely about accuracy; it’s about consistency, reliability, and safety, especially in critical fields like medicine. As newer versions of LLMs emerge, evaluations should not only test their prowess in answering correctly but also in ensuring the absence of misinformation. Furthermore, the ability of LLMs to improve upon feedback and adapt makes continuous evaluation a necessity. Continuous evaluation will ensure that as LLMs evolve, they remain aligned with the highest standards of safety and accuracy. Limitations of this study were identified. The questions were asked only once, without rephrasing them or asking for clarifications, and the ChatGPT was not allowed to correct itself (often ChatGPT, as reported on the OpenAI website, if asked again on a topic, can answer correctly) . Moreover, the three questions were asked in sequence within the same chat, and this may have led to progressively more precise answers. Disease selection was also performed so that the 5 most common diseases from each subspecialty would be included to maximize relevance, for instance, purposefully selecting cataract. Doing so rather than grading all diseases in the database or selecting diseases at random introduces the possibility for selection bias. Additionally, our grading scale was comprised of an ordinal system rather than continuous or percentage-based variables. This affects the interpretation of the data as the intervals between each of the grades are not objectively equivalent. Although the grading system was clearly defined prior to initiating data collection, the subjective nature of graders reading ChatGPT’s responses and comparing them to the AAO guideline standard may introduce some inherent bias. From this study, it appears that artificial intelligence may be a valuable adjunct to patient education, but it is not sufficient without concomitant human medical supervision. On its own, it currently provides incomplete, incorrect, and potentially harmful information about common ophthalmic conditions. Further studies using different prompts and evaluation methods will be needed to better assess the accuracy of the information provided by ChatGPT in ophthalmology and other fields of medicine. As the use of chatbots increases, human medical supervision of the reliability and accuracy of the information they provide will be essential to ensure patient’s proper understanding of their disease and prevent any potential harm to the patient’s health or well-being. What was known before ChatGPT is an artificial intelligence chatbot that has gained considerable popularity online and is becoming an increasingly common way for people to seek information of all kinds online. Seeking health information online is a steadily increasing trend. It is likely that people will utilize ChatGPT to ask medical questions, but the quality of the answers is uncertain. What this study adds This study assesses 120 questions related to patient health information in Ophthalmology and compares it to the information for patients provided by the AAO. This study shows that ChatGPT is on the right track and has utility in patient education but currently, it is not sufficient without concomitant human medical supervision. ChatGPT is an artificial intelligence chatbot that has gained considerable popularity online and is becoming an increasingly common way for people to seek information of all kinds online. Seeking health information online is a steadily increasing trend. It is likely that people will utilize ChatGPT to ask medical questions, but the quality of the answers is uncertain. This study assesses 120 questions related to patient health information in Ophthalmology and compares it to the information for patients provided by the AAO. This study shows that ChatGPT is on the right track and has utility in patient education but currently, it is not sufficient without concomitant human medical supervision. Supplemental Table |
Optical mapping and optogenetics in cardiac electrophysiology research and therapy: a state-of-the-art review | caf6b503-e1fe-45fb-bce7-172d4e831065 | 10847904 | Physiology[mh] | Optical approaches for studying electrical function in the heart have fundamentally shaped our understanding of cardiac electrophysiology for 50 years ( Figure ). The origins of electrophysiology can be traced back to Galvani’s experiments in the 18th century, demonstrating intrinsic electrical activity generates muscle contraction in frog legs. Development of the capillary electrometer then allowed first recordings of cardiac electrical activity, leading to Einthoven’s refinement of the electrocardiogram (ECG). Microelectrode recordings from Purkinje fibres generated the first recorded cardiac action potentials. Subsequent single-cell techniques, such as voltage and patch-clamping, informed our understanding of distinct action potential phases and respective currents. However, the limited scalability and throughput prompted the development of multi-electrode arrays (MEAs), enabling measurement of electrical propagation. Nevertheless, MEAs only record extracellular potential and spatial resolution is constrained by electrode distance. Optical imaging overcomes these limitations in cardiac electrophysiological interrogation, offering unmatched spatiotemporal resolution. Cardiac optical mapping uses voltage and calcium (Ca 2+ )-sensitive dye to image multi-cellular preparations at high spatiotemporal resolutions. Cultured cardiomyocytes, induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs), engineered heart tissue (EHT), myocardial slices, isolated atria, and whole hearts have all been optically mapped. It has been crucial in deciphering the role of rotors in atrial fibrillation, the virtual electrode phenomena, and autonomic regulation of cardiac arrhythmias. Optical mapping is continually advancing with novel dye variants, improved hardware, motion tracking, and analysis tools. Optogenetics uses light to actuate transmembrane ion movement to modulate cardiac excitability in tissues expressing light-sensitive proteins called opsins. Optogenetic applications include precise control of pacing and arrhythmia induction or termination. Optogenetics provides huge potential for cardiac pacemaker development and defibrillation. Unlike chemical and electrical stimulation, optogenetics utilizes contactless, cell-selective pacing with minimal cytotoxicity. More recently, optical imaging and optogenetics have been combined to realize ‘all-optical’ electrophysiology, enabling precise control and measurement of cardiac function using light alone. , This review focuses on the applications, recent advances, and limitations of optogenetics, optical mapping and all-optical imaging systems for cardiac electrophysiology mechanistic research and translational applications. Optical mapping fluorescent sensors and illumination Optical mapping is a fluorescence-based technique that visualizes electrophysiological properties of multi-cellular preparations at unparalleled spatiotemporal resolution ( Figure ). This method involves the infusion of cardiac preparations, ranging from cellular monolayers to whole hearts, with voltage and/or Ca 2+ fluorescent sensors. Once these indicators are illuminated [e.g. by light-emitting diodes (LEDs)], the resulting fluorescence is captured by high-speed cameras. Consequently, optical recordings of cardiac action potentials, Ca 2+ transient morphology, and conduction are obtained that give crucial electrophysiological insights in health and disease. Established and novel optical mapping probes Voltage-sensitive dyes The coordinated generation and propagation of cardiac action potentials form the electrical basis of the heartbeat. For this reason, synthetic voltage-sensitive dyes are the most used in optical mapping. These dyes [e.g. di-4-aminonaphthylethenylpyridinium (di-4-ANEPPS)] respond to changes in voltage in the picosecond range, enabling accurate optical recording of surface cardiac electrophysiology. Voltage dyes are crucial tools for optical mapping; however, they are not without limitation. The fractional change in fluorescence output is low, which can generate low-quality signals. Moreover, most commonly used dyes are optimally excited by blue/green light, limiting penetration depth and promoting phototoxic tissue interactions. Further, synthetic dyes can have potential adverse effects on electrophysiology. For example, voltage-sensitive dye di-4-ANEPPS has demonstrated reduced spontaneous heart rate, sodium current, T-wave amplitude, and AV-node conduction in ex vivo and in vitro preparations. , Red-shifted dyes help overcome some of these issues. Di-4-ANBDQBS is a potentiometric dye with red excitation (660 nm) and near-infrared emission, for greater penetration depth. Di-4-ANBDQBS has been used to capture information from the endocardium with minimal effects on cardiac electrophysiology and cardiotoxicity. , Most synthetic sensors show adequate stability for acute electrophysiology investigation but exhibit signal decay over time due to photobleaching and dye leakage. Recently developed photo-electron transfer dyes, such as fluoVolt, show high photostability, rapid response time (pico- to nanoseconds), and high fractional changes. Additionally, high sensitivity makes these dyes better suited to two-photon imaging application, allowing greater penetration depth for transmural investigation. Calcium probes Ca 2+ couples the cardiomyocyte action potential to contraction. Imaging intracellular Ca 2+ ([Ca 2+ ] i ) for Ca 2+ transient recording is another significant application of cardiac optical mapping, often conducted simultaneously with voltage imaging. Ca 2+ probes typically consist of a fluorophore, chelator, and conjugator to quantify [Ca 2+ ] i . The most common of these are rhodamine (Rhod-2)-based probes such as rhod-2-AM, used to image cytosolic [Ca 2+ ] i . Recent advances have enabled organelle-specific [Ca 2+ ] i imaging. Valverde et al. recorded sarcoplasmic reticulum Ca 2+ transients alongside cytosolic Ca 2+ transients in isolated murine whole hearts using pulsed local-field fluorescence microscopy of mag-fluo-4 AM and rhod-2-AM. Trollinger et al. developed a novel technique to achieve mitochondrial-specific [Ca 2+ ] i measurement via a cold/warm rhod-2-AM loading protocol while simultaneously recording cytosolic [Ca 2+ ] i using fluo3. Several Ca 2+ indicators are also suitable for ratiometry, measuring the ratio of emission signals at different excitation wavelengths (‘excitation ratiometry’), as they exhibit wavelength-dependent fluorescence output. This enables more accurate quantification of absolute [Ca 2+ ] i in single-cell models and Ca 2+ amplitudes in whole heart optical mapping. Furthermore ratiometric dyes, including voltage-sensitive dyes, can significantly mitigate system noise and motion artefacts (see section). Genetically encoded voltage and calcium sensors Genetically encoded voltage and Ca 2+ sensors (GEVI/GECI) can achieve durable, cell-specific expression for long-term cardiac electrophysiology in vitro studies with reduced cytotoxicity, offering unique capabilities compared with synthetic indicators. However, their adoption is hindered by slower response times compared with ‘fast’ synthetic dyes and the necessity for genetic encoding. Different sensor classes offer distinct properties and advantages depending on intended use that cannot be fully explored here. Broyles et al. provide a comprehensive review of available dyes for optical mapping, while we have previously summarized dyes that are spectrally suitable for dual optical mapping and optogenetics . Optical mapping hardware Optical mapping systems integrate various sophisticated components to capture cardiac electrophysiology. Figure outlines key components, namely excitation sources, optical components, and high-speed camera(s). Common excitation sources include LEDs, tungsten–halogen lamps, mercury/xeon arc lamps, and lasers. Optical filters can narrow excitation wavelengths to avoid spectral cross-talk and effectively filter photons for imaging. In multi-parametric set-ups (see section), further filters are employed to deconvolve the fluorescent signals. Optical lenses are used to focus excitation and emission photons. Charge-coupled device (CCD) and complementary metal-oxide semiconductor (CMOS) cameras are most frequently used in optical mapping, and recent advances in CMOS technology have provided low-cost and portable imaging systems. , Key camera metrics include quantum efficiency, dynamic range, signal-to-noise ratio, sampling rate, and pixel size. A high quantum efficiency, defined as the ratio of photogenerated electrons to incoming photons within each pixel, enables sufficient signal-to-noise ratio even with low fractional changes. More recently, back-illuminated CMOS cameras and electron multiplying charge-coupled devices (EMCCD) have been introduced to reduce noise and amplify emission signal, respectively, for increased sensitivity. A key consideration in optical mapping set-ups is the balance between spatial and temporal resolution. High temporal resolution (>500 Hz) is needed to capture millisecond scale changes in voltage and/or [Ca 2+ ] i , while high spatial resolution (pixel size ≤ 100 s of microns) is required for detailed imaging of complex conduction patterns. Hardware constraints necessitate a trade-off between spatial and temporal resolution, where higher spatial resolution limits maximum sampling rate and vice versa. More sophisticated optical mapping designs include panoramic, often multi-camera, all-optical stimulation and imaging, capturing the entire surface topology. Rieger et al. implemented a customized LED light source with 294 optical fibres for panoramic optical mapping of mouse hearts, expressing GEVIs. Importantly, panoramic view was captured through two lenses directing optical emission bands onto a single CMOS camera, thereby surpassing logistical complications of a multi-camera set-up. Probes with higher quantum yield, such as fluoVolt, require fewer photons for adequate signal quality and are compatible with two-photon microscopy techniques, capable of capturing tissue at greater depth. Multi-photon, optical coherence tomography and light sheet fluorescence microscopy techniques have allowed between 400 μm and 4 mm depth ranges, capturing transmural activation. Future advancements are required to optimize camera sensor quantum efficiency at near-infrared wavelengths for three-dimensional (3D) reconstruction (Z axis profiling) and transmural optical imaging. Data analysis Short exposure times, small fluorescent changes, small pixel areas, and technical artefacts (e.g. motion and signal ‘blurring’ due to wavelength-dependent photon scattering ) all complicate processing and analysis of optical mapping data. Several approaches are applied to improve signal quality, including spatial and temporal filtering, temporal oversampling, and baseline correction. However, misapplication of these approaches (for example, by ‘over smoothing’ signals) can lead to misinterpretation, and the reader is directed to relevant literature that outline effective handling of optical mapping data. , Recent advances have seen the development of several open-source options for optical mapping data analysis. These include general all-purpose software and more specialized options for arrhythmia analysis, , panoramic imaging, conduction, and alternans. Further automation (e.g. machine learning–based approaches for automated artefact detection), combined with technical advances in minimizing post-processing needs, will further reduce the risk of misinterpretation of optical signals. Multi-parametric imaging set-up Key physiological insights can be gained by combining voltage and Ca 2+ imaging or other optically measurable parameters. Dual optical mapping is achieved by simultaneously exciting voltage and Ca 2+ dyes with distinguishable emission wavelengths. Rh237 and rhod-2 are well suited for dual voltage and Ca 2+ transient optical mapping set-ups, with similar peak excitation wavelength bands but distinct emission spectra. Furthermore, dual voltage and sarcoplasmic reticulum Ca 2+ mapping can be achieved using rh237 and fluo-5N AM dyes. Dual mapping enables voltage–Ca 2+ coupling analysis, including voltage–Ca 2+ latency, important for elucidating arrhythmic risk. Triple-parametric optical mapping, recording voltage, Ca 2+ , and autofluorescent nicotinamide adenine dinucleotide (NADH), has been performed to investigate metabolism–excitation–contraction coupling. Optical mapping is also compatible with other imaging techniques. Caldwell et al. measured cAMP activity and voltage simultaneously using a combined optical mapping and Förster resonance energy transfer (FRET) set-up in murine whole hearts. Reactive oxygen species, oxygen, and mitochondrial membrane potential optical probes can provide additional metabolic insights. Motion tracking The heart is a dynamic organ. This presents a problem for optical mapping as motion can distort signal morphology. Therefore, cardiac optical mapping is usually carried out on non-beating hearts, omitting physiologically important bidirectional electromechanical feedback and altering metabolic demand. The pharmacological electromechanical uncoupler blebbistatin selectively inhibits myosin II isoforms, abolishing contraction. Blebbistatin has been reported to alter cardiac physiology, increasing action potential durations and Ca 2+ transient upstroke rise times while reducing NADH autofluorescence in isolated murine hearts. However, the effect of blebbistatin on cardiac physiology is still disputed and contradictory findings may be due to species differences, blebbistatin concentration, or incorrect use (e.g. blebbistatin precipitation). , Motion tracking by computational signal correction or ratiometry has made it possible to optically map the freely beating heart. , Markers can be used to track and correct for movement. Motion correction can also be achieved without the use of fiducial markers, for example, by optical flow methods, which compute displacement vectors to quantify pixel movement and motion. , Christoph & Luther et al. showed up to a 80% decrease in motion artefacts when using marker-free motion tracking in optical mapping videos of contracting hearts. However, two-dimensional (2D) motion tracking exclusively captures movements along the horizontal and vertical axes, omitting movements along the depth axis, thereby limiting accuracy. Zhang et al. performed 3D marker-based motion tracking to measure epicardial strain and deformation. Recent studies have shown how graphical processing units can be utilized to accelerate the application of open-source motion correction algorithms, demonstrating real-time correction of optical signals. Ratiometry requires two signals, generated either during excitation or emission, which are similarly distorted by motion but differentially respond to, for example, voltage or Ca 2+ . The effects of motion on the time series signal, and artefacts due to uneven dye loading and illumination, can therefore be eradicated. However, ratiometry alone does not ensure spatial coupling, so can only be used in preparations with minimal dispersion. There are also inherent limitations, including reduced effective frame rates. Several studies have combined motion tracking, ratiometry techniques and tracking images within each spectral band for further reduction in motion artefacts. , Optical mapping of the freely beating heart is still a relatively specialized application; however, recent advancements will pave the way for innovative optical mapping investigations with enhanced physiological relevance. Optical mapping is a fluorescence-based technique that visualizes electrophysiological properties of multi-cellular preparations at unparalleled spatiotemporal resolution ( Figure ). This method involves the infusion of cardiac preparations, ranging from cellular monolayers to whole hearts, with voltage and/or Ca 2+ fluorescent sensors. Once these indicators are illuminated [e.g. by light-emitting diodes (LEDs)], the resulting fluorescence is captured by high-speed cameras. Consequently, optical recordings of cardiac action potentials, Ca 2+ transient morphology, and conduction are obtained that give crucial electrophysiological insights in health and disease. Voltage-sensitive dyes The coordinated generation and propagation of cardiac action potentials form the electrical basis of the heartbeat. For this reason, synthetic voltage-sensitive dyes are the most used in optical mapping. These dyes [e.g. di-4-aminonaphthylethenylpyridinium (di-4-ANEPPS)] respond to changes in voltage in the picosecond range, enabling accurate optical recording of surface cardiac electrophysiology. Voltage dyes are crucial tools for optical mapping; however, they are not without limitation. The fractional change in fluorescence output is low, which can generate low-quality signals. Moreover, most commonly used dyes are optimally excited by blue/green light, limiting penetration depth and promoting phototoxic tissue interactions. Further, synthetic dyes can have potential adverse effects on electrophysiology. For example, voltage-sensitive dye di-4-ANEPPS has demonstrated reduced spontaneous heart rate, sodium current, T-wave amplitude, and AV-node conduction in ex vivo and in vitro preparations. , Red-shifted dyes help overcome some of these issues. Di-4-ANBDQBS is a potentiometric dye with red excitation (660 nm) and near-infrared emission, for greater penetration depth. Di-4-ANBDQBS has been used to capture information from the endocardium with minimal effects on cardiac electrophysiology and cardiotoxicity. , Most synthetic sensors show adequate stability for acute electrophysiology investigation but exhibit signal decay over time due to photobleaching and dye leakage. Recently developed photo-electron transfer dyes, such as fluoVolt, show high photostability, rapid response time (pico- to nanoseconds), and high fractional changes. Additionally, high sensitivity makes these dyes better suited to two-photon imaging application, allowing greater penetration depth for transmural investigation. Calcium probes Ca 2+ couples the cardiomyocyte action potential to contraction. Imaging intracellular Ca 2+ ([Ca 2+ ] i ) for Ca 2+ transient recording is another significant application of cardiac optical mapping, often conducted simultaneously with voltage imaging. Ca 2+ probes typically consist of a fluorophore, chelator, and conjugator to quantify [Ca 2+ ] i . The most common of these are rhodamine (Rhod-2)-based probes such as rhod-2-AM, used to image cytosolic [Ca 2+ ] i . Recent advances have enabled organelle-specific [Ca 2+ ] i imaging. Valverde et al. recorded sarcoplasmic reticulum Ca 2+ transients alongside cytosolic Ca 2+ transients in isolated murine whole hearts using pulsed local-field fluorescence microscopy of mag-fluo-4 AM and rhod-2-AM. Trollinger et al. developed a novel technique to achieve mitochondrial-specific [Ca 2+ ] i measurement via a cold/warm rhod-2-AM loading protocol while simultaneously recording cytosolic [Ca 2+ ] i using fluo3. Several Ca 2+ indicators are also suitable for ratiometry, measuring the ratio of emission signals at different excitation wavelengths (‘excitation ratiometry’), as they exhibit wavelength-dependent fluorescence output. This enables more accurate quantification of absolute [Ca 2+ ] i in single-cell models and Ca 2+ amplitudes in whole heart optical mapping. Furthermore ratiometric dyes, including voltage-sensitive dyes, can significantly mitigate system noise and motion artefacts (see section). Genetically encoded voltage and calcium sensors Genetically encoded voltage and Ca 2+ sensors (GEVI/GECI) can achieve durable, cell-specific expression for long-term cardiac electrophysiology in vitro studies with reduced cytotoxicity, offering unique capabilities compared with synthetic indicators. However, their adoption is hindered by slower response times compared with ‘fast’ synthetic dyes and the necessity for genetic encoding. Different sensor classes offer distinct properties and advantages depending on intended use that cannot be fully explored here. Broyles et al. provide a comprehensive review of available dyes for optical mapping, while we have previously summarized dyes that are spectrally suitable for dual optical mapping and optogenetics . The coordinated generation and propagation of cardiac action potentials form the electrical basis of the heartbeat. For this reason, synthetic voltage-sensitive dyes are the most used in optical mapping. These dyes [e.g. di-4-aminonaphthylethenylpyridinium (di-4-ANEPPS)] respond to changes in voltage in the picosecond range, enabling accurate optical recording of surface cardiac electrophysiology. Voltage dyes are crucial tools for optical mapping; however, they are not without limitation. The fractional change in fluorescence output is low, which can generate low-quality signals. Moreover, most commonly used dyes are optimally excited by blue/green light, limiting penetration depth and promoting phototoxic tissue interactions. Further, synthetic dyes can have potential adverse effects on electrophysiology. For example, voltage-sensitive dye di-4-ANEPPS has demonstrated reduced spontaneous heart rate, sodium current, T-wave amplitude, and AV-node conduction in ex vivo and in vitro preparations. , Red-shifted dyes help overcome some of these issues. Di-4-ANBDQBS is a potentiometric dye with red excitation (660 nm) and near-infrared emission, for greater penetration depth. Di-4-ANBDQBS has been used to capture information from the endocardium with minimal effects on cardiac electrophysiology and cardiotoxicity. , Most synthetic sensors show adequate stability for acute electrophysiology investigation but exhibit signal decay over time due to photobleaching and dye leakage. Recently developed photo-electron transfer dyes, such as fluoVolt, show high photostability, rapid response time (pico- to nanoseconds), and high fractional changes. Additionally, high sensitivity makes these dyes better suited to two-photon imaging application, allowing greater penetration depth for transmural investigation. Ca 2+ couples the cardiomyocyte action potential to contraction. Imaging intracellular Ca 2+ ([Ca 2+ ] i ) for Ca 2+ transient recording is another significant application of cardiac optical mapping, often conducted simultaneously with voltage imaging. Ca 2+ probes typically consist of a fluorophore, chelator, and conjugator to quantify [Ca 2+ ] i . The most common of these are rhodamine (Rhod-2)-based probes such as rhod-2-AM, used to image cytosolic [Ca 2+ ] i . Recent advances have enabled organelle-specific [Ca 2+ ] i imaging. Valverde et al. recorded sarcoplasmic reticulum Ca 2+ transients alongside cytosolic Ca 2+ transients in isolated murine whole hearts using pulsed local-field fluorescence microscopy of mag-fluo-4 AM and rhod-2-AM. Trollinger et al. developed a novel technique to achieve mitochondrial-specific [Ca 2+ ] i measurement via a cold/warm rhod-2-AM loading protocol while simultaneously recording cytosolic [Ca 2+ ] i using fluo3. Several Ca 2+ indicators are also suitable for ratiometry, measuring the ratio of emission signals at different excitation wavelengths (‘excitation ratiometry’), as they exhibit wavelength-dependent fluorescence output. This enables more accurate quantification of absolute [Ca 2+ ] i in single-cell models and Ca 2+ amplitudes in whole heart optical mapping. Furthermore ratiometric dyes, including voltage-sensitive dyes, can significantly mitigate system noise and motion artefacts (see section). Genetically encoded voltage and Ca 2+ sensors (GEVI/GECI) can achieve durable, cell-specific expression for long-term cardiac electrophysiology in vitro studies with reduced cytotoxicity, offering unique capabilities compared with synthetic indicators. However, their adoption is hindered by slower response times compared with ‘fast’ synthetic dyes and the necessity for genetic encoding. Different sensor classes offer distinct properties and advantages depending on intended use that cannot be fully explored here. Broyles et al. provide a comprehensive review of available dyes for optical mapping, while we have previously summarized dyes that are spectrally suitable for dual optical mapping and optogenetics . Optical mapping systems integrate various sophisticated components to capture cardiac electrophysiology. Figure outlines key components, namely excitation sources, optical components, and high-speed camera(s). Common excitation sources include LEDs, tungsten–halogen lamps, mercury/xeon arc lamps, and lasers. Optical filters can narrow excitation wavelengths to avoid spectral cross-talk and effectively filter photons for imaging. In multi-parametric set-ups (see section), further filters are employed to deconvolve the fluorescent signals. Optical lenses are used to focus excitation and emission photons. Charge-coupled device (CCD) and complementary metal-oxide semiconductor (CMOS) cameras are most frequently used in optical mapping, and recent advances in CMOS technology have provided low-cost and portable imaging systems. , Key camera metrics include quantum efficiency, dynamic range, signal-to-noise ratio, sampling rate, and pixel size. A high quantum efficiency, defined as the ratio of photogenerated electrons to incoming photons within each pixel, enables sufficient signal-to-noise ratio even with low fractional changes. More recently, back-illuminated CMOS cameras and electron multiplying charge-coupled devices (EMCCD) have been introduced to reduce noise and amplify emission signal, respectively, for increased sensitivity. A key consideration in optical mapping set-ups is the balance between spatial and temporal resolution. High temporal resolution (>500 Hz) is needed to capture millisecond scale changes in voltage and/or [Ca 2+ ] i , while high spatial resolution (pixel size ≤ 100 s of microns) is required for detailed imaging of complex conduction patterns. Hardware constraints necessitate a trade-off between spatial and temporal resolution, where higher spatial resolution limits maximum sampling rate and vice versa. More sophisticated optical mapping designs include panoramic, often multi-camera, all-optical stimulation and imaging, capturing the entire surface topology. Rieger et al. implemented a customized LED light source with 294 optical fibres for panoramic optical mapping of mouse hearts, expressing GEVIs. Importantly, panoramic view was captured through two lenses directing optical emission bands onto a single CMOS camera, thereby surpassing logistical complications of a multi-camera set-up. Probes with higher quantum yield, such as fluoVolt, require fewer photons for adequate signal quality and are compatible with two-photon microscopy techniques, capable of capturing tissue at greater depth. Multi-photon, optical coherence tomography and light sheet fluorescence microscopy techniques have allowed between 400 μm and 4 mm depth ranges, capturing transmural activation. Future advancements are required to optimize camera sensor quantum efficiency at near-infrared wavelengths for three-dimensional (3D) reconstruction (Z axis profiling) and transmural optical imaging. Short exposure times, small fluorescent changes, small pixel areas, and technical artefacts (e.g. motion and signal ‘blurring’ due to wavelength-dependent photon scattering ) all complicate processing and analysis of optical mapping data. Several approaches are applied to improve signal quality, including spatial and temporal filtering, temporal oversampling, and baseline correction. However, misapplication of these approaches (for example, by ‘over smoothing’ signals) can lead to misinterpretation, and the reader is directed to relevant literature that outline effective handling of optical mapping data. , Recent advances have seen the development of several open-source options for optical mapping data analysis. These include general all-purpose software and more specialized options for arrhythmia analysis, , panoramic imaging, conduction, and alternans. Further automation (e.g. machine learning–based approaches for automated artefact detection), combined with technical advances in minimizing post-processing needs, will further reduce the risk of misinterpretation of optical signals. Key physiological insights can be gained by combining voltage and Ca 2+ imaging or other optically measurable parameters. Dual optical mapping is achieved by simultaneously exciting voltage and Ca 2+ dyes with distinguishable emission wavelengths. Rh237 and rhod-2 are well suited for dual voltage and Ca 2+ transient optical mapping set-ups, with similar peak excitation wavelength bands but distinct emission spectra. Furthermore, dual voltage and sarcoplasmic reticulum Ca 2+ mapping can be achieved using rh237 and fluo-5N AM dyes. Dual mapping enables voltage–Ca 2+ coupling analysis, including voltage–Ca 2+ latency, important for elucidating arrhythmic risk. Triple-parametric optical mapping, recording voltage, Ca 2+ , and autofluorescent nicotinamide adenine dinucleotide (NADH), has been performed to investigate metabolism–excitation–contraction coupling. Optical mapping is also compatible with other imaging techniques. Caldwell et al. measured cAMP activity and voltage simultaneously using a combined optical mapping and Förster resonance energy transfer (FRET) set-up in murine whole hearts. Reactive oxygen species, oxygen, and mitochondrial membrane potential optical probes can provide additional metabolic insights. The heart is a dynamic organ. This presents a problem for optical mapping as motion can distort signal morphology. Therefore, cardiac optical mapping is usually carried out on non-beating hearts, omitting physiologically important bidirectional electromechanical feedback and altering metabolic demand. The pharmacological electromechanical uncoupler blebbistatin selectively inhibits myosin II isoforms, abolishing contraction. Blebbistatin has been reported to alter cardiac physiology, increasing action potential durations and Ca 2+ transient upstroke rise times while reducing NADH autofluorescence in isolated murine hearts. However, the effect of blebbistatin on cardiac physiology is still disputed and contradictory findings may be due to species differences, blebbistatin concentration, or incorrect use (e.g. blebbistatin precipitation). , Motion tracking by computational signal correction or ratiometry has made it possible to optically map the freely beating heart. , Markers can be used to track and correct for movement. Motion correction can also be achieved without the use of fiducial markers, for example, by optical flow methods, which compute displacement vectors to quantify pixel movement and motion. , Christoph & Luther et al. showed up to a 80% decrease in motion artefacts when using marker-free motion tracking in optical mapping videos of contracting hearts. However, two-dimensional (2D) motion tracking exclusively captures movements along the horizontal and vertical axes, omitting movements along the depth axis, thereby limiting accuracy. Zhang et al. performed 3D marker-based motion tracking to measure epicardial strain and deformation. Recent studies have shown how graphical processing units can be utilized to accelerate the application of open-source motion correction algorithms, demonstrating real-time correction of optical signals. Ratiometry requires two signals, generated either during excitation or emission, which are similarly distorted by motion but differentially respond to, for example, voltage or Ca 2+ . The effects of motion on the time series signal, and artefacts due to uneven dye loading and illumination, can therefore be eradicated. However, ratiometry alone does not ensure spatial coupling, so can only be used in preparations with minimal dispersion. There are also inherent limitations, including reduced effective frame rates. Several studies have combined motion tracking, ratiometry techniques and tracking images within each spectral band for further reduction in motion artefacts. , Optical mapping of the freely beating heart is still a relatively specialized application; however, recent advancements will pave the way for innovative optical mapping investigations with enhanced physiological relevance. The advances outlined above have furthered the role of cardiac optical mapping as a central research tool, facilitating several important insights in cardiac electrophysiology and arrhythmogenesis, for example, the genetic basis for atrial fibrillation, and insights into the chamber specificity of anti-arrhythmics. , Human iPSC-CMs are an increasingly popular model for cardiac research and drug screening, capable of modelling patient-specific cell lines. The field is moving towards organoid EHTs, often utilizing human iPSC-CMs, to mimic cell–cell interactions for higher physiological relevance. However, chemical Ca 2+ dyes have been demonstrated to significantly impair blebbistatin efficacy in cardiac single-cell models, including human iPSC-CMs, increasing risk of motion artefacts. While microinjection has been suggested to reduce intracellular dye concentrations and subsequent adverse effects, this is not scalable. Mapping of beating human iPSC-CMs and EHTs using computational motion tracking has provided a solution, additionally capturing Ca 2+ –contraction coupling, for accurate mapping of these models. The advent of both red-shifted dyes (e.g. di-4-ANBDQBS) and higher quantum yield dyes (e.g. fluoVolt) has increased single wavelength penetration depths from 0.5–1 mm using standard (green–red) dye to ∼1–4 mm in cardiac tissue using ‘near-infrared optical mapping’. , Previously, studies used cardiac wedge preparations or myocardial slices to provide a 2D transmural surface , or two CCD cameras on either side of the myocardium in ventricular wall preparations. Mitrea et al. applied near-infrared di-4-ANBDQBS for improved transillumination to record signals from four different layers of the myocardial wall. Furthermore, longer wavelengths of near-infrared dyes demonstrate reduced absorption by blood, offering opportunity for in vivo cardiac optical mapping. Hansen et al. performed the first in vivo cardiac optical mapping using near-infrared di-4-ANBDQBS dye for successful activation mapping of the canine left atrium during sinus rhythm and fibrillation. More recent studies have used excitation ratiometry or mechanical stabilization to reduce motion artefacts during in vivo cardiac optical mapping , ; however, neither approach was sufficient for broad-area, accurate APD measurement. Advances in motion tracking algorithms, combined with red-shifted dyes and novel signal processing, have improved optical mapping capabilities for in vivo preparations. However, challenges remain including invasive surgery, limited optical view determined by the surgical thoracic window, and adverse effects of anaesthesia on cardiac electrophysiology. Multi-parametric optical imaging unveils novel sequences of events in excitation–contraction coupling in cardiac disease. Hypertrophic murine hearts demonstrated prolonged voltage–Ca 2+ latency, indicating altered electrical–contraction coupling. Dual optical mapping also captures cross-talk between electrical and Ca 2+ alternans, revealing positive or negative electromechanical coupling. Voltage and sarcoplasmic reticulum Ca 2+ signals simultaneously recorded in isolated rabbit hearts during ventricular fibrillation indicated ryanodine receptor refractoriness triggered sarcoplasmic reticulum Ca 2+ alternans, subsequently leading to electrical alternans. Triple-parametric optical mapping using Ca 2+ -sensitive rhod-2-AM, voltage-sensitive rh237, and endogenous NADH fluorophores revealed metabolic changes precede cardiac electrophysiology changes in murine ischaemic hearts. Recently, Caldwell et al. measured cAMP and electrical activity simultaneously, reporting higher apical phosphodiesterase activity in females vs. males which may contribute to sex-dependent electrophysiological variation. Mechanistic understanding from optical mapping Mapping of the freely beating heart has enabled excitation–contraction coupling and electromechanical feedback to be captured. Christoph et al. studied the spatiotemporal dynamics and topological organization of electrical and mechanical rotors in sinus rhythm and ventricular fibrillation. Three-dimensional mechanical scroll waves of contracting pig hearts were captured using combined panoramic optical mapping and four-dimensional (4D) ultrasound. Marker-free motion tracking for simultaneous optical mapping of voltage, [Ca 2+ ] i , and mechanical strain demonstrated cardiac tissue deformation was related to onset of electrical activation. Additionally, optical mapping of isolated, contracting pig hearts using excitation ratiometry demonstrated electromechanical decoupling during atrial fibrillation and highlighted Ca 2+ remodelling as an important mediator in early stages of atrial fibrillation. Combining optical mapping with multi-modal 3D structural analysis has facilitated significant advances in mechanistic understanding of pathophysiological conduction and arrhythmia. Optical voltage mapping and tissue clearing approaches have revealed neurocardiac and myofibre remodelling post-infarct unique to the border zone region, which contributes to heterogeneous conduction and therefore arrhythmia risk. Transmural near-infrared optical mapping has been combined with MRI, for improved clinical detection of re-entrant atrial fibrillation drivers. Optical action potentials were recorded over a depth of ∼4 mm to capture intramural microanatomic re-entry, commonly misinterpreted as focal patterns by ‘surface’ multi-electrode mapping. Combination with MRI assessment of intramural fibrosis enables delineation between re-entrant AF drivers and non-drivers, important for improving the efficacy of driver-targeted ablation. Optical recording of the transmembrane voltage has the unique advantage of not being corrupted by signal artefacts during electrical stimulation, for example, during defibrillation. Therefore, optical mapping has been fundamental in investigating the virtual electrode phenomenon, whereby an external stimulus polarizes the cardiac tissue, generating positive and negative virtual electrodes. Virtual electrode theory can be harnessed to better understand mechanisms of cardiac defibrillation, where the interaction between the shock-induced virtual electrodes and the ongoing electrical activity can terminate re-entrant circuits or, conversely, reinduce arrhythmia. Efimov et al. demonstrated optimal biphasic shocks achieved successful defibrillation in rabbit hearts without arrhythmia reinduction. Designing effective shock waveforms that capitalize on virtual electrode effects may improve defibrillation success and minimize potential side effects. Mapping of the freely beating heart has enabled excitation–contraction coupling and electromechanical feedback to be captured. Christoph et al. studied the spatiotemporal dynamics and topological organization of electrical and mechanical rotors in sinus rhythm and ventricular fibrillation. Three-dimensional mechanical scroll waves of contracting pig hearts were captured using combined panoramic optical mapping and four-dimensional (4D) ultrasound. Marker-free motion tracking for simultaneous optical mapping of voltage, [Ca 2+ ] i , and mechanical strain demonstrated cardiac tissue deformation was related to onset of electrical activation. Additionally, optical mapping of isolated, contracting pig hearts using excitation ratiometry demonstrated electromechanical decoupling during atrial fibrillation and highlighted Ca 2+ remodelling as an important mediator in early stages of atrial fibrillation. Combining optical mapping with multi-modal 3D structural analysis has facilitated significant advances in mechanistic understanding of pathophysiological conduction and arrhythmia. Optical voltage mapping and tissue clearing approaches have revealed neurocardiac and myofibre remodelling post-infarct unique to the border zone region, which contributes to heterogeneous conduction and therefore arrhythmia risk. Transmural near-infrared optical mapping has been combined with MRI, for improved clinical detection of re-entrant atrial fibrillation drivers. Optical action potentials were recorded over a depth of ∼4 mm to capture intramural microanatomic re-entry, commonly misinterpreted as focal patterns by ‘surface’ multi-electrode mapping. Combination with MRI assessment of intramural fibrosis enables delineation between re-entrant AF drivers and non-drivers, important for improving the efficacy of driver-targeted ablation. Optical recording of the transmembrane voltage has the unique advantage of not being corrupted by signal artefacts during electrical stimulation, for example, during defibrillation. Therefore, optical mapping has been fundamental in investigating the virtual electrode phenomenon, whereby an external stimulus polarizes the cardiac tissue, generating positive and negative virtual electrodes. Virtual electrode theory can be harnessed to better understand mechanisms of cardiac defibrillation, where the interaction between the shock-induced virtual electrodes and the ongoing electrical activity can terminate re-entrant circuits or, conversely, reinduce arrhythmia. Efimov et al. demonstrated optimal biphasic shocks achieved successful defibrillation in rabbit hearts without arrhythmia reinduction. Designing effective shock waveforms that capitalize on virtual electrode effects may improve defibrillation success and minimize potential side effects. Optogenetics, initially developed to study complex neuron interactions, has since been translated to cardiac applications with great effect. On wavelength-dependent illumination, opsins enable ionic transport across the cell membrane in a similar manner to voltage-gated, ligand-gated, or mechano-sensitive ion channels and pumps ( Figure ). Optogenetics has made possible several mechanistic insights and may potentially advance to clinical application. Opsin design and mechanism Channelrhodopsins The most popular family of opsins are channelrhodopsins (ChR). ChR2 is the most common of these, a low-selectivity cation channel with high photosensitivity at ∼480 nm (blue) excitation wavelengths enabling passive ion movement for dynamic stimulation and depolarization. Several studies have used ChR2 for pacing cardiac preparations. , , Since the first deployment of ChR2, several mutations have been genetically engineered. Chen et al. designed and validated ChRmine, a red-shifted excitation wavelength opsin, for non-invasive, in vivo optogenetic cardiac pacing in mice. Lin et al. engineered red-shifted opsin ReaChR with reduced tissue absorption and scattering, resulting in greater photocurrents. Additionally, other opsin variants with greater photocurrent conductance have been published, such as ChR2-H134R, ChR2-T159C and ChR-XXL, and CatCh. ChR2 containing a mitochondria-targeting sequence at its N-terminus has been shown to successfully reach the inner mitochondrial membrane in neonatal rat cardiomyocyte cells, to optically control mitochondrial membrane potential and ATP synthesis. Additionally, sensor co-expression has enabled dual functionality of optical stimulation and transduction signalling recording using FRET. Halorhodopsins, anion, and Kalium channelrhodopsins Conversely, opsins such as halorhodopsins drive anion transport. Halorhodopsins Natronomonas pharaonic halorhodopsin (NpHR) and Halobacterium salinarum halorhodopsin (HsHR) exhibit extracellular chloride affinity and pump chloride ions (Cl − ) into the cell upon light stimulation causing hyperpolarization. Anion ChRs such as GtARC1 enable Cl − conductance and can be used to depolarize cardiomyocytes or, through sustained depolarization, block re-excitation and inhibit cardiac action potentials, although this may be proarrhythmic due to sodium (Na + )/Ca 2+ overload. HcKCR1, a natural Kalium ChR (KCR), was recently discovered as the first ChR which selectively conducts potassium (K + ) over Na + ions, rapidly hyperpolarizing (<1 ms response time) murine cortical neurons. WiChR, another KCR, was recently expressed in human iPSC atrial cardiomyocytes and inhibited action potentials, reversibly suppressing spontaneous contraction during blue light illumination. Spectrally distinct depolarizing and hyperpolarizing opsins can be expressed and actuated in tandem, to enable bidirectional optical modulation. Limitations of opsins Cytotoxicity can be a concentration-dependent side effect of opsins. Additionally, after repetitive stimulation, several ChR variants can demonstrate desensitization, including reduced peak and plateau currents, thereby reducing opsin efficacy. Interval switching protocols or complete dark adaption of ChRs can prevent desensitization or eliminate bias, although may alter photocurrent. Further optimization of opsin safety, efficacy, and durability is required to expand utility of optogenetics, including potential clinical applications. Opsin delivery Key to the application of optogenetics is the expression of opsins by target cells/tissues. Model species, targeted cell type, vector genome size, duration of expression, and expression level required must all be considered for opsin delivery efficiency. The main methods of opsin delivery are summarized in Table . The model and method of opsin delivery must be compatible. For example, lentivirus undergoes host genome integration unlike adeno-associated virus, providing a compatible vector for mitotic and stem cell optogenetic models. The most widely used method is viral transduction using adeno-associated virus (AAV), offering rapid and consistent transduction with cell specificity (viral tropism). , Cardiac troponin T is a commonly selected promoter for cardiomyocyte-targeted opsin expression. Adeno-associated virus vectors display longer, more stable expression of opsins and minimal inflammatory response in post-mitotic cells, such as neurons or cardiomyocytes, compared with lentivirus. Self-complementary AAV vectors have recently been designed, eradicating the DNA synthesis step which AAVs normally undergo, increasing viral transduction speed. While non-viral opsin delivery methods including physical (electroporation) and chemical (liposomes, tandem-cell unit, nanoparticles, and transgenic models) approaches usually allow high stability of opsin expression, they demonstrate lower specificity and/or poorer clinical translatability. Illumination Direction and focus of light aids region-specific photostimulation, alongside genetic targeting. Most opsins are activated by shorter, ‘blue’ wavelengths, associated with reduced tissue penetration, providing only surface actuation. Therefore, transmural conductance and dyssynchrony cannot be optically modulated, despite established importance in arrhythmogenesis. , Additionally, studies have demonstrated the significance of depth of photostimulation to terminate re-entrant arrhythmias. , Strategies with deeper penetration capabilities have been explored, including ultra-thin injectable optoelectronic devices, flexible biocompatible membranes, and integration of opsins with X-ray excitable nanophosphors or upconversion nanoparticles. , Ultra-thin, waterproof micro-LED arrays on flexible substrate can be implanted in vivo to deliver cardiac illumination for transmural optical stimulation. Near-infrared wavelengths allow significantly greater tissue penetration and reduced light-induced cytotoxicity. Upconversion nanoparticles can transduce low-energy near-infrared photons to high-energy infrared, visible, or ultraviolet photons. Yu et al. demonstrated addition of all-trans-retinal (ATR) photosensitizer to ChR2-H134R-expressing cardiomyocytes in vitro could significantly increase ChR2 membrane expression and reduce optical pacing energy. However, techniques altering opsin spectral sensitivity and ATR treatment can compromise opsin kinetics and cardiac electrophysiology respectively. Unlike optical mapping where homogeneous illumination is preferred, optogenetics often employs directed illumination signals. Liquid crystal and digital micromirror devices, incorporating an array of microscopic mirrors, can acutely control LED-derived light impulses to narrow focus and increase spatial targeting. Moreover, light impulse patterns in optogenetics have shown to be important in arrhythmia termination studies. Standard optical sources, such as LEDs, are bulky and therefore must be positioned externally to any in vivo preparations. Multi-LED probes in the form of LED-chips have been implanted into the septum of ex vivo mouse hearts, expressing ChR2, and enabled stable optogenetic pacing including endocardial actuation. While this still requires a complex surgery and risk of complications, these set-ups provide higher clinical translatability. Channelrhodopsins The most popular family of opsins are channelrhodopsins (ChR). ChR2 is the most common of these, a low-selectivity cation channel with high photosensitivity at ∼480 nm (blue) excitation wavelengths enabling passive ion movement for dynamic stimulation and depolarization. Several studies have used ChR2 for pacing cardiac preparations. , , Since the first deployment of ChR2, several mutations have been genetically engineered. Chen et al. designed and validated ChRmine, a red-shifted excitation wavelength opsin, for non-invasive, in vivo optogenetic cardiac pacing in mice. Lin et al. engineered red-shifted opsin ReaChR with reduced tissue absorption and scattering, resulting in greater photocurrents. Additionally, other opsin variants with greater photocurrent conductance have been published, such as ChR2-H134R, ChR2-T159C and ChR-XXL, and CatCh. ChR2 containing a mitochondria-targeting sequence at its N-terminus has been shown to successfully reach the inner mitochondrial membrane in neonatal rat cardiomyocyte cells, to optically control mitochondrial membrane potential and ATP synthesis. Additionally, sensor co-expression has enabled dual functionality of optical stimulation and transduction signalling recording using FRET. Halorhodopsins, anion, and Kalium channelrhodopsins Conversely, opsins such as halorhodopsins drive anion transport. Halorhodopsins Natronomonas pharaonic halorhodopsin (NpHR) and Halobacterium salinarum halorhodopsin (HsHR) exhibit extracellular chloride affinity and pump chloride ions (Cl − ) into the cell upon light stimulation causing hyperpolarization. Anion ChRs such as GtARC1 enable Cl − conductance and can be used to depolarize cardiomyocytes or, through sustained depolarization, block re-excitation and inhibit cardiac action potentials, although this may be proarrhythmic due to sodium (Na + )/Ca 2+ overload. HcKCR1, a natural Kalium ChR (KCR), was recently discovered as the first ChR which selectively conducts potassium (K + ) over Na + ions, rapidly hyperpolarizing (<1 ms response time) murine cortical neurons. WiChR, another KCR, was recently expressed in human iPSC atrial cardiomyocytes and inhibited action potentials, reversibly suppressing spontaneous contraction during blue light illumination. Spectrally distinct depolarizing and hyperpolarizing opsins can be expressed and actuated in tandem, to enable bidirectional optical modulation. Limitations of opsins Cytotoxicity can be a concentration-dependent side effect of opsins. Additionally, after repetitive stimulation, several ChR variants can demonstrate desensitization, including reduced peak and plateau currents, thereby reducing opsin efficacy. Interval switching protocols or complete dark adaption of ChRs can prevent desensitization or eliminate bias, although may alter photocurrent. Further optimization of opsin safety, efficacy, and durability is required to expand utility of optogenetics, including potential clinical applications. The most popular family of opsins are channelrhodopsins (ChR). ChR2 is the most common of these, a low-selectivity cation channel with high photosensitivity at ∼480 nm (blue) excitation wavelengths enabling passive ion movement for dynamic stimulation and depolarization. Several studies have used ChR2 for pacing cardiac preparations. , , Since the first deployment of ChR2, several mutations have been genetically engineered. Chen et al. designed and validated ChRmine, a red-shifted excitation wavelength opsin, for non-invasive, in vivo optogenetic cardiac pacing in mice. Lin et al. engineered red-shifted opsin ReaChR with reduced tissue absorption and scattering, resulting in greater photocurrents. Additionally, other opsin variants with greater photocurrent conductance have been published, such as ChR2-H134R, ChR2-T159C and ChR-XXL, and CatCh. ChR2 containing a mitochondria-targeting sequence at its N-terminus has been shown to successfully reach the inner mitochondrial membrane in neonatal rat cardiomyocyte cells, to optically control mitochondrial membrane potential and ATP synthesis. Additionally, sensor co-expression has enabled dual functionality of optical stimulation and transduction signalling recording using FRET. Conversely, opsins such as halorhodopsins drive anion transport. Halorhodopsins Natronomonas pharaonic halorhodopsin (NpHR) and Halobacterium salinarum halorhodopsin (HsHR) exhibit extracellular chloride affinity and pump chloride ions (Cl − ) into the cell upon light stimulation causing hyperpolarization. Anion ChRs such as GtARC1 enable Cl − conductance and can be used to depolarize cardiomyocytes or, through sustained depolarization, block re-excitation and inhibit cardiac action potentials, although this may be proarrhythmic due to sodium (Na + )/Ca 2+ overload. HcKCR1, a natural Kalium ChR (KCR), was recently discovered as the first ChR which selectively conducts potassium (K + ) over Na + ions, rapidly hyperpolarizing (<1 ms response time) murine cortical neurons. WiChR, another KCR, was recently expressed in human iPSC atrial cardiomyocytes and inhibited action potentials, reversibly suppressing spontaneous contraction during blue light illumination. Spectrally distinct depolarizing and hyperpolarizing opsins can be expressed and actuated in tandem, to enable bidirectional optical modulation. Cytotoxicity can be a concentration-dependent side effect of opsins. Additionally, after repetitive stimulation, several ChR variants can demonstrate desensitization, including reduced peak and plateau currents, thereby reducing opsin efficacy. Interval switching protocols or complete dark adaption of ChRs can prevent desensitization or eliminate bias, although may alter photocurrent. Further optimization of opsin safety, efficacy, and durability is required to expand utility of optogenetics, including potential clinical applications. Key to the application of optogenetics is the expression of opsins by target cells/tissues. Model species, targeted cell type, vector genome size, duration of expression, and expression level required must all be considered for opsin delivery efficiency. The main methods of opsin delivery are summarized in Table . The model and method of opsin delivery must be compatible. For example, lentivirus undergoes host genome integration unlike adeno-associated virus, providing a compatible vector for mitotic and stem cell optogenetic models. The most widely used method is viral transduction using adeno-associated virus (AAV), offering rapid and consistent transduction with cell specificity (viral tropism). , Cardiac troponin T is a commonly selected promoter for cardiomyocyte-targeted opsin expression. Adeno-associated virus vectors display longer, more stable expression of opsins and minimal inflammatory response in post-mitotic cells, such as neurons or cardiomyocytes, compared with lentivirus. Self-complementary AAV vectors have recently been designed, eradicating the DNA synthesis step which AAVs normally undergo, increasing viral transduction speed. While non-viral opsin delivery methods including physical (electroporation) and chemical (liposomes, tandem-cell unit, nanoparticles, and transgenic models) approaches usually allow high stability of opsin expression, they demonstrate lower specificity and/or poorer clinical translatability. Direction and focus of light aids region-specific photostimulation, alongside genetic targeting. Most opsins are activated by shorter, ‘blue’ wavelengths, associated with reduced tissue penetration, providing only surface actuation. Therefore, transmural conductance and dyssynchrony cannot be optically modulated, despite established importance in arrhythmogenesis. , Additionally, studies have demonstrated the significance of depth of photostimulation to terminate re-entrant arrhythmias. , Strategies with deeper penetration capabilities have been explored, including ultra-thin injectable optoelectronic devices, flexible biocompatible membranes, and integration of opsins with X-ray excitable nanophosphors or upconversion nanoparticles. , Ultra-thin, waterproof micro-LED arrays on flexible substrate can be implanted in vivo to deliver cardiac illumination for transmural optical stimulation. Near-infrared wavelengths allow significantly greater tissue penetration and reduced light-induced cytotoxicity. Upconversion nanoparticles can transduce low-energy near-infrared photons to high-energy infrared, visible, or ultraviolet photons. Yu et al. demonstrated addition of all-trans-retinal (ATR) photosensitizer to ChR2-H134R-expressing cardiomyocytes in vitro could significantly increase ChR2 membrane expression and reduce optical pacing energy. However, techniques altering opsin spectral sensitivity and ATR treatment can compromise opsin kinetics and cardiac electrophysiology respectively. Unlike optical mapping where homogeneous illumination is preferred, optogenetics often employs directed illumination signals. Liquid crystal and digital micromirror devices, incorporating an array of microscopic mirrors, can acutely control LED-derived light impulses to narrow focus and increase spatial targeting. Moreover, light impulse patterns in optogenetics have shown to be important in arrhythmia termination studies. Standard optical sources, such as LEDs, are bulky and therefore must be positioned externally to any in vivo preparations. Multi-LED probes in the form of LED-chips have been implanted into the septum of ex vivo mouse hearts, expressing ChR2, and enabled stable optogenetic pacing including endocardial actuation. While this still requires a complex surgery and risk of complications, these set-ups provide higher clinical translatability. Optogenetics has revolutionized cardiovascular research, with Bruegmann et al. first demonstrating its application for in vivo murine heart pacing in 2010. Subsequently, continued advances have broadened optogenetics application to several domains. For example, red-shifted opsins (e.g. ChRmine) have enabled non-invasive, in vivo optogenetic cardiac pacing in freely moving mice wearing micro-LEDs. Implantable multi-LED devices delivering apical or transthoracic illumination with a closed chest have enabled in vivo arrhythmia termination in pathologically remodelled rat hearts, expressing ReaChR. , Patterned illumination techniques have been employed for precise regional stimulation of the heart. For example, Arrenberg et al. located and stimulated zebra fish cardiac pacemaker cells through altering light patterns and selective plane illumination. Additionally, acute photostimulation has been employed to dynamically alter action potential duration, ranging from depolarizing opsins at precise phases of the action potential to eliminate arrhythmia to complete silencing of action potential firing. Targeted delivery methods facilitating cell-specific opsin expression have paved the way for comprehensive electrophysiological characterization and studying interactions between different cardiac cell types. Zaglia et al. crossed double-floxed ChR2-tdTomato mice with connexin-40-Cre mice to induce cardiac conduction system-specific opsin expression and reported myocardial ectopy sites correlated with Purkinje fibre connection. Nussinovitch et al. demonstrated the ability of ChR2-expressing mouse embryonic fibroblasts to pace neonatal rat-derived cardiomyocytes in co-culture, in response to blue light flashes. Optogenetics has also offered a unique approach to study the cardiac autonomic nervous system. For instance, Moreno et al. demonstrated heart rate reduction via stimulation of ChR2-expressing intrinsic parasympathetic neurons in mouse hearts. This illustrates the potential for optogenetics in mimicking complex pathophysiology underlying cardiac disease. In an in vivo study, Rao et al. successfully paced rat hearts using upconversion nanoparticles embedded in flexible polydimethylsiloxane films attached to the ventricle. This approach effectively modified near-infrared light spectra for ChR2 activation. Furthermore, pacing efficiencies were comparable with blue light pulsed stimulation, presenting a promising non-invasive method of cardiac rhythm modulation. Finally, optogenetics plays a critical role in inducing, studying, and terminating arrhythmias. , For instance, Lemme et al. used optogenetic methods to pace 3D EHTs derived from human iPSC-CMs, simulating conditions like chronic tachypacing-induced cardiac dysfunction. Chronic optogenetic tachypacing correlated with greater vulnerability of EHTs to electrical burst pacing-induced tachycardia, as well as reduced action potential duration and effective refractory period. Further, optogenetic defibrillation with continuous blue light was successful in terminating arrhythmias induced by burst pacing. The application of cardiac all-optical imaging, combining optogenetics and optical mapping for acute control and sensing of cardiac electrophysiology, is expanding. Cells expressing opsins in conjunction with fluorescent dyes, or genetically encoded indicators for all-genetic delivery, enable contactless, cell-selective all-optical control and sensing. For an all-optical set-up, compatible opsins and sensors are essential, ensuring minimal spectral overlap. All-optical set-ups have been used for pacing and imaging human iPSC-CMs, , cardiomyocyte subpopulations, ventricular slices, and ex vivo hearts. Red-shifted dyes facilitate all-optical electrophysiology as they are spectrally distinct from ChR2. Optical mapping, using di-4-ANBDQBS, of optically paced ChR2-expressing rat hearts demonstrated reduced total ventricular activation time and improved homogeneity of ventricular depolarization vs. electrical pacing. These findings suggest that broad-area optical stimulation may offer an improved method for cardioversion therapy in patients with mechanical cardiac dyssynchrony. Although still far from clinical application, optogenetic pacing and defibrillation show promising translatability. The current gold standard for restoring sinus rhythm is defibrillation by electrical cardioversion, which is painful and not cell selective. Whilst effective, standard clinical pacemaker devices require complex surgery. One study reported 9.5% of cases show post-surgery complications within 6 months, such as pacing lead dislodgement and infection. Consequently, there has been a large research effort exploring optogenetic cell-selective and pain-free arrhythmia termination. Optoelectronic devices for cardiac pacing and defibrillation Optoelectronic devices provide opportunity for accurate, real-time modulation of cardiac electrical activity for ex vivo or in vivo set-ups, offering exciting research and future clinical applications. Miniaturization of optoelectronic devices has enabled previously unattainable in vivo application in rodents. Implanted, bioresorbable optoelectronic devices have achieved ex vivo concurrent cardiac pacing and electrophysiology measurement , ( Figure ) and in vivo acute and chronic cardiac pacing. , Recently engineered, transparent graphene electrode arrays were compatible with ex vivo cardiac optical sensing and stimulation, in addition to generating electrograms. , Xu et al. developed 3D elastic membranes to envelop the rabbit heart with flexible arrays overlaid, consisting of micro-LEDs for optical pacing and multi-functional sensors for pH, temperature, and strain measurement. Closed-loop all-optical pacing and modulation has been performed in real time (≤2 ms response time) to control electrical activity in ex vivo mouse hearts and restore sinus rhythm in abnormal conditions. Multi-functional optoelectronic devices may offer advanced cardiac diagnostic and treatment applications; however, feasibility of a closed-loop, implantable clinical pacemaker device may be limited due to the amount of spatiotemporal data to be processed in real time. These limitations however may be overcome by advancements in ‘on-chip’ computing technologies. Nevertheless, closed-loop optogenetics provides a promising research tool for previously unverifiable hypotheses, such as determining the level of discordance in alternans required to trigger arrhythmia. Pre-clinical cardiotoxicity drug screening All-optical interrogation could offer an improved pre-clinical cardiotoxicity screening platform for novel drugs. The current gold standard is electrode-based or patch-clamp recording to measure QT-prolongation and human ether-a-go-go-related gene channel inhibition, which provides limited information on conduction or contractility defects. A multi-parametric, all-optical set-up would enable multiple facets of cardiac physiology to be tested, including metabolism–excitation–contraction coupling, for more accurate cardiotoxicity predictions. , , Challenges for clinical application of optical methods A key barrier to clinical optogenetics is the safety and efficacy of opsin delivery and expression. Ongoing clinical trials are assessing the long-term safety of gene therapy to induce opsin expression, marking a crucial step in advancing this field. A considerable limitation of optogenetic-based therapy is light attenuation due to photon scattering and absorption, causing transmural gradients and potentially arrhythmia. Thus, implementing systems and opsins that allow greater penetration depth at safe irradiances must be prioritized. Advancing optoelectronic devices for clinical pacing and defibrillation also requires optimizing illumination strategies for successful and energy-efficient defibrillation of different heart rhythm pathologies using pre-clinical models. Device sensitivity for arrhythmia detection should be improved using artificial intelligence (AI) approaches, although developing and validating algorithms will require a large training data set. The long-term effects of optogenetic pacing on cardiac electrophysiology and structural remodelling must be investigated. Additionally, future studies should confirm whether optical pacemakers require less energy since they only target a subgroup of cells, which may improve longevity over standard clinical pacemakers. Combined near-infrared optical mapping and 3D functional imaging provides promising potential for improving arrhythmic driver-targeted catheter-based ablation . However, this technique requires further in vivo study before progressing to clinic. Additionally, while pre-clinical in vivo optical mapping studies have been performed successfully, inherent limitations including field of view, invasive surgery, and penetration depth remain. Similarly, although all-optical imaging provides a contactless, high-resolution research method for precise cardiac electrophysiology actuation and measurement, the logistical limitations of optical mapping imaging systems restrict clinical translatability. Optoelectronic devices provide opportunity for accurate, real-time modulation of cardiac electrical activity for ex vivo or in vivo set-ups, offering exciting research and future clinical applications. Miniaturization of optoelectronic devices has enabled previously unattainable in vivo application in rodents. Implanted, bioresorbable optoelectronic devices have achieved ex vivo concurrent cardiac pacing and electrophysiology measurement , ( Figure ) and in vivo acute and chronic cardiac pacing. , Recently engineered, transparent graphene electrode arrays were compatible with ex vivo cardiac optical sensing and stimulation, in addition to generating electrograms. , Xu et al. developed 3D elastic membranes to envelop the rabbit heart with flexible arrays overlaid, consisting of micro-LEDs for optical pacing and multi-functional sensors for pH, temperature, and strain measurement. Closed-loop all-optical pacing and modulation has been performed in real time (≤2 ms response time) to control electrical activity in ex vivo mouse hearts and restore sinus rhythm in abnormal conditions. Multi-functional optoelectronic devices may offer advanced cardiac diagnostic and treatment applications; however, feasibility of a closed-loop, implantable clinical pacemaker device may be limited due to the amount of spatiotemporal data to be processed in real time. These limitations however may be overcome by advancements in ‘on-chip’ computing technologies. Nevertheless, closed-loop optogenetics provides a promising research tool for previously unverifiable hypotheses, such as determining the level of discordance in alternans required to trigger arrhythmia. All-optical interrogation could offer an improved pre-clinical cardiotoxicity screening platform for novel drugs. The current gold standard is electrode-based or patch-clamp recording to measure QT-prolongation and human ether-a-go-go-related gene channel inhibition, which provides limited information on conduction or contractility defects. A multi-parametric, all-optical set-up would enable multiple facets of cardiac physiology to be tested, including metabolism–excitation–contraction coupling, for more accurate cardiotoxicity predictions. , , A key barrier to clinical optogenetics is the safety and efficacy of opsin delivery and expression. Ongoing clinical trials are assessing the long-term safety of gene therapy to induce opsin expression, marking a crucial step in advancing this field. A considerable limitation of optogenetic-based therapy is light attenuation due to photon scattering and absorption, causing transmural gradients and potentially arrhythmia. Thus, implementing systems and opsins that allow greater penetration depth at safe irradiances must be prioritized. Advancing optoelectronic devices for clinical pacing and defibrillation also requires optimizing illumination strategies for successful and energy-efficient defibrillation of different heart rhythm pathologies using pre-clinical models. Device sensitivity for arrhythmia detection should be improved using artificial intelligence (AI) approaches, although developing and validating algorithms will require a large training data set. The long-term effects of optogenetic pacing on cardiac electrophysiology and structural remodelling must be investigated. Additionally, future studies should confirm whether optical pacemakers require less energy since they only target a subgroup of cells, which may improve longevity over standard clinical pacemakers. Combined near-infrared optical mapping and 3D functional imaging provides promising potential for improving arrhythmic driver-targeted catheter-based ablation . However, this technique requires further in vivo study before progressing to clinic. Additionally, while pre-clinical in vivo optical mapping studies have been performed successfully, inherent limitations including field of view, invasive surgery, and penetration depth remain. Similarly, although all-optical imaging provides a contactless, high-resolution research method for precise cardiac electrophysiology actuation and measurement, the logistical limitations of optical mapping imaging systems restrict clinical translatability. Advanced computational approaches have been paramount in optical electrophysiology and are increasingly important for clinical translation. Deep learning predictive modelling has been implemented to compute phase maps and phase singularities in real-time using brief temporal sequences of electrical activity from optical maps, which may enable more accurate analysis of rotors and arrhythmia. Additionally, these models can predict future phase maps and phase singularity positions, targeting the ‘excitable gap’ as a low-energy method for arrhythmia termination. Computational optogenetic models have also provided insight into spatial targeting for successful arrhythmia termination. Models revealed light pulses should last longer than the arrhythmia cycle length or provide precise optogenetic spatial–temporal control to drag rotors to non-excitable boundaries to reliably terminate arrhythmia. Computational optogenetics has also been applied and validated to study opsin kinetics and dynamics, across a range of voltage and irradiance conditions, potentially accelerating the development of clinically suitable candidates. Optical mapping and optogenetics have undoubtedly revolutionized the field of cardiac electrophysiology research, offering unique insights into cell–cell interactions, arrhythmia development, and defibrillation strategies. All-optical imaging has facilitated contactless cardiac electrophysiology investigation, and implantable optoelectronic cardiac pacemakers could enable pain-free, cell-selective defibrillation. Advances in optical imaging technologies and tools, to improve penetration depth, alongside long-term safety studies will further enhance clinical applications of optical imaging. |
Pediatric Oncology Hospice: A Comprehensive Review | 63ce37d1-b172-4c78-898d-e9d03c237a57 | 11425979 | Pediatrics[mh] | The institution of modern hospice care (HC) dates back to 1967, when the first hospice clinic opened in London, England and the first US hospice home-based service was founded in 1974. These outstanding institutions were designed exclusively to serve adults with terminal and debilitating conditions with certain illnesses. This unique field of medicine has grown gradually, embracing a broader spectrum of patients, numerous specialized organizations, hospital wards, voluntary groups and allied health professionals. However, it took many years for the establishment of child-centered services, even in developed countries. The first North American free-standing pediatric hospice center was established in 1995, in British Columbia, Canada. This fact shows the lagging process of knowledge and practice transference from adult setting to pediatrics, considering the importance of pediatric hospice care (PHC) as an exclusive medical entity. Neoplastic disorders are among the leading causes of death in pediatric populations. , Accordingly, a majority of applicants for pediatric hospice includes those families affected by childhood cancers. There are some features in the group of children with cancer that makes them contrast with populations suffering from other life-threatening diseases, at least to some extent. Patients in this category commonly and chronically experience pain, somatic symptoms, disturbed mood, excessive fear and anxiety, lack of resilient, self-worth issues besides severe coping challenges due to their age-related limitations. Also, with a sick child who is dying of cancer all the other members of the family may suffer in a chronic way, resulting in a huge load of stress in their mutual life time or afterwards. In addition, cancer is a unique entity among the human pathologies, regarding the care approaches and outcomes. There is a divergent spectrum of malignancies with different, and sometimes unpredictable cure rates in pediatrics. The advent of novel techniques has changed the expectations of survival in the affected patients. This fact, increased the level of challenges and uncertainties in this field. Considering the complications and hurdles in pediatric hospice and childhood oncology the requirement of pediatric-plus-oncology care as a co-specialty in this common ground, makes establishing the facilities for these groups more challenging, compared to the care level provided to each setting, separately. Pediatric oncology hospice is generally a new terminology in current medical literature. Despite the fact that it involves a very wide spectrum of services, specialties and target populations, the infrastructures for providing adequate and efficient care are still debatable. Even the terminology to address the issue is not well-defined in medical literature. Besides several other aetiologies, these factors cause many end stage children with cancer to undergo frequent futile aggressive therapies, all over the world. Hence, they experience a considerable level of suffering in their last days, without adequate symptom control. Unfortunately, many countries around the world are deprived from such significantly vital frameworks. In fact, different geographic territories have a very diverse range for quality and quantity of hospice palliative care, globally. This defect stems from different underlying causes including economic, academic and cultural shortcomings. Also, governments have a big role in adopting these specialized care strategies in the national health plans. This observation is more true about certain locations in Africa, Asia , and South America. Lack of knowledge and experience in the form of education and goal-directed studies, can be a prominent etiology in this worldwide inadequacy. Otherwise, children have always been cherished and supported from almost all the officially accredited cultural, national, and global parties. For example, a systematic analysis reported a global funding for childhood cancer research around US$2 billion in the United States in less than a decade. Also, in China, provincial governments cover 50 to100% of medical expenses for pediatric cancer patients. It can be stated that the dimensions of a standard PHC have not been precisely defined, yet. - Complexity of the situation in pediatric oncology boosts the level of uncertainties for providing hospice care. In this review, we tried to gather, categorize, and utilize the available, but sparse and limited data about pediatric hospice with a focus on cancer-related hospice care. However, the results and discussions are widely applicable for PHC in other disabling conditions. Generally, any chronic terminal illness with similar features to cancer including preserved consciousness and pain perception, limited mobility, requirement for invasive procedures, age groups, bodily deformities, mental stress, having a defined illness trajectory and care expenses can be approached in the same way. , Serious infections, end stage heart and respiratory diseases, metabolic complications and genetic malformations are among these illnesses. However, the data for several of these entries are not as abundant as for cancer hospice. Our goals in this research are demonstrating the standard approaches, presenting the key parameters, emphasizing on the gaps and finally, developing a scientifically reliable groundwork for the future experiments and practice in pediatric oncology hospice. In order to achieve a well-founded data pool for our study, we explored the accredited scientific database, PubMed, with this search strategy: ((“Child”[Mesh]) AND (“Hospice Care”[Mesh])) AND “Neoplasms”[Mesh]. The word neoplasm was chosen as a standard MeSH term to embrace all the cancer-oriented terminologies in the database. Choosing child as the MeSH term, reliably gave us articles with pediatric-oriented topics and keywords, as well. Then the yielded journal articles were screened one by one, and our desired information was extracted from the most relevant entries. The categories provided in the following parts are derived from the context of the titles first, and then revising them after exploring the abstracts and full-texts. The articles without accessible full-text or those with foreign language text were backed up by extracted articles from Google Scholar within the same topics. The interchangeability of hospice care and palliative care in some of the studies has also been meticulously considered. Searching PubMed with the keywords “child,” “hospice care” and “neoplasms” provided 94 journal articles. Majority of those articles were included in this study and several additional ones were utilized for enhancement of the explanations and support of justifications. Countries of origin for the yielded articles from PubMed searching, based on the affiliations of corresponding authors are United States of America, United Kingdom, Canada, Australia, Germany, Italy, China, Republic of Korea, Japan, Taiwan, India, Greece, Turkey, France, Hong Kong, Denmark, Irland, Finland and Uganda, that confirms the international perspective of this review. Numerous studies have tried to explore different aspects of pediatric end-of-life (EOL) care in cancer settings. , According to the results from our database searching, the first noticeable point is the interchangeable usage of hospice and palliation in many articles. As stated by some institutions like Canadian Hospice Palliative Care Association, “both terms are used to refer to the same thing; however, people often use the term hospice care to describe care that is offered in the community rather than in hospitals.” Some references introduced hospice as a part of palliative care. Nevertheless, interchangeable use of these two terms with inattention cannot be endorsed. In other words, these two terms cannot be used interchangeably unless we are confident that our conclusions, statements or recommendations will have the same effect on both settings or the settings are combined (hospice/palliative care). The reason for this prohibition underlies in differences between the two settings. Although the definitions in different references are not consistent, the start or trigger points of the care, treatment types, place of care, prognosis and the insurance coverage are dissimilar for hospice and palliative care, despite their considerable overlaps. For instance, nowadays, hospice services are being provided in many hospitals, too, so it should not be considered as a merely home-based care. Based on definitions provided by American National Institute of Aging the most important differentiating factor is that in many regulatory jurisdictions curative approaches are not being implemented in hospice, while they are totally acceptable in palliative or hospice/palliative care. However, life-sustaining treatments are commonly utilized at hospice level of care. Therefore, local and situational definitions and concepts should be considered when discussing about this issue. Hospice services are now available for different groups of patients including those with cancers, AIDS, neurologic disorders, heart or lung diseases and severe rheumatologic complications. While approaches for pediatric oncology cases can be applicable for many other situations, specialized strategic planning might be needed for others like those with multiple complex chronic conditions. The amount of hospice experience for cancer, is generally more than other ailments. Unfortunately, many centers rarely reported their admissions for pediatric patients, over long periods of time that has resulted in a major loss of valuable data in this setting. Screening of the titles, abstracts and full-texts in the list of our references guided us to categorize the whole material into 9 specified categories. These parts are discussed separately under the following subheadings including general approaches, population specifications, role of parents and family, psychosocial issues, financial complications, locations of service, involved specialties, regulations, and quality improvement. The practical applicability and specifications of each hospice feature in pediatric oncology is summarized in . General Considerations for Pediatric Hospice Care There is adequate level of evidence in medical literature that proves the necessity of HC in oncology. It has been shown frequently that hospice philosophies are beneficial for the patients, caregivers and health care system. A majority of this evidence is derived from adult-based studies that compared outcomes of patient referrals to hospice services to the outcomes of staying in general or intensive care units. - As mentioned before, the boundaries and definitions of hospice/palliative care is not well defined. Hence, establishing clear-cut points in the continuum of concurrent care transitioning from palliation and curative measures into hospice and EOL care is essential for improving the quality of life and death in this specific population. However, physiologic, emotional, and social characteristics of children bear specific requirements that mandates exclusive services. Due to lack of exclusive experience in pediatric hospice, many of the currently implemented approaches are derived or copied from adult guidelines. In a study on a diverse range of advanced cancer patients, it was shown that lower age was associated with longer survival time in hospice setting. Such facts dictate a specialized planning for PHC. The first and most important role of hospice is changing “pain and distress” with “joy and closeness” for the patient and the family. Therefore, finding the earliest but most conservative time to start the service can be immensely beneficial for the patients, families, and healthcare system. - This can be accomplished through defining the indications or eligibility criteria, setting the priorities, addressing symptoms such as pain, nausea, vomiting, loss of appetite, respiratory failure, tiredness, anxiety, or depression, and decision-making about location and level of care, quality standards, modified expectations, and logical and legal issues. Demography Knowing the statistics and demography of children who used PHC versus non-users can be helpful in long-term policy making for improving the establishment of this service as a quality-of-life-saving strategy. It is notable that some minorities might have lower knowledge or accessibility to these services. In a study in USA from 2011 to 2013, factors like non-Hispanic White race, complex chronic comorbidities, mental issues, technology dependence, specific tumors, and residence in rural areas with lower socioeconomic environments were considered as hindrance factors to hospice admissions. In another study in 2013, it was shown that race, ethnicity, payor status, patient diagnosis, and religion are significantly associated with hospice enrollment and its outcome. Cancer type, hospital size, insurance status and geographic location are other defined factors in this field. Considering the value of home-based care in pediatric hospice, some predictive demographics can be found among the above-mentioned factors to forecast the risk for hospital death. While data in this regard for pediatric oncology hospice is very limited, information from other child-oriented hospice settings can be useful. Age lower than a year, presence of congenital causes, certain locations, proximity to tertiary hospitals are associated with dying in hospital. Focusing on pediatric intensive care unit (PICU) death rates among children with cancer, risk factors such as Hispanic ethnicity, hematologic malignancies, hematopoietic stem cell transplantation, cumulative number of PICU admissions, receiving cancer therapies during the final month, and delayed palliative care involvement have been discovered. In our data pool for this review, we detected a variety of demographic features for pediatric cancer patients under hospice care. The information derived from the references included data for age groups (neonates, children, and young adults), , , - disease type and conditions, , , , , , - types and levels of medical interventions, , , , - access to specialized care and sufficient communication, - geographical location, , , , - physical, mental and behavioural comorbidities, , , , gender, , , , previous hospitalization unit or ward, , , , health insurance type, , , , clinical phase (stable, unstable, deteriorating, terminal), , cost of care, , race and ethnicity, , , financial background, , familial characteristics and siblings, cultural background, , number of admissions, access to palliative care before hospice, , symptomatology, , , and access to general hospitals. It should be noted that some of the studies included non-pediatric-cancer populations, as well. Some of these parameters were associated with different outcomes and success rates of hospice services, which are discussed in different sections of this review. Based on the findings it can be assumed that middle categories of age, non-hematologic malignancies at lower stages, receiving fewer intensive therapies, absence of comorbidities, access to insurance coverage, advanced care and education, less frequent history of admissions and utilization of palliative care are associated with better outcomes regarding the EOL quality of life (QOL) and hospice attendance or success. Parents and Family Members Parents are major determinants for initiation, process, place, and quality of hospice care in pediatric settings. In such complicated situations, parents are the first-line legal authorities to adopt this kind of lifestyle for their loved ones. This produces is a huge burden of mental stress for them that can threaten their efficient critical-thinking and proper decision making. It has been shown that these parents, suffer more from psychiatric disorders, including anxiety and depression compared to general population. These guardians have to plan their caregiving strategy, deal with financial issues, and handle the familial and social complications of entering a hospice-associated life. This fact is even more complicated among the families that have children with cancer, both due to debilitating expenses and bombardment with every-day advertised experimental and uncertain therapeutic options. One of the greatest challenges for parents is digesting the dilemma, when their decision is not in accordance with their dying child; this makes it even more difficult for the healthcare team to reach a certain opinion, as well. These unwanted and unplanned consequences, plus feelings of regret, continue even after their great loss, and can be accompanied by prolonged grief and disturbed QOL. The role of hospice in family care is to ease the pressure and remove this burden. Taking care of parents of children with terminal illnesses is as important as paying attention to the patients. After the diagnosis, there is a huge risk for total collapse of personal, financial, and mental health of a family dealing with the impending death of a younger family member. The most common aspects of supportive needs among parents have financial, psychosocial, and legal sources. Besides the patients and their parents, other family members deserve to be acknowledged in this process. Young brothers and sisters are priorities in this regard due to their vulnerable situation and minor authoritative role in this family hazard. A literature review by Ridley and Frache demonstrated that there is an alarming gap in coverage of care for youngster siblings of deceased children. Several approaches have been developed in cancer hospice settings, due to their higher prevalence. Group sessions, recreational camps and educational activities have been tried for bereaving children with promising results, but availability of such facilities has not been globally established. In addition, the role of young siblings as assisting caregivers in hospice has been neglected, almost completely. It can be concluded that concordant sponsorships from medical, social, and governmental entities are required to comprehensively support families and decrease the ravaging effects of this issue on parents and other family member. Psychosocial Issues Psychological and social aspects of PHC are as important as its somatic features. The process of providing PHC starts with this portion of care. In recent years there has been a major change from the state of disappointing expectations from cure-oriented thinking that may have a destructive effect on the EOL management procedure, to optimism in a variety of available options for improving the near-death trajectory of a child and the family. This mindset will boost their mental and spiritual strength, significantly. After building a clear and honest bond, hospice systems should provide the most peaceful environment to control and ease the understanding of the end and grief for parents, siblings and caregivers. Even the decisions about the array of medical interventions in this phase should be finalized based on the emotional and relational responses from the care recipients. This issue is very critical in cancer settings in which lots of curative and palliative interventions are available. Then they should provide bereavement aid including emotional support, recognition of abnormal and pathologic responses, and follow-ups. These maneuvers generally target parents or caregivers, rather than their young patients. However, social interactions of a dying child should be addressed as a very complicated issue, as well. In order to plan effectively, determinants of psychological risks should be defined. Studies have shown that besides a rewarding and meaningfulness feeling, caregiving is commonly accompanied by adversities like anxiety, depression, intensified grief reactions, physical and mental health dysfunction and even, increased caregiver mortality compared to general population of parents or those with children affected by non-terminal illnesses. With the use of indicators such as relationship with the patient, age at caregiving, accessibility of help, financial support, and history of loss, high risk entities can be more feasibly detected and supported through psychosocial back up programs in pediatric cancer hospice. Reports from UK childhood cancer centers shows that there are multiple areas for providing these kinds of supportive measures but generally no firm standard has been established to make them continuously and comprehensively accessible. These areas include education, staffing, facilities for children and families, and psychosocial services in terms of group activities and transition plans. In this way, presence of psychiatrists and psychologists can be undeniably beneficial regarding they role in implementation of specialized assessments, diagnosis, interventions and research, both for the child and the family. , However, due to lack of extensive experience in this field the true value of these collaborations has not been fully elucidated in PHC. In addition, death education, initiated in schools can have a preventive effect on mental health disorders for families who will expose to such tragedies, unexpectedly. Continuing this type of education through the society with meticulously supervised methods can act as a public immunization and enhance the society with self-care tools to decrease the devastating consequences, from before entrance to the trajectory of a child’s death. At the end it should be stated that this harsh pathway is more impassable to go through alone. Therefore, the value of mentoring and support groups from those who had the same experience cannot be overemphasized. In this pathway, instead of being considered as hopeless, the patient and the family will be hopeful for spending nice and memorable days together. Families will enjoy the memories of those valuable last days to harness their bitter bereavement. Also, the role of faith which has been introduced in adult settings, should be further explored in PHC, as well. Budgeting and Insurance EOL care is associated with a huge financial burden, both for the involved family and the supporting heath care system. These budgeting issues are powerful to the extent that they can change the pathway of care for a dying child. Decreasing the amount of expenses for pointless therapies or unnecessary admissions is hidden in the philosophy of PHC to facilitate the passage of the family through this harsh course. It has been shown in adult settings that implementation of HC in the form of home based care is significantly more cost-saving than hospital-located conventional care. However, the hospice process can be very expensive by itself. Therefore, financial pre-planning for critical situations is mandatory. Logically, families cannot actively be involved in the investment procedures for these plans, especially if the time period between the diagnosis and EOL situation is short. EOL insurance is not a common or routine investment among families who cannot predict such unexpected grave occasions. Also, decision making in this situation about the EOL plan can be immensely difficult. Hence, governmental authorities, social organizations, insurance companies and finally health care systems have to be prepared for providing this kind of support. Regarding the fact that PHC is still a newborn science and entangled with the multi-aspect complications of pediatricEOL services, the rubrics of budgeting and financial planning for it has not been fully developed in many countries. However, there are a limited numbers of studies that tried to explore this new field and propose guidelines. The first step for establishment of an effective financial plan is recognition of the sources of cost. This can be achieved through cost-evaluation studies. Such studies have proposed that the intensity of treatment protocols, physician care, nursing care and presence of some clinical symptoms including asthenia, anorexia, bedsore, nausea and vomiting, can result in rising costs. There is a considerable risk that governments do not have the capacity to provide for all the hospice costs. In this way, the role of support from charities will be vital. Nowadays, a considerable amount of national hospice expenses in countries like England and Wales is provided by charities for PHC. However, this might persuade the government to step back from their true supportive stands. Additionally, it should be noted that governments may devote a considerable budget to HC but recognition of a suitable share for PHC needs more organized amendments. Insurance organizations play a great role in handling of PHC costs. Insurance coverage can be considered as a determinant in choosing the strategic style for the EOL care. Generally, public governmental insurance plans cover these types of expenses. Therefore, those patients who are covered only by the private ones might be deprived of such services. There are territory-based local eligibility criteria for utilization of hospice insurances. For example a certificate from an accredited practitioner to announce that the patient is within the terminal 6 months is mandatory, in some settings. All these diversities in coverages can affect the delivery of PHC. Different jurisdictions have different insurance coverage policies for hospice from zero to 100% of the expenses. , Also, there is a difference between coverage for hospice and palliative care in some places and the policies are evolving. For example, in an American setting, palliative care is paid by personal insurances, involving limitations such as standard co-pay and deductions. A partial coverage is also provided by Medicaid and Medicare. But for hospice, expenses are fully paid by Medicaid or Medicare and coverage includes medications, medical equipment, nursing care, social and spiritual services. The same is true with certain personal insurances. The Place for Death One of the key features of PHC is the place in which the young patient passes away. This issue has been the focus of a considerable portion of studies. In fact, the matter of location for spending the last days can be the leading determinant for planning the strategy of HC. HPC services can be delivered at different locational or strategic points. Supervised home care is one of the most appealing approaches. Specialized hospice wards are also available in limited numbers that provide a combination of medical and spiritual care. In resource limited situations, as a last resort, regular, pre-critical, or critical care units can be adjusted, as much as possible, to accept the young patients for this purpose. In addition, medical procedures are dispensed at different levels of intensity. Serving the patient and family with hospice alone is one option. Due to scarcity and costly nature of these services in some places, simple private nursing care is substituted on some occasions, generally with low satisfaction rates. Another option for the pediatric patients is admission in “concurrent” HC. Although, concurrent palliation is a better terminology in this regard, due to continuation of therapeutic and curative measures. In this setting the patient can be benefitted from both life prolonging care and pain reducing hospice. Because this approach is more resource consuming, developing certain criteria for its recipients is fundamental. The decision about where a patient passes away is affected by numerous factors from different entities of this challenge. Each one of the patient, family and medical team may have a divergent opinion about it. The usual available options include home, general wards, critical care or emergency settings, general hospice, and pediatric hospice facilities. For children with cancer, oncology wards are also a common place to die. However, accessibility to these choices varies in different geographical zones. In addition, continuation of hospitalization in the regular oncology or pediatric wards cannot be recommended. It has been shown that home is the most appealing place for the ending days, both from parents and clinicians’ point of view. Statistical data shows that this happens for a majority of patients under PHC, as well. Although, the actual and the preferred location can be incongruent, specifically for children with haematologic malignancies. On the other hand, presence of an active palliative hospice care can increase the congruency. Notably, in some cases, parents showed more tendency to choose hospital-base care compared to specialized PHC institutions, compared to clinicians. This implies the lack of adequate introduction and popularization of PHC to the public audience. Establishment of HC teams in healthcare centers has pushed the curve toward home-based or hospice-based admissions. Studies demonstrated that cultural, racial, and clinical backgrounds of patients can alter the aforementioned assumptions. For instance, in study in UK that categorized the destinations into hospital and hospice or home, it was shown ethnic minorities were less likely to die in hospice, while a lower portion of White children died in the hospital setting. It should be noted that the optimal choice for the final place can be totally different case-by-case. But based on the aforementioned studies that considering home-death as a preferrable outcome, some arrangements highly improve the QOL and feasibility of this type of care for the involved families. The term “shared care” emphasizes a collaboration of family and the medical team during this period. A wise approach would be educating the household members to handle the major requirements of caring to stay independent from frequent professional visits. The share of healthcare team in this situation is more educational, rather than therapeutic. This method both save the financial status of the bereaving family and spare the limited number of trained staff to attend more cases. In addition, the number of unnecessary admissions and hospital deaths can be diminished through such a goal-directed programming. Insurance policies and governmental financial strategies can also significantly change the trends in this regard. It should be kept in mind that providing hospice services is accompanied by environmental, cultural, economic and religious issues that need preplanning. Also, it is notable that some authors have argued against considering hospice as a location of care and they defined it as a philosophy of compassionate, multidisciplinary care that can be delivered at any place. Specialists As mentioned frequently in the previous parts, PHC is a teamwork, and it will be handled optimally if the whole team be present with adequate capabilities, expertise and numbers. Besides their pragmatic function, the accessibility to this team will be highly comforting for both the child and the stressed caregivers involved. It should be kept in mind that there is no specific hierarchical order for roles and duties in this procedure and the importance for all of the elements are totally overlapping. Considering the majority of studies in this field, nurses are the mainstays of this procedure for two reasons. First, their professional dexterity in handling every-day needs of hospice children makes them the frontiers of PHC team. Secondly, they are health practitioners that provide care at a high professional level with considerably longer continuation, due to the nature of their occupational characteristics and history of trainings. In this way, their persistent active presence will help a lot in management of a near death child and the family. On the other hand, lack of adequate nursing staff or availability of home visits significantly decrease the quality of lives of the challenged household in that terminal hectic stage. Physicians are also critical in easing the last days with helping to relieve the pain, giving medical consultations and guiding the caregivers for an optimized decision making. Palliative and hospice care specialists are valuable agents in PHC, but regarding their scarcity, most centers lack the access to such ideally trained professionals. Considering the fact that majority of children on hospice settings are transferred from specialized care services, usually the in-charge physicians are doctors with specialties and subspecialities. Oncology fields are the major referring wards for pediatric hospice; therefore, oncologists have the highest exposure and experience with these cases. Stem cell transplant specialists also are in contact with a big population of children with rare and terminal diseases. However, pediatricians and oncology pediatricians are the most eligible clinical entities in this regard, with deep insight into somatic, psychological, and social features of PHC. Other practitioners like surgeons and anaesthesiologists also have positive effects. In addition, the presence of mental health clinicians such as psychologists and psychiatrists are non-negotiable in preparing the child and family for their destiny. An optimal example in this regard is an organized team involving pediatric oncologists, palliative care physicians, HPC trained nurses, psychologists, child life specialists, and medical social workers. Also, pharmacists can improve the quality of medication management in pediatric EOL care. The last but not least group involves social workers, spiritualists, lawyers, and other non-clinical support coordinators to help the parents navigate socio-economic-legal challenges. This list can be elongated with more clinicians and non-clinical supporters but considering the sensitivity of this issue, empathy and moral devotion are the main requirements of being a team member in hospice services for children and their caregivers. Regulations Considering the fact that the whole procedure is dealing with life of a person who can be the most valuable belonging of an entire family, every single act or even comment will bear a huge burden of legislative responsibilities for the PHC team. Ethical regulations for EOL care in pediatrics is considerably debatable, fluid and sometimes ambiguous. This fact stems from the uncertainties about the principle of autonomy in this age group. Physical, mental and social development and the chance for survival into adolescence can be different for children with the same age. , Therefore, in some territories there is no designated age limit for eligibility to consent to, or refuse a medical approach. In addition, cultural background affects the criteria of capability and role in consenting for child and the caregiver in such procedures. In many cases there is no specific eligibility criteria for pediatric hospice and the rules for adults are being implemented. Because HC is not a curative therapeutic approach, instead of “indications”, guidelines usually use the term “eligibility” criteria. However, these criteria are usually conditional and negotiable in hospice field, rather than fixed and strict rules. This matter is even more ambiguous in PHC. In place of indication or eligibility, the American academy of pediatrics have proposed “trigger points” for initiation of such EOL services. These include severe fetal or traumatic disabilities, a progressive or unresponsive disease state with poor prognosis or therapy-resistance, prolonged or frequent hospitalizations, and reaching to certain ages for decision making strategies. For pediatric cancer patients, globally or even nationally applicable standards are very limited and local criteria is usually established by private institutions. Some centers advocate for accessibility to EOL care regardless of the diagnosis, while disease-based prioritization might be necessary in resource limited settings. Considering the broad spectrum of needs of pediatric hospice patients, selection and prioritization of major symptoms for treatment would be necessary. Addressing pain is one of the top-rank common priorities. , Amelioration of nausea, vomiting, loss of appetite, respiratory failure, tiredness, anxiety, and depression should also be adequately managed. In a study in pediatric cancer patients, bleeding, dyspnea, pain, seizures, and delirium were reported to be the most prevalent symptoms during the last days of life, so, addressing these discomforts may be prioritized in oncology hospice settings. Eligibility criteria for receiving this sub-specialized and multi-compartmental intervention should be defined. Each and every hospice center has its own local admission parameters and scoring systems. The format and availability of these criteria is divergent based on cancer type, as well. Exclusive guidelines for pediatric patients in cancer setting has not been clearly provided. Some general instructions which are developed by insurance authorities include being terminally ill with a medical prognosis of 6 months or less and consenting for hospice limitations. One of the key issues and trigger points in hospice care is the “do not resuscitate” or “DNR” code. This code had been a prerequisite in some centers for hospice, but nowadays EOL care receivers can choose their demand for resuscitation. While having an official DNR code is not a universal amenity, in countries with developed health care systems including Canada and USA, signing a DNR agreement is no longer mandatory for hospice services. There are clinical practice guidelines to guide the initiation and progression of hospice/palliative care in oncology. Unfortunately, the recommendation have not been tailored for pediatric settings but they generally include endorsement for early initiation of palliative care process (not hospice), referrals to specialized teams, use of scoring systems for performance status evaluation, and hospice consideration for those with less than 6 months prognosis. , For palliative care (not hospice alone), usually the guidelines are more detailed and provide criteria for both on-diagnosis and during-the-treatment in pediatric oncology setting. As a common point in this area, pro-active legal involvement of the patient, surrogate decision-makers or parents and the healthcare providers is recommended. , Neurobiological (instead of chronological) parameters of adulthood have been suggested as a practical way of assessing a child’s capability for consenting in such situations but this requirs more evaluations. One of the most neglected groups in HC fields includes adolescents and young adults (AYAs). Considering their maturity state, this population needs to be more involved in decision makings. Due to their borderline aging category, a lack of recognition and subsequent heterogeneity for their management is very common. Furthermore, after death of a child the process may continue with more intense issues like requirement for an autopsy. Hence, involvement of a lawyer or forensic physicians has been endorsed in studies. , Quality Improvement Pediatric hospice is a recent concept which needs continuous trimming and reassessment. Methodologies of investigation in pediatric hospice/palliative care are growing and new approaches are under exploration to address the specific needs in this setting. The efficiency and satisfactoriness of these services can be scientifically measured through some standardized tools. Patient and family reports, staff observations and retrospective data analysis are highly valuable resources for these evaluations. Unfortunately, exclusive data to define optimum quality of care in pediatrics is not as accessible as adults. Some evaluation determinants which have been derived from interviews from PHC oncology settings include communication and guidance, inclusion of interdisciplinary professionals, symptom-based measures, and considerations for family preferences. Interestingly, late-stage chemo-radiation with the goal of symptom relief can be considered as a part of palliation. However, the decision in this direction should be debated with highest attention in order to prevent negating the value of painless EOL care. In fact, continuation of anti-cancer therapy in children has been announced as one of the key causes of delay or deprivation of children in need of PHC. In addition, maximization of the parent-children closure time has been announced as the number one priority in some observations. Notably, goal-directed programs for education of clinical staff including physicians and nurses for hospice has shown improvement in PHC success rates. , Also, availability of hospice services in the same clinical institution of admission for the primary disease can boost efficiency. Therefore, a risk-benefit analysis is required before strategic planning in this regard. In a quality measurement study in 2022 on pediatric cancer cases from multiple centers in the United States parameters such as utilization of chemotherapy, mechanical ventilation, intensive care units, and counts of reported distressing symptoms in the last 30 days of life were checked. It was shown that a great proportion of non-hospice users were among the families with lower income. This means that they chose more costly and distressing care approaches for their children. Based on these reports, we can state that there is a huge gap for optimization of PHC. Hence, quality improvement with a socio-educational perspective should be considered as a constant attachment of hospice in achieving the desired goals, in a more time- and cost-efficient manner. Development of high-tech methods such as artificial intelligence has opened new horizons for more strict decision making and referrals to hospice care. However, deep emotional aspects of this issue prohibits carefree establishment of machine-guided approaches in PHC. As the last step, periodical evaluations are required for assessment of trends in this evolutionary discipline. It is obvious that data obtained from care recipients including patients and families can be very enlightening for success evaluation of PHC teams and facilities. These evaluations can be implemented both for general strategic areas of PHC or for detailed issues such as personnels’ eligibility, knowledge and skills. Regarding the short history of modern PHC implementation in the whole world, involvement and assistance of experienced specialists from adult hospice services can be constructive for upgrading the PHC procedures. There is adequate level of evidence in medical literature that proves the necessity of HC in oncology. It has been shown frequently that hospice philosophies are beneficial for the patients, caregivers and health care system. A majority of this evidence is derived from adult-based studies that compared outcomes of patient referrals to hospice services to the outcomes of staying in general or intensive care units. - As mentioned before, the boundaries and definitions of hospice/palliative care is not well defined. Hence, establishing clear-cut points in the continuum of concurrent care transitioning from palliation and curative measures into hospice and EOL care is essential for improving the quality of life and death in this specific population. However, physiologic, emotional, and social characteristics of children bear specific requirements that mandates exclusive services. Due to lack of exclusive experience in pediatric hospice, many of the currently implemented approaches are derived or copied from adult guidelines. In a study on a diverse range of advanced cancer patients, it was shown that lower age was associated with longer survival time in hospice setting. Such facts dictate a specialized planning for PHC. The first and most important role of hospice is changing “pain and distress” with “joy and closeness” for the patient and the family. Therefore, finding the earliest but most conservative time to start the service can be immensely beneficial for the patients, families, and healthcare system. - This can be accomplished through defining the indications or eligibility criteria, setting the priorities, addressing symptoms such as pain, nausea, vomiting, loss of appetite, respiratory failure, tiredness, anxiety, or depression, and decision-making about location and level of care, quality standards, modified expectations, and logical and legal issues. Knowing the statistics and demography of children who used PHC versus non-users can be helpful in long-term policy making for improving the establishment of this service as a quality-of-life-saving strategy. It is notable that some minorities might have lower knowledge or accessibility to these services. In a study in USA from 2011 to 2013, factors like non-Hispanic White race, complex chronic comorbidities, mental issues, technology dependence, specific tumors, and residence in rural areas with lower socioeconomic environments were considered as hindrance factors to hospice admissions. In another study in 2013, it was shown that race, ethnicity, payor status, patient diagnosis, and religion are significantly associated with hospice enrollment and its outcome. Cancer type, hospital size, insurance status and geographic location are other defined factors in this field. Considering the value of home-based care in pediatric hospice, some predictive demographics can be found among the above-mentioned factors to forecast the risk for hospital death. While data in this regard for pediatric oncology hospice is very limited, information from other child-oriented hospice settings can be useful. Age lower than a year, presence of congenital causes, certain locations, proximity to tertiary hospitals are associated with dying in hospital. Focusing on pediatric intensive care unit (PICU) death rates among children with cancer, risk factors such as Hispanic ethnicity, hematologic malignancies, hematopoietic stem cell transplantation, cumulative number of PICU admissions, receiving cancer therapies during the final month, and delayed palliative care involvement have been discovered. In our data pool for this review, we detected a variety of demographic features for pediatric cancer patients under hospice care. The information derived from the references included data for age groups (neonates, children, and young adults), , , - disease type and conditions, , , , , , - types and levels of medical interventions, , , , - access to specialized care and sufficient communication, - geographical location, , , , - physical, mental and behavioural comorbidities, , , , gender, , , , previous hospitalization unit or ward, , , , health insurance type, , , , clinical phase (stable, unstable, deteriorating, terminal), , cost of care, , race and ethnicity, , , financial background, , familial characteristics and siblings, cultural background, , number of admissions, access to palliative care before hospice, , symptomatology, , , and access to general hospitals. It should be noted that some of the studies included non-pediatric-cancer populations, as well. Some of these parameters were associated with different outcomes and success rates of hospice services, which are discussed in different sections of this review. Based on the findings it can be assumed that middle categories of age, non-hematologic malignancies at lower stages, receiving fewer intensive therapies, absence of comorbidities, access to insurance coverage, advanced care and education, less frequent history of admissions and utilization of palliative care are associated with better outcomes regarding the EOL quality of life (QOL) and hospice attendance or success. Parents are major determinants for initiation, process, place, and quality of hospice care in pediatric settings. In such complicated situations, parents are the first-line legal authorities to adopt this kind of lifestyle for their loved ones. This produces is a huge burden of mental stress for them that can threaten their efficient critical-thinking and proper decision making. It has been shown that these parents, suffer more from psychiatric disorders, including anxiety and depression compared to general population. These guardians have to plan their caregiving strategy, deal with financial issues, and handle the familial and social complications of entering a hospice-associated life. This fact is even more complicated among the families that have children with cancer, both due to debilitating expenses and bombardment with every-day advertised experimental and uncertain therapeutic options. One of the greatest challenges for parents is digesting the dilemma, when their decision is not in accordance with their dying child; this makes it even more difficult for the healthcare team to reach a certain opinion, as well. These unwanted and unplanned consequences, plus feelings of regret, continue even after their great loss, and can be accompanied by prolonged grief and disturbed QOL. The role of hospice in family care is to ease the pressure and remove this burden. Taking care of parents of children with terminal illnesses is as important as paying attention to the patients. After the diagnosis, there is a huge risk for total collapse of personal, financial, and mental health of a family dealing with the impending death of a younger family member. The most common aspects of supportive needs among parents have financial, psychosocial, and legal sources. Besides the patients and their parents, other family members deserve to be acknowledged in this process. Young brothers and sisters are priorities in this regard due to their vulnerable situation and minor authoritative role in this family hazard. A literature review by Ridley and Frache demonstrated that there is an alarming gap in coverage of care for youngster siblings of deceased children. Several approaches have been developed in cancer hospice settings, due to their higher prevalence. Group sessions, recreational camps and educational activities have been tried for bereaving children with promising results, but availability of such facilities has not been globally established. In addition, the role of young siblings as assisting caregivers in hospice has been neglected, almost completely. It can be concluded that concordant sponsorships from medical, social, and governmental entities are required to comprehensively support families and decrease the ravaging effects of this issue on parents and other family member. Psychological and social aspects of PHC are as important as its somatic features. The process of providing PHC starts with this portion of care. In recent years there has been a major change from the state of disappointing expectations from cure-oriented thinking that may have a destructive effect on the EOL management procedure, to optimism in a variety of available options for improving the near-death trajectory of a child and the family. This mindset will boost their mental and spiritual strength, significantly. After building a clear and honest bond, hospice systems should provide the most peaceful environment to control and ease the understanding of the end and grief for parents, siblings and caregivers. Even the decisions about the array of medical interventions in this phase should be finalized based on the emotional and relational responses from the care recipients. This issue is very critical in cancer settings in which lots of curative and palliative interventions are available. Then they should provide bereavement aid including emotional support, recognition of abnormal and pathologic responses, and follow-ups. These maneuvers generally target parents or caregivers, rather than their young patients. However, social interactions of a dying child should be addressed as a very complicated issue, as well. In order to plan effectively, determinants of psychological risks should be defined. Studies have shown that besides a rewarding and meaningfulness feeling, caregiving is commonly accompanied by adversities like anxiety, depression, intensified grief reactions, physical and mental health dysfunction and even, increased caregiver mortality compared to general population of parents or those with children affected by non-terminal illnesses. With the use of indicators such as relationship with the patient, age at caregiving, accessibility of help, financial support, and history of loss, high risk entities can be more feasibly detected and supported through psychosocial back up programs in pediatric cancer hospice. Reports from UK childhood cancer centers shows that there are multiple areas for providing these kinds of supportive measures but generally no firm standard has been established to make them continuously and comprehensively accessible. These areas include education, staffing, facilities for children and families, and psychosocial services in terms of group activities and transition plans. In this way, presence of psychiatrists and psychologists can be undeniably beneficial regarding they role in implementation of specialized assessments, diagnosis, interventions and research, both for the child and the family. , However, due to lack of extensive experience in this field the true value of these collaborations has not been fully elucidated in PHC. In addition, death education, initiated in schools can have a preventive effect on mental health disorders for families who will expose to such tragedies, unexpectedly. Continuing this type of education through the society with meticulously supervised methods can act as a public immunization and enhance the society with self-care tools to decrease the devastating consequences, from before entrance to the trajectory of a child’s death. At the end it should be stated that this harsh pathway is more impassable to go through alone. Therefore, the value of mentoring and support groups from those who had the same experience cannot be overemphasized. In this pathway, instead of being considered as hopeless, the patient and the family will be hopeful for spending nice and memorable days together. Families will enjoy the memories of those valuable last days to harness their bitter bereavement. Also, the role of faith which has been introduced in adult settings, should be further explored in PHC, as well. EOL care is associated with a huge financial burden, both for the involved family and the supporting heath care system. These budgeting issues are powerful to the extent that they can change the pathway of care for a dying child. Decreasing the amount of expenses for pointless therapies or unnecessary admissions is hidden in the philosophy of PHC to facilitate the passage of the family through this harsh course. It has been shown in adult settings that implementation of HC in the form of home based care is significantly more cost-saving than hospital-located conventional care. However, the hospice process can be very expensive by itself. Therefore, financial pre-planning for critical situations is mandatory. Logically, families cannot actively be involved in the investment procedures for these plans, especially if the time period between the diagnosis and EOL situation is short. EOL insurance is not a common or routine investment among families who cannot predict such unexpected grave occasions. Also, decision making in this situation about the EOL plan can be immensely difficult. Hence, governmental authorities, social organizations, insurance companies and finally health care systems have to be prepared for providing this kind of support. Regarding the fact that PHC is still a newborn science and entangled with the multi-aspect complications of pediatricEOL services, the rubrics of budgeting and financial planning for it has not been fully developed in many countries. However, there are a limited numbers of studies that tried to explore this new field and propose guidelines. The first step for establishment of an effective financial plan is recognition of the sources of cost. This can be achieved through cost-evaluation studies. Such studies have proposed that the intensity of treatment protocols, physician care, nursing care and presence of some clinical symptoms including asthenia, anorexia, bedsore, nausea and vomiting, can result in rising costs. There is a considerable risk that governments do not have the capacity to provide for all the hospice costs. In this way, the role of support from charities will be vital. Nowadays, a considerable amount of national hospice expenses in countries like England and Wales is provided by charities for PHC. However, this might persuade the government to step back from their true supportive stands. Additionally, it should be noted that governments may devote a considerable budget to HC but recognition of a suitable share for PHC needs more organized amendments. Insurance organizations play a great role in handling of PHC costs. Insurance coverage can be considered as a determinant in choosing the strategic style for the EOL care. Generally, public governmental insurance plans cover these types of expenses. Therefore, those patients who are covered only by the private ones might be deprived of such services. There are territory-based local eligibility criteria for utilization of hospice insurances. For example a certificate from an accredited practitioner to announce that the patient is within the terminal 6 months is mandatory, in some settings. All these diversities in coverages can affect the delivery of PHC. Different jurisdictions have different insurance coverage policies for hospice from zero to 100% of the expenses. , Also, there is a difference between coverage for hospice and palliative care in some places and the policies are evolving. For example, in an American setting, palliative care is paid by personal insurances, involving limitations such as standard co-pay and deductions. A partial coverage is also provided by Medicaid and Medicare. But for hospice, expenses are fully paid by Medicaid or Medicare and coverage includes medications, medical equipment, nursing care, social and spiritual services. The same is true with certain personal insurances. One of the key features of PHC is the place in which the young patient passes away. This issue has been the focus of a considerable portion of studies. In fact, the matter of location for spending the last days can be the leading determinant for planning the strategy of HC. HPC services can be delivered at different locational or strategic points. Supervised home care is one of the most appealing approaches. Specialized hospice wards are also available in limited numbers that provide a combination of medical and spiritual care. In resource limited situations, as a last resort, regular, pre-critical, or critical care units can be adjusted, as much as possible, to accept the young patients for this purpose. In addition, medical procedures are dispensed at different levels of intensity. Serving the patient and family with hospice alone is one option. Due to scarcity and costly nature of these services in some places, simple private nursing care is substituted on some occasions, generally with low satisfaction rates. Another option for the pediatric patients is admission in “concurrent” HC. Although, concurrent palliation is a better terminology in this regard, due to continuation of therapeutic and curative measures. In this setting the patient can be benefitted from both life prolonging care and pain reducing hospice. Because this approach is more resource consuming, developing certain criteria for its recipients is fundamental. The decision about where a patient passes away is affected by numerous factors from different entities of this challenge. Each one of the patient, family and medical team may have a divergent opinion about it. The usual available options include home, general wards, critical care or emergency settings, general hospice, and pediatric hospice facilities. For children with cancer, oncology wards are also a common place to die. However, accessibility to these choices varies in different geographical zones. In addition, continuation of hospitalization in the regular oncology or pediatric wards cannot be recommended. It has been shown that home is the most appealing place for the ending days, both from parents and clinicians’ point of view. Statistical data shows that this happens for a majority of patients under PHC, as well. Although, the actual and the preferred location can be incongruent, specifically for children with haematologic malignancies. On the other hand, presence of an active palliative hospice care can increase the congruency. Notably, in some cases, parents showed more tendency to choose hospital-base care compared to specialized PHC institutions, compared to clinicians. This implies the lack of adequate introduction and popularization of PHC to the public audience. Establishment of HC teams in healthcare centers has pushed the curve toward home-based or hospice-based admissions. Studies demonstrated that cultural, racial, and clinical backgrounds of patients can alter the aforementioned assumptions. For instance, in study in UK that categorized the destinations into hospital and hospice or home, it was shown ethnic minorities were less likely to die in hospice, while a lower portion of White children died in the hospital setting. It should be noted that the optimal choice for the final place can be totally different case-by-case. But based on the aforementioned studies that considering home-death as a preferrable outcome, some arrangements highly improve the QOL and feasibility of this type of care for the involved families. The term “shared care” emphasizes a collaboration of family and the medical team during this period. A wise approach would be educating the household members to handle the major requirements of caring to stay independent from frequent professional visits. The share of healthcare team in this situation is more educational, rather than therapeutic. This method both save the financial status of the bereaving family and spare the limited number of trained staff to attend more cases. In addition, the number of unnecessary admissions and hospital deaths can be diminished through such a goal-directed programming. Insurance policies and governmental financial strategies can also significantly change the trends in this regard. It should be kept in mind that providing hospice services is accompanied by environmental, cultural, economic and religious issues that need preplanning. Also, it is notable that some authors have argued against considering hospice as a location of care and they defined it as a philosophy of compassionate, multidisciplinary care that can be delivered at any place. As mentioned frequently in the previous parts, PHC is a teamwork, and it will be handled optimally if the whole team be present with adequate capabilities, expertise and numbers. Besides their pragmatic function, the accessibility to this team will be highly comforting for both the child and the stressed caregivers involved. It should be kept in mind that there is no specific hierarchical order for roles and duties in this procedure and the importance for all of the elements are totally overlapping. Considering the majority of studies in this field, nurses are the mainstays of this procedure for two reasons. First, their professional dexterity in handling every-day needs of hospice children makes them the frontiers of PHC team. Secondly, they are health practitioners that provide care at a high professional level with considerably longer continuation, due to the nature of their occupational characteristics and history of trainings. In this way, their persistent active presence will help a lot in management of a near death child and the family. On the other hand, lack of adequate nursing staff or availability of home visits significantly decrease the quality of lives of the challenged household in that terminal hectic stage. Physicians are also critical in easing the last days with helping to relieve the pain, giving medical consultations and guiding the caregivers for an optimized decision making. Palliative and hospice care specialists are valuable agents in PHC, but regarding their scarcity, most centers lack the access to such ideally trained professionals. Considering the fact that majority of children on hospice settings are transferred from specialized care services, usually the in-charge physicians are doctors with specialties and subspecialities. Oncology fields are the major referring wards for pediatric hospice; therefore, oncologists have the highest exposure and experience with these cases. Stem cell transplant specialists also are in contact with a big population of children with rare and terminal diseases. However, pediatricians and oncology pediatricians are the most eligible clinical entities in this regard, with deep insight into somatic, psychological, and social features of PHC. Other practitioners like surgeons and anaesthesiologists also have positive effects. In addition, the presence of mental health clinicians such as psychologists and psychiatrists are non-negotiable in preparing the child and family for their destiny. An optimal example in this regard is an organized team involving pediatric oncologists, palliative care physicians, HPC trained nurses, psychologists, child life specialists, and medical social workers. Also, pharmacists can improve the quality of medication management in pediatric EOL care. The last but not least group involves social workers, spiritualists, lawyers, and other non-clinical support coordinators to help the parents navigate socio-economic-legal challenges. This list can be elongated with more clinicians and non-clinical supporters but considering the sensitivity of this issue, empathy and moral devotion are the main requirements of being a team member in hospice services for children and their caregivers. Considering the fact that the whole procedure is dealing with life of a person who can be the most valuable belonging of an entire family, every single act or even comment will bear a huge burden of legislative responsibilities for the PHC team. Ethical regulations for EOL care in pediatrics is considerably debatable, fluid and sometimes ambiguous. This fact stems from the uncertainties about the principle of autonomy in this age group. Physical, mental and social development and the chance for survival into adolescence can be different for children with the same age. , Therefore, in some territories there is no designated age limit for eligibility to consent to, or refuse a medical approach. In addition, cultural background affects the criteria of capability and role in consenting for child and the caregiver in such procedures. In many cases there is no specific eligibility criteria for pediatric hospice and the rules for adults are being implemented. Because HC is not a curative therapeutic approach, instead of “indications”, guidelines usually use the term “eligibility” criteria. However, these criteria are usually conditional and negotiable in hospice field, rather than fixed and strict rules. This matter is even more ambiguous in PHC. In place of indication or eligibility, the American academy of pediatrics have proposed “trigger points” for initiation of such EOL services. These include severe fetal or traumatic disabilities, a progressive or unresponsive disease state with poor prognosis or therapy-resistance, prolonged or frequent hospitalizations, and reaching to certain ages for decision making strategies. For pediatric cancer patients, globally or even nationally applicable standards are very limited and local criteria is usually established by private institutions. Some centers advocate for accessibility to EOL care regardless of the diagnosis, while disease-based prioritization might be necessary in resource limited settings. Considering the broad spectrum of needs of pediatric hospice patients, selection and prioritization of major symptoms for treatment would be necessary. Addressing pain is one of the top-rank common priorities. , Amelioration of nausea, vomiting, loss of appetite, respiratory failure, tiredness, anxiety, and depression should also be adequately managed. In a study in pediatric cancer patients, bleeding, dyspnea, pain, seizures, and delirium were reported to be the most prevalent symptoms during the last days of life, so, addressing these discomforts may be prioritized in oncology hospice settings. Eligibility criteria for receiving this sub-specialized and multi-compartmental intervention should be defined. Each and every hospice center has its own local admission parameters and scoring systems. The format and availability of these criteria is divergent based on cancer type, as well. Exclusive guidelines for pediatric patients in cancer setting has not been clearly provided. Some general instructions which are developed by insurance authorities include being terminally ill with a medical prognosis of 6 months or less and consenting for hospice limitations. One of the key issues and trigger points in hospice care is the “do not resuscitate” or “DNR” code. This code had been a prerequisite in some centers for hospice, but nowadays EOL care receivers can choose their demand for resuscitation. While having an official DNR code is not a universal amenity, in countries with developed health care systems including Canada and USA, signing a DNR agreement is no longer mandatory for hospice services. There are clinical practice guidelines to guide the initiation and progression of hospice/palliative care in oncology. Unfortunately, the recommendation have not been tailored for pediatric settings but they generally include endorsement for early initiation of palliative care process (not hospice), referrals to specialized teams, use of scoring systems for performance status evaluation, and hospice consideration for those with less than 6 months prognosis. , For palliative care (not hospice alone), usually the guidelines are more detailed and provide criteria for both on-diagnosis and during-the-treatment in pediatric oncology setting. As a common point in this area, pro-active legal involvement of the patient, surrogate decision-makers or parents and the healthcare providers is recommended. , Neurobiological (instead of chronological) parameters of adulthood have been suggested as a practical way of assessing a child’s capability for consenting in such situations but this requirs more evaluations. One of the most neglected groups in HC fields includes adolescents and young adults (AYAs). Considering their maturity state, this population needs to be more involved in decision makings. Due to their borderline aging category, a lack of recognition and subsequent heterogeneity for their management is very common. Furthermore, after death of a child the process may continue with more intense issues like requirement for an autopsy. Hence, involvement of a lawyer or forensic physicians has been endorsed in studies. , Pediatric hospice is a recent concept which needs continuous trimming and reassessment. Methodologies of investigation in pediatric hospice/palliative care are growing and new approaches are under exploration to address the specific needs in this setting. The efficiency and satisfactoriness of these services can be scientifically measured through some standardized tools. Patient and family reports, staff observations and retrospective data analysis are highly valuable resources for these evaluations. Unfortunately, exclusive data to define optimum quality of care in pediatrics is not as accessible as adults. Some evaluation determinants which have been derived from interviews from PHC oncology settings include communication and guidance, inclusion of interdisciplinary professionals, symptom-based measures, and considerations for family preferences. Interestingly, late-stage chemo-radiation with the goal of symptom relief can be considered as a part of palliation. However, the decision in this direction should be debated with highest attention in order to prevent negating the value of painless EOL care. In fact, continuation of anti-cancer therapy in children has been announced as one of the key causes of delay or deprivation of children in need of PHC. In addition, maximization of the parent-children closure time has been announced as the number one priority in some observations. Notably, goal-directed programs for education of clinical staff including physicians and nurses for hospice has shown improvement in PHC success rates. , Also, availability of hospice services in the same clinical institution of admission for the primary disease can boost efficiency. Therefore, a risk-benefit analysis is required before strategic planning in this regard. In a quality measurement study in 2022 on pediatric cancer cases from multiple centers in the United States parameters such as utilization of chemotherapy, mechanical ventilation, intensive care units, and counts of reported distressing symptoms in the last 30 days of life were checked. It was shown that a great proportion of non-hospice users were among the families with lower income. This means that they chose more costly and distressing care approaches for their children. Based on these reports, we can state that there is a huge gap for optimization of PHC. Hence, quality improvement with a socio-educational perspective should be considered as a constant attachment of hospice in achieving the desired goals, in a more time- and cost-efficient manner. Development of high-tech methods such as artificial intelligence has opened new horizons for more strict decision making and referrals to hospice care. However, deep emotional aspects of this issue prohibits carefree establishment of machine-guided approaches in PHC. As the last step, periodical evaluations are required for assessment of trends in this evolutionary discipline. It is obvious that data obtained from care recipients including patients and families can be very enlightening for success evaluation of PHC teams and facilities. These evaluations can be implemented both for general strategic areas of PHC or for detailed issues such as personnels’ eligibility, knowledge and skills. Regarding the short history of modern PHC implementation in the whole world, involvement and assistance of experienced specialists from adult hospice services can be constructive for upgrading the PHC procedures. Our findings highlight that our understanding and delivery of pediatric hospice resembles an ocean of uncertainties, doubts and unknown issues. Regarding the maturation of knowledge, this field of medicine and patient care is in its infancy. This immaturity makes reaching a consensus on guidelines with straightforward instructions seemingly impossible. While the majority of the current rubrics in this field are generalizable to a diverse spectrum of ailments, individualization for each group is clearly required. Among these groups pediatric cancer patients are exclusively considerable due their huge population and special needs. However, as mentioned before, the complications of hospice care in pediatric oncology include lack of child-oriented guidelines, psychosocial issues, challenges of parents and other family members, uncertainties for the place and intensity of care including continuation or discontinuation or modification of cancer therapies, financial obstacles, inadequacy in the number of specialized practitioners and ambiguity of monitoring parameters for quality improvement. The same can be said for all other life-limiting conditions experienced by children and their families. Each and every one of these issues requires a well-targeted approach for comprehension, analysis and research to find evidence-based and logical solutions for easing the physical, mental, emotional and spiritual pain of the involved entities in a child’s final palliative and end of life journey. |
COVID-19: a further step forward in the long journey of Occupational Medicine | 2dd43e8e-e5d1-4ff8-8a8a-3d829e0911a5 | 8223939 | Preventive Medicine[mh] | |
A foundation model for clinical-grade computational pathology and rare cancers detection | eea1a661-c9ec-48e6-b829-dcc3cd91b5d3 | 11485232 | Pathology[mh] | Pathologic analysis of tissue is essential for the diagnosis and treatment of cancer. Increasingly, the traditional histological preparations used for light microscopy examination are being replaced by their digital counterparts, also known as whole-slide images (WSIs), which enables the use of computational pathology – to move from primarily academic proof points to routine tools in clinical practice. Computational pathology applies artificial intelligence (AI) to digitized WSIs to support the diagnosis, characterization and understanding of disease , . Initial work has focused on clinical decision support tools to enhance current workflows – , and in 2021 the first Food and Drug Administration-approved AI pathology system was launched . However, given the incredible gains in performance of computer vision, a subfield of AI focused on images, more recent studies – attempt to unlock new insights from routine WSIs and reveal undiscovered outcomes such as prognosis and therapeutic response . If successful, such efforts would enhance the utility of hematoxylin and eosin (H&E)-stained WSIs and reduce reliance on specialized and often expensive immunohistochemistry (IHC) or genomic testing . A major factor in the performance gains of computer vision models has been the creation of large-scale deep neural networks, termed foundation models. Foundation models are trained on enormous datasets—orders of magnitude greater than any used historically for computational pathology—using a family of algorithms, referred to as self-supervised learning (for example, refs. – ), which do not require curated labels. Foundation models generate data representations, called embeddings, that can generalize well to diverse predictive tasks . This offers a distinct advantage over current diagnostic-specific methods in computational pathology, which, limited to a subset of pathology images, are less likely to reflect the full spectrum of variations in tissue morphology and laboratory preparations necessary for adequate generalization in practice. The value of generalization from large datasets is even greater for applications with inadequate quantities of data to develop bespoke models, as is the case for the detection of uncommon or rare tumor types, as well as for less common diagnostic tasks such as the prediction of specific genomic alterations, clinical outcomes and therapeutic response. A successful pathology foundation model should capture a broad spectrum of patterns, including cellular morphology, tissue architecture, staining characteristics, nuclear morphology, mitotic figures, necrosis, inflammatory response, neovascularization and biomarker expression and therefore would be well-suited to predicting a wide variety of WSI characteristics. If trained with a sufficiently large quantity of digitized WSIs in the pathology domain, such a model could form the basis for clinically robust prediction of both common and rare cancers, as well as for other critical tasks such as subtyping of cancer, quantification of biomarkers, counting of cellular instances and events and the prediction of therapeutic response. Foundation model performance crucially depends on dataset and model size, as demonstrated by scaling law results – . Modern foundation models in the natural image domain use millions of images (for example, ImageNet , JFT-300M and LVD-142M ) to train models with hundreds of millions to billions of parameters (for example, vision transformers (ViTs) ). Despite the challenges in collecting large-scale datasets in the pathology domain, recent pioneering works have utilized datasets ranging from 30,000 to 400,000 WSIs to train foundation models ranging in size from 28 million to 307 million parameters – (see Supplementary Note for a detailed summary of recent models). These works demonstrate that image features produced with self-supervised learning of pathology images outperform image features trained on natural images and that performance improves with scale. Here, we present a million-image-scale pathology foundation model, Virchow, named in honor of Rudolf Virchow, who is regarded as the father of modern pathology , and proposed the first theory of cellular pathology . Virchow is trained on data from approximately 100,000 patients corresponding to approximately 1.5 million H&E stained WSIs acquired from Memorial Sloan Kettering Cancer Center (MSKCC), which is 4–10× more WSIs than in prior training datasets in pathology (detailed in Fig. and ‘Million-scale training dataset’ in ). The training data are composed of cancerous and benign tissues, collected via biopsy (63%) and resection (37%), from 17 high-level tissues (Fig. ). Virchow, a 632 million parameter ViT model, is trained using the DINO v.2 algorithm , a multiview student–teacher self-supervised algorithm (Fig. ; see ‘Virchow architecture and training’ in for training details). DINO v.2 leverages global and local regions of tissue tiles to learn to produce embeddings of WSI tiles (Fig. ), which can be aggregated across slides and used to train a variety of downstream predictive tasks (Fig. ). Motivated by highlighting the potential clinical impact of a pathology foundation model, we assess the performance of a pan-cancer model trained using the Virchow embeddings to predict specimen-level cancer across different tissues. Virchow embeddings outperform or match all baseline models on all tested cancer types, notably including rare cancers and out-of-distribution (OOD) data. Quantitative comparison to three specialized clinical-grade AI products demonstrates that the pan-cancer model performs nearly as well as the clinical products in general and outperforms them on some rare variants of cancers. To provide evidence for potential focus areas for future advances in computational pathology, qualitative analysis is also performed, characterizing the error patterns where the AI model fails to identify or falsely identifies cancerous cells. Motivated by simplifying clinical workflows, we evaluated the use of Virchow embeddings to train biomarker prediction, generally outperforming other models. Overall, our results provide evidence that large-scale foundation models can be the basis for robust results in a new frontier of computational pathology. The Virchow model embeddings were evaluated on two categories of slide-level computational pathology applications: pan-cancer detection (‘Virchow enables pan-cancer detection’ and ‘Towards clinical-grade performance’) and biomarker prediction (‘Biomarker detection in routine imaging obviates additional testing’). These tasks require training a weakly supervised aggregator model to group tile embeddings to slide-level predictions. A series of tile-level linear probing benchmarks were also performed to directly assess the embeddings on individual tissue tiles (‘Tile-level benchmarks and qualitative analysis demonstrate generalizability’). Virchow enables pan-cancer detection A key aim of our work was to develop a single model to detect cancer, including rare cancers (defined by the National Cancer Institute (NCI) as cancers with an annual incidence in the United States of fewer than 15 people per 100,000 (ref. )), across various tissues. The pan-cancer detection model infers the presence of cancer using Virchow embeddings as input. For evaluation, slides from MSKCC and slides submitted for consultation to MSKCC from numerous external sites globally are used. Stratified performance across nine common and seven rare cancer types is reported. Embeddings generated by Virchow, UNI , Phikon and CTransPath are evaluated. Pan-cancer aggregators are trained using specimen-level labels, maintaining the same training protocol for all embeddings (see ‘Pan-cancer detection’ in for data and training details). Virchow embeddings yielded the best cancer detection performance on all cancer types (Fig. ). Pan-cancer detection using UNI embeddings achieved statistically similar performance ( P < 0.05) for eight of the nine common cancer types and five of the seven rare cancer types; nevertheless, in all but one case, the specific area under (the receiver operating characteristic) curve (AUC) score was lower. Overall the pan-cancer model achieved an AUC of 0.950 with Virchow embeddings, 0.940 with UNI embeddings, 0.932 with Phikon embeddings and 0.907 with CTransPath embeddings (Fig. ; all significantly different with P < 0.0001). See Extended Data Fig. for more detailed AUC and specificity metrics, stratified by cancer type. Rare cancer detection performance is particularly noteworthy. Compared to the aforementioned AUC of 0.950 overall, Virchow embeddings yielded an AUC of 0.937 on rare cancers (Fig. ), demonstrating generalization to rare data. Performance across the individual rare cancers was, however, non-uniform, with detection of cervical and bone cancers proving more challenging (AUC < 0.9) irrespective of the embeddings used (Fig. ). Virchow embeddings improved cervix detection to 0.875 AUC compared with 0.830, 0.810 or 0.753 when using UNI, Phikon or CTransPath embeddings, respectively. Similarly, Virchow embeddings yielded 0.841 AUC for bone cancer detection, compared to 0.813, 0.822 and 0.728 with UNI, Phikon and CTransPath, respectively. At 95% sensitivity, we show that a pan-cancer detection model using Virchow embeddings can achieve 72.5% specificity, compared to 68.9%, 62.9% or 52.3% using UNI, Phikon or CTransPath embeddings, respectively, trained on less data (Fig. ). The robustness of Virchow embeddings to data sampled from a different population than the training set (OOD data) is evaluated directly with data from institutions other than MSKCC (both Virchow and the pan-cancer aggregator were trained only on data from MSKCC) and indirectly by including data from tissues which were not observed during training (Fig. ). As AUC measures cannot be exactly compared across different data subsets (due to different positive to negative sample ratios), we report AUC for all pan-cancer models on all data or rare cancers (Fig. ), as well as on internal or external data (Fig. ), and demonstrate that the AUC differences across models remain consistent in each subpopulation. This demonstrates that Virchow embeddings generalize well to new or rare data and outperform the others consistently. Although AUC cannot be exactly compared across data subsets, we can observe that all models achieve a similar AUC on both internal and external data, suggesting that they generalize well as external data can be challenging because it is submitted to MSKCC for consultation. Furthermore, cervix, testis and head and neck (H&N) are tissues not seen during training, and Virchow embeddings still outperform competing models. Overall, pan-cancer detection generalizes across cancer types, including rare cancers, as well as on OOD data when using foundation model embeddings. The comparison of pan-cancer performance based on different foundation model embeddings reveals that performance scales with the size of the foundation model and the size of the training data (Fig. ). Cancer detection was found to scale approximately logarithmically with the number of model parameters (Fig. , top); although performance scaled with the number of training tile samples, the trend (Fig. bottom) suggests diminishing returns. Although the training datasets, model architectures and optimization strategies differ across Virchow, UNI, Phikon and CTransPath, there are enough similarities to motivate the scaling analysis. All models are transformer-based: CTransPath uses a Swin transformer , and the rest use ViTs of different sizes. Phikon was trained using the iBOT algorithm , and both Virchow and UNI were trained using the DINO v.2 algorithm with similar hyperparameters. iBOT and DINO v.2 are related approaches as the latter builds on the masked image modeling proposal of the former. CTransPath is differentiated in terms of training algorithm as it used a contrastive learning algorithm based on MoCov3 (ref. ). To learn about the effect of dataset size independent of model size, we direct the reader to the study in ref. . Toward clinical-grade performance A promise of foundation models is improved generalization; however, this claim is difficult to verify without access to rigorously trained and tested tissue-specific specialist models. To this end, we conducted a comparative analysis between the Virchow-based pan-cancer detection model and specialist commercial models, specifically Paige Prostate, Paige Breast and Paige Breast Lymph Node (BLN). The comparison focuses on the AUC for cancer detection, specifically for prostate cancer, invasive breast cancer and metastases of breast cancer in lymph nodes. These commercial models were trained using multiple-instance weakly supervised learning as described in refs. , specifically for cancer detection. The evaluation was performed in two settings: (1) product testing datasets and (2) rare cancer variant datasets in the respective tissues (Fig. ). The Virchow-based pan-cancer detection model, trained on cancers across numerous tissues, performs nearly as well as the prostate, breast and BLN clinical specialist models (Fig. ) while outperforming them on many rare variants of cancers (Fig. ). It is important to note that the pan-cancer training set did not benefit from the same refinement as the product training sets, such as enrichment for subpopulations and label quality control. Furthermore, the pan-cancer model was trained on fewer tissue-specific specimens than the clinical models (Fig. and Extended Data Fig. ). Concretely, Paige Prostate was trained on 66,713 blocks, Paige Breast was trained on 44,588 specimens and BLN on 8150 specimens, whereas pan-cancer (using Virchow embeddings) was trained on only 35,387 groups of slides (blocks or specimens) in total, of which 2,829 are prostate, 1,626 are breast and 1,441 are lymph node. The pan-cancer model achieves an AUC of 0.980, 0.985 and 0.971 on prostate, breast and BLN, respectively. This performance approaches that of commercial models; however, it is still surpassed by the Food and Drug Administration-approved Paige Prostate model (0.980 versus 0.995 AUC, P < 0.05) and the Paige Breast model (0.985 versus 0.992 AUC, P < 0.01). On the other hand, it is statistically significantly better at detecting macrometastases than Paige BLN (0.999 versus 0.994 AUC, P < 0.05). Furthermore, there is no statistically significant difference ( P < 0.05) in the other BLN comparisons or some of the stratified breast cancer comparisons (Fig. ). In addition to approaching the specialist models in terms of overall AUCs, the pan-cancer model matches or outperforms these models on rare variants of cancers, as shown in Fig. . In prostate and lymph node tissues, the pan-cancer model is capable of detecting lymphoma. This is particularly noteworthy because none of the models were trained in hematolymphoid malignancies. Owing to their different lineage (carcinomas originate from epithelial cells, whereas lymphomas arise from lymphoid tissue) their morphologic appearance tends to be quite different. In two of the four lymphoma variants, the pan-cancer model outperforms the specialized model. Improved detection of diffused large B-cell lymphoma is noteworthy as this variant is particularly aggressive. In breast tissue, the pan-cancer model outperforms the Paige Breast model overall and especially on some rare histological variants, including adenoid cystic carcinoma, carcinoma with apocrine differentiation ( P < 0.05), metaplastic carcinoma spindle cell ( P < 0.01), metaplastic carcinoma squamous cell and the exceptionally unusual secretory carcinoma. We note that due to the rarity of these variants of cancers, rare variants prediction lacks the statistical power of the product datasets. To comprehend the error patterns of the pan-cancer model across various tissues, a pathologist examined the error cases within a curated set of evaluation WSIs (see ‘Pan-tissue product benchmark’ in the section ‘Clinical evaluation datasets’ in ). The operating point for each tissue was selected to achieve approximately 95% sensitivity and 85% specificity on a tuning dataset. These error patterns were documented using free text first, which was subsequently categorized to provide a comprehensive summary. We posit that these patterns could be beneficial to similar cancer detection studies, providing valuable insights for the enhancement of future foundational models and clinical AI applications. The false positive and false negative patterns were analyzed separately, as depicted in Fig. . Upon analysis of the false positive and false negative cases, it was discerned that a substantial proportion could be attributed to specific findings. Histological preparations that contained only small tumoral foci constituted the majority (45.2%) of the false negatives. Certain neoplasms, undetected as cancer (11.9%), were of borderline malignant potential, such as gastrointestinal stromal tumors or borderline serous neoplasm of the ovary. Others (9.5%), such as low-grade astrocytoma, exhibited only very subtle histologic features of malignancy. Treatment effects, extensive necrosis and tissue artifacts obscuring the cancer accounted for a few false negatives. In 11 cases (26.2%), there was more than minimal cancer within the specimen, and the negative result of the model could not be explained. The majority of the false positive cases fell into two categories. Precursor lesions in specimens lacking invasive cancer constituted most (53.2%) of the false positives. These were found most frequently in the bladder, breast, cervix, skin (squamous dysplasia) and esophagus. Most detected precursors exhibited high-grade dysplasia, with cytologic features resembling those of invasive carcinoma, although some foci of low-grade dysplasia were also detected in the gastroesophageal junction and skin. The second most common (17.0%) cause of false positive results was tissue artifacts, especially crush artifacts (in which non-neoplastic cells are physically crushed during sample preparation, resulting in a characteristic streaming effect of the nuclei), tissue folds and out-of-focus regions. Reactive alterations within the stroma or lymphoid components, constituting 14.9%, and in non-neoplastic epithelial tissue, representing 11.7%, were also responsible for false positive results. A number of these findings, such as biopsy site changes, reactive epithelial atypia, glandular atrophy and acellular stromal mucin, are well-recognized malignant mimics that challenge pathologists as well. Three cases (3.2%) were benign neoplasms misidentified as cancer. These included benign gastrointestinal stromal tumors, hepatic angiomyolipomas and serous cystadenomas of the pancreas. Biomarker detection in routine imaging obviates additional testing The prediction of biomarkers from standard H&E stained images can reduce the reliance on testing using additional methods and the associated substantial delays in returning results to patients (Fig. ). The status of a biomarker in a specimen is predicted using an aggregator network with the foundation model embeddings as input. These biomarkers play a crucial role in the diagnosis and treatment of various cancers, and each is described in further detail in ‘Biomarker detection’ in (see also Supplementary Table and Fig. ). The biomarker detection datasets consist of WSIs from the histological sections matching the blocks used for DNA extraction and MSK-integrated mutation profiling of actionable targets (MSK-IMPACT) sequencing , the latter of which was analyzed to determine the status of genetic alterations and establish a binary label indicating the presence or absence of the variants: that is, the biomarker (Fig. ). Similar to the pan-cancer evaluation, the publicly available UNI , Phikon and CTransPath models are used as baseline models for comparisons. We note that the biomarker prediction results lacked sufficient statistical power to assess statistically significant differences across models; instead, we conclude relative model performance from evaluating many different biomarker predictions. In our comparative analysis shown in Fig. , Virchow embeddings demonstrated superior performance in seven of the nine evaluated digital biomarkers, achieving AUC scores that exceeded those of the nearest baseline foundation models. This performance underscores the robustness of Virchow embeddings across a diverse range of biomarkers. Even in the categories of prostate–androgen receptor (AR) and ovarian–fraction of genome altered (FGA), where Virchow did not secure the top position, it remained a strong contender, with AUCs of 0.849 and 0.847, respectively. These findings underscore the potential of Virchow embeddings to accurately represent H&E histologic phenotypes, offering predictive insights into biomarkers that are traditionally identified through DNA extraction and MSK-IMPACT sequencing. Tile-level benchmarks and qualitative analysis demonstrate generalizability To directly evaluate tile-level embeddings without the confounder of training an aggregator network, we evaluated Virchow performance on a set of tile-level benchmarks by linear probing. Linear probe evaluation aims to gauge the quality and separability of representations learned by a self-supervised model. We compare Virchow embeddings to baseline model embeddings by applying the same linear probing protocol for each model, using the same training, validation and testing data splits (see ‘Tile-level benchmarking’ in for further details). The analysis is performed both on public datasets and on an internal MSKCC pan-cancer dataset. The internal multitissue dataset for pan-cancer detection at the tile level (referred to as PanMSK) is an in-distribution benchmark, as it is composed of annotations on a held out set of patients across the entire diverse set of tissue groups selected for training (Fig. ). The public datasets are OOD benchmarks and are described in the ‘Tile-level benchmarking’ section in . In addition to UNI , Phikon and CTransPath , DINO p =8 (ref. ) (49 million parameter model trained using The Cancer Genome Atlas (TCGA) and an internal dataset), PLIP (87 million parameter model trained using pathology image-text pairs) and NatImg (1.1 billion parameter model trained on 142 million natural images) are evaluated. As shown in Fig. , Virchow embeddings match or surpass the performance of other embeddings in seven of the eight benchmark tasks (Fig. ; see Supplementary Table for additional metrics). The closest competing models are UNI and Phikon, with UNI scoring in the top 1 three times and in the top 2 for all tasks and Phikon scoring in among the top 2 three times. Virchow demonstrates strong OOD performance as measured by the WILDS and ‘CRC (no norm)’ tasks. The WILDS test data is sourced from a hospital that is not encountered in the training set. The ‘CRC (no norm)’ task introduces a distribution shift from the stain-normalized training set by avoiding stain normalization on the testing set. Without normalization, Virchow’s performance declines by only −0.005 in weighted F 1 score, indicating robustness to variations in data preprocessing. To qualitatively evaluate whether the embeddings learned by Virchow tend to separate the image into semantically meaningful clusters of features, we performed an unsupervised feature analysis similar to the procedure in ref. using the CoNSeP dataset , which contains H&E stained slides of colorectal adenocarcinoma (detailed under ‘Qualitative feature analysis’ in ). We observe approximate semantic segmentation of the cell types in the CoNSeP images (Fig. ). In both examples, the first principal component highlighted malignant epithelium (red) cells. The second principal component, respectively, highlighted miscellaneous cells (yellow) and inflammatory (magenta) cells. DINO v.2 was shown to learn a similar semantic feature separation on natural images, allowing foreground/background separation (for example, discriminating a bus or a bird from the background) as well as part annotation (for example, wheels versus windows in a bus) . Here, we show that this emerging property of the model carries over to the pathology domain. This encouraging result supports our expectation that the unsupervised features learned by Virchow are meaningful and interpretable for a wide range of downstream tasks. A key aim of our work was to develop a single model to detect cancer, including rare cancers (defined by the National Cancer Institute (NCI) as cancers with an annual incidence in the United States of fewer than 15 people per 100,000 (ref. )), across various tissues. The pan-cancer detection model infers the presence of cancer using Virchow embeddings as input. For evaluation, slides from MSKCC and slides submitted for consultation to MSKCC from numerous external sites globally are used. Stratified performance across nine common and seven rare cancer types is reported. Embeddings generated by Virchow, UNI , Phikon and CTransPath are evaluated. Pan-cancer aggregators are trained using specimen-level labels, maintaining the same training protocol for all embeddings (see ‘Pan-cancer detection’ in for data and training details). Virchow embeddings yielded the best cancer detection performance on all cancer types (Fig. ). Pan-cancer detection using UNI embeddings achieved statistically similar performance ( P < 0.05) for eight of the nine common cancer types and five of the seven rare cancer types; nevertheless, in all but one case, the specific area under (the receiver operating characteristic) curve (AUC) score was lower. Overall the pan-cancer model achieved an AUC of 0.950 with Virchow embeddings, 0.940 with UNI embeddings, 0.932 with Phikon embeddings and 0.907 with CTransPath embeddings (Fig. ; all significantly different with P < 0.0001). See Extended Data Fig. for more detailed AUC and specificity metrics, stratified by cancer type. Rare cancer detection performance is particularly noteworthy. Compared to the aforementioned AUC of 0.950 overall, Virchow embeddings yielded an AUC of 0.937 on rare cancers (Fig. ), demonstrating generalization to rare data. Performance across the individual rare cancers was, however, non-uniform, with detection of cervical and bone cancers proving more challenging (AUC < 0.9) irrespective of the embeddings used (Fig. ). Virchow embeddings improved cervix detection to 0.875 AUC compared with 0.830, 0.810 or 0.753 when using UNI, Phikon or CTransPath embeddings, respectively. Similarly, Virchow embeddings yielded 0.841 AUC for bone cancer detection, compared to 0.813, 0.822 and 0.728 with UNI, Phikon and CTransPath, respectively. At 95% sensitivity, we show that a pan-cancer detection model using Virchow embeddings can achieve 72.5% specificity, compared to 68.9%, 62.9% or 52.3% using UNI, Phikon or CTransPath embeddings, respectively, trained on less data (Fig. ). The robustness of Virchow embeddings to data sampled from a different population than the training set (OOD data) is evaluated directly with data from institutions other than MSKCC (both Virchow and the pan-cancer aggregator were trained only on data from MSKCC) and indirectly by including data from tissues which were not observed during training (Fig. ). As AUC measures cannot be exactly compared across different data subsets (due to different positive to negative sample ratios), we report AUC for all pan-cancer models on all data or rare cancers (Fig. ), as well as on internal or external data (Fig. ), and demonstrate that the AUC differences across models remain consistent in each subpopulation. This demonstrates that Virchow embeddings generalize well to new or rare data and outperform the others consistently. Although AUC cannot be exactly compared across data subsets, we can observe that all models achieve a similar AUC on both internal and external data, suggesting that they generalize well as external data can be challenging because it is submitted to MSKCC for consultation. Furthermore, cervix, testis and head and neck (H&N) are tissues not seen during training, and Virchow embeddings still outperform competing models. Overall, pan-cancer detection generalizes across cancer types, including rare cancers, as well as on OOD data when using foundation model embeddings. The comparison of pan-cancer performance based on different foundation model embeddings reveals that performance scales with the size of the foundation model and the size of the training data (Fig. ). Cancer detection was found to scale approximately logarithmically with the number of model parameters (Fig. , top); although performance scaled with the number of training tile samples, the trend (Fig. bottom) suggests diminishing returns. Although the training datasets, model architectures and optimization strategies differ across Virchow, UNI, Phikon and CTransPath, there are enough similarities to motivate the scaling analysis. All models are transformer-based: CTransPath uses a Swin transformer , and the rest use ViTs of different sizes. Phikon was trained using the iBOT algorithm , and both Virchow and UNI were trained using the DINO v.2 algorithm with similar hyperparameters. iBOT and DINO v.2 are related approaches as the latter builds on the masked image modeling proposal of the former. CTransPath is differentiated in terms of training algorithm as it used a contrastive learning algorithm based on MoCov3 (ref. ). To learn about the effect of dataset size independent of model size, we direct the reader to the study in ref. . A promise of foundation models is improved generalization; however, this claim is difficult to verify without access to rigorously trained and tested tissue-specific specialist models. To this end, we conducted a comparative analysis between the Virchow-based pan-cancer detection model and specialist commercial models, specifically Paige Prostate, Paige Breast and Paige Breast Lymph Node (BLN). The comparison focuses on the AUC for cancer detection, specifically for prostate cancer, invasive breast cancer and metastases of breast cancer in lymph nodes. These commercial models were trained using multiple-instance weakly supervised learning as described in refs. , specifically for cancer detection. The evaluation was performed in two settings: (1) product testing datasets and (2) rare cancer variant datasets in the respective tissues (Fig. ). The Virchow-based pan-cancer detection model, trained on cancers across numerous tissues, performs nearly as well as the prostate, breast and BLN clinical specialist models (Fig. ) while outperforming them on many rare variants of cancers (Fig. ). It is important to note that the pan-cancer training set did not benefit from the same refinement as the product training sets, such as enrichment for subpopulations and label quality control. Furthermore, the pan-cancer model was trained on fewer tissue-specific specimens than the clinical models (Fig. and Extended Data Fig. ). Concretely, Paige Prostate was trained on 66,713 blocks, Paige Breast was trained on 44,588 specimens and BLN on 8150 specimens, whereas pan-cancer (using Virchow embeddings) was trained on only 35,387 groups of slides (blocks or specimens) in total, of which 2,829 are prostate, 1,626 are breast and 1,441 are lymph node. The pan-cancer model achieves an AUC of 0.980, 0.985 and 0.971 on prostate, breast and BLN, respectively. This performance approaches that of commercial models; however, it is still surpassed by the Food and Drug Administration-approved Paige Prostate model (0.980 versus 0.995 AUC, P < 0.05) and the Paige Breast model (0.985 versus 0.992 AUC, P < 0.01). On the other hand, it is statistically significantly better at detecting macrometastases than Paige BLN (0.999 versus 0.994 AUC, P < 0.05). Furthermore, there is no statistically significant difference ( P < 0.05) in the other BLN comparisons or some of the stratified breast cancer comparisons (Fig. ). In addition to approaching the specialist models in terms of overall AUCs, the pan-cancer model matches or outperforms these models on rare variants of cancers, as shown in Fig. . In prostate and lymph node tissues, the pan-cancer model is capable of detecting lymphoma. This is particularly noteworthy because none of the models were trained in hematolymphoid malignancies. Owing to their different lineage (carcinomas originate from epithelial cells, whereas lymphomas arise from lymphoid tissue) their morphologic appearance tends to be quite different. In two of the four lymphoma variants, the pan-cancer model outperforms the specialized model. Improved detection of diffused large B-cell lymphoma is noteworthy as this variant is particularly aggressive. In breast tissue, the pan-cancer model outperforms the Paige Breast model overall and especially on some rare histological variants, including adenoid cystic carcinoma, carcinoma with apocrine differentiation ( P < 0.05), metaplastic carcinoma spindle cell ( P < 0.01), metaplastic carcinoma squamous cell and the exceptionally unusual secretory carcinoma. We note that due to the rarity of these variants of cancers, rare variants prediction lacks the statistical power of the product datasets. To comprehend the error patterns of the pan-cancer model across various tissues, a pathologist examined the error cases within a curated set of evaluation WSIs (see ‘Pan-tissue product benchmark’ in the section ‘Clinical evaluation datasets’ in ). The operating point for each tissue was selected to achieve approximately 95% sensitivity and 85% specificity on a tuning dataset. These error patterns were documented using free text first, which was subsequently categorized to provide a comprehensive summary. We posit that these patterns could be beneficial to similar cancer detection studies, providing valuable insights for the enhancement of future foundational models and clinical AI applications. The false positive and false negative patterns were analyzed separately, as depicted in Fig. . Upon analysis of the false positive and false negative cases, it was discerned that a substantial proportion could be attributed to specific findings. Histological preparations that contained only small tumoral foci constituted the majority (45.2%) of the false negatives. Certain neoplasms, undetected as cancer (11.9%), were of borderline malignant potential, such as gastrointestinal stromal tumors or borderline serous neoplasm of the ovary. Others (9.5%), such as low-grade astrocytoma, exhibited only very subtle histologic features of malignancy. Treatment effects, extensive necrosis and tissue artifacts obscuring the cancer accounted for a few false negatives. In 11 cases (26.2%), there was more than minimal cancer within the specimen, and the negative result of the model could not be explained. The majority of the false positive cases fell into two categories. Precursor lesions in specimens lacking invasive cancer constituted most (53.2%) of the false positives. These were found most frequently in the bladder, breast, cervix, skin (squamous dysplasia) and esophagus. Most detected precursors exhibited high-grade dysplasia, with cytologic features resembling those of invasive carcinoma, although some foci of low-grade dysplasia were also detected in the gastroesophageal junction and skin. The second most common (17.0%) cause of false positive results was tissue artifacts, especially crush artifacts (in which non-neoplastic cells are physically crushed during sample preparation, resulting in a characteristic streaming effect of the nuclei), tissue folds and out-of-focus regions. Reactive alterations within the stroma or lymphoid components, constituting 14.9%, and in non-neoplastic epithelial tissue, representing 11.7%, were also responsible for false positive results. A number of these findings, such as biopsy site changes, reactive epithelial atypia, glandular atrophy and acellular stromal mucin, are well-recognized malignant mimics that challenge pathologists as well. Three cases (3.2%) were benign neoplasms misidentified as cancer. These included benign gastrointestinal stromal tumors, hepatic angiomyolipomas and serous cystadenomas of the pancreas. The prediction of biomarkers from standard H&E stained images can reduce the reliance on testing using additional methods and the associated substantial delays in returning results to patients (Fig. ). The status of a biomarker in a specimen is predicted using an aggregator network with the foundation model embeddings as input. These biomarkers play a crucial role in the diagnosis and treatment of various cancers, and each is described in further detail in ‘Biomarker detection’ in (see also Supplementary Table and Fig. ). The biomarker detection datasets consist of WSIs from the histological sections matching the blocks used for DNA extraction and MSK-integrated mutation profiling of actionable targets (MSK-IMPACT) sequencing , the latter of which was analyzed to determine the status of genetic alterations and establish a binary label indicating the presence or absence of the variants: that is, the biomarker (Fig. ). Similar to the pan-cancer evaluation, the publicly available UNI , Phikon and CTransPath models are used as baseline models for comparisons. We note that the biomarker prediction results lacked sufficient statistical power to assess statistically significant differences across models; instead, we conclude relative model performance from evaluating many different biomarker predictions. In our comparative analysis shown in Fig. , Virchow embeddings demonstrated superior performance in seven of the nine evaluated digital biomarkers, achieving AUC scores that exceeded those of the nearest baseline foundation models. This performance underscores the robustness of Virchow embeddings across a diverse range of biomarkers. Even in the categories of prostate–androgen receptor (AR) and ovarian–fraction of genome altered (FGA), where Virchow did not secure the top position, it remained a strong contender, with AUCs of 0.849 and 0.847, respectively. These findings underscore the potential of Virchow embeddings to accurately represent H&E histologic phenotypes, offering predictive insights into biomarkers that are traditionally identified through DNA extraction and MSK-IMPACT sequencing. To directly evaluate tile-level embeddings without the confounder of training an aggregator network, we evaluated Virchow performance on a set of tile-level benchmarks by linear probing. Linear probe evaluation aims to gauge the quality and separability of representations learned by a self-supervised model. We compare Virchow embeddings to baseline model embeddings by applying the same linear probing protocol for each model, using the same training, validation and testing data splits (see ‘Tile-level benchmarking’ in for further details). The analysis is performed both on public datasets and on an internal MSKCC pan-cancer dataset. The internal multitissue dataset for pan-cancer detection at the tile level (referred to as PanMSK) is an in-distribution benchmark, as it is composed of annotations on a held out set of patients across the entire diverse set of tissue groups selected for training (Fig. ). The public datasets are OOD benchmarks and are described in the ‘Tile-level benchmarking’ section in . In addition to UNI , Phikon and CTransPath , DINO p =8 (ref. ) (49 million parameter model trained using The Cancer Genome Atlas (TCGA) and an internal dataset), PLIP (87 million parameter model trained using pathology image-text pairs) and NatImg (1.1 billion parameter model trained on 142 million natural images) are evaluated. As shown in Fig. , Virchow embeddings match or surpass the performance of other embeddings in seven of the eight benchmark tasks (Fig. ; see Supplementary Table for additional metrics). The closest competing models are UNI and Phikon, with UNI scoring in the top 1 three times and in the top 2 for all tasks and Phikon scoring in among the top 2 three times. Virchow demonstrates strong OOD performance as measured by the WILDS and ‘CRC (no norm)’ tasks. The WILDS test data is sourced from a hospital that is not encountered in the training set. The ‘CRC (no norm)’ task introduces a distribution shift from the stain-normalized training set by avoiding stain normalization on the testing set. Without normalization, Virchow’s performance declines by only −0.005 in weighted F 1 score, indicating robustness to variations in data preprocessing. To qualitatively evaluate whether the embeddings learned by Virchow tend to separate the image into semantically meaningful clusters of features, we performed an unsupervised feature analysis similar to the procedure in ref. using the CoNSeP dataset , which contains H&E stained slides of colorectal adenocarcinoma (detailed under ‘Qualitative feature analysis’ in ). We observe approximate semantic segmentation of the cell types in the CoNSeP images (Fig. ). In both examples, the first principal component highlighted malignant epithelium (red) cells. The second principal component, respectively, highlighted miscellaneous cells (yellow) and inflammatory (magenta) cells. DINO v.2 was shown to learn a similar semantic feature separation on natural images, allowing foreground/background separation (for example, discriminating a bus or a bird from the background) as well as part annotation (for example, wheels versus windows in a bus) . Here, we show that this emerging property of the model carries over to the pathology domain. This encouraging result supports our expectation that the unsupervised features learned by Virchow are meaningful and interpretable for a wide range of downstream tasks. The value of a pathology foundation model is twofold: generalizability and training data efficiency. In our study, we demonstrate both of these benefits. Virchow-based pan-cancer prediction generalized well to tissue types or slides submitted from institutions not observed in the training data. Rare histological subtypes of cancer were detected nearly as well as common variants. The same pan-cancer detection model was shown to almost match the performance of clinical-grade models overall (AUC from 0.001 to 0.007 behind clinical products, P < 0.01) and surpassed them in the detection of some rare variants of cancers, despite training with fewer tissue-specific labels. This result is even more impressive when noting that the training dataset of the pan-cancer model, as a proof of concept, lacks the quality control and subpopulation enrichment of data and labels that are typically done for commercially available AI models. Finally, we note that Virchow embeddings were not fine-tuned, and models used simple aggregator architectures to make predictions. These results build confidence that, with sufficient scale, foundation models will serve as the building blocks for the future development of a wide variety of downstream tasks. There are a few areas in which we anticipate particularly high-value impact. In clinical practice, where most biopsy samples are benign, a pan-cancer detection system can prioritize cases to help reduce diagnostic turnaround. With decreasing training data requirements, clinical-grade products for less common cancers could be developed. Biomarker prediction using routine H&E WSIs would increase screening rates; reduce intrusive, tissue-destructive testing; and rapidly provide the data needed to make more informed treatment decisions. Virchow embeddings demonstrated sufficiently high performance to suggest these tools are achievable. Indeed, Virchow unlocks the ability to accurately and precisely detect unusual histological variants of cancer as well as biomarker status, something that is difficult to achieve with cancer- or biomarker-specific training due to the limited amount of associated training data. Despite the observed improvements, there are still aspects of Virchow’s development that merit further discussion. Histopathology data differs from natural image data in key ways: the long-tailed distribution of pathologic entities and histological structures, the lack of object scale diversity and the restricted color space. Self-supervised learning algorithms attempt to match the inductive biases of the learning algorithm to the data distribution; however, in this work, as in many other works in self-supervised learning for computational pathology, algorithmic and training settings are largely based on what was successful in the natural image domain. Further study may reveal that altering these design choices will further improve performance in the pathology domain. It remains an open question at what point the model and data scale are saturated. We found that pan-cancer detection performance scales with model and dataset size (Fig. ), which is consistent with observations of prior foundation models in other domains – . The improvement in performance with respect to model size appears to still be in an approximately log-linear range; however, sub-log-linear trends were observed as a function of training data. Trends in training data size may be oversimplified as they do not capture the tradeoff between increasing the number of WSIs versus tiles. The setting is too complex to draw precise conclusions about the effect of dataset diversity, although we posit that increased diversity helps to learn robust and rare features. Indeed, it has been shown that training a model on multiple tissues or cancer variants can improve detection performance for each cancer , as many morphological features are observed across cancers from different topographies . Overall, our investigation into scaling behavior suggests that increasing the number of model parameters remains a salient axis to explore. Our work has several limitations. The training dataset is acquired from one center with limited scanner types. As with most histopathology self-supervised models, embeddings are generated at the tile level using ×20 magnification (0.5 mpp) as opposed to the slide level and therefore require training an aggregation model. Although scaling up the size of a tile-level foundation model may improve performance, it is likely that such models must be extended to the slide level to achieve the data efficiency required for low-data tasks such as the prediction of biomarkers, treatment response or clinical outcome. A deep investigation of aggregator architectures and training procedures is beyond the scope of this work. As is the case for all models aiming for clinical application, thorough stratified performance validation is required. Furthermore, hardware considerations must be made toward the deployment of models the size of Virchow or larger; model distillation may be appropriate for some tasks. Due to the scale of training, our study has not been able to fully explore the effectiveness of data-balancing and -distillation strategies. The challenge of curating training data that preserves rare features while reducing redundancy remains an open question. Considering the long-tail distribution in digital pathology, we question the suitability of clustering-based data distillation methods such as those used in the original DINO v.2 model for natural images . Recent advances in computational pathology have been supported by increased dataset scale and reduced reliance on labels. Using multiple-instance learning , , with labels at the level of groups of slides has enabled clinically relevant diagnostics by scaling to training datasets on the order of 10,000 WSIs – . These earlier works typically initialized the model’s embedding parameters using pretrained model weights, often those trained on ImageNet in a supervised setting. This process, called transfer learning, was motivated by the observation that model performance critically depends on the model’s ability to capture image features. In-domain transfer learning was not possible given the limited availability of labeled pathology datasets. Now self-supervised learning is enabling in-domain transfer by removing the label requirement, driving a second wave of scaling to tens of thousands of WSIs to inform image representation – , . Virchow marks a major increase in training data scale to 1.5 million WSIs—a volume of data that is over 3,000 times the size of ImageNet as measured by the total number of pixels. This large scale of data in turn motivates large models that can capture the diversity of image features in WSIs. In this work, we have demonstrated that this approach can form the foundation for clinical-grade models in cancer pathology. Million-scale training dataset Institutional review board review was not applicable for the research described in this study. This research study was conducted retrospectively from deidentified data licensed to Paige.AI, Inc. from MSKCC. The data used in this study were all collected originally for clinical use by MSKCC in the practice setting and are therefore considered secondary data. Only data previously deidentified by MSKCC were utilized in the analysis, and unique patient identifiers were completely removed from the analytical dataset. To the best of our knowledge, MSKCC has not transferred any data for which the applicable patient has not consented to or otherwise agreed to MSKCC’s Notice of Privacy Practices or a substantially similar notice, waiver or consent. The training digital pathology dataset comprises 1,488,550 WSIs derived from 119,629 patients. These WSIs are all stained with H&E, a routine stain that stains the nuclei blue and the extracellular matrix and cytoplasm pink. The WSIs are scanned at ×20 resolution or 0.5 mpp using Leica scanners. Seventeen high-level tissue groups are included, as illustrated in Fig. . WSIs are gigapixels in size and are challenging to use directly during training. Instead, Virchow was trained on tissue tiles that were sampled from foreground tissue in each WSI. To detect foreground, each WSI was downsampled 16× with bilinear interpolation, and every pixel of the downsampled image was evaluated as to whether its hue, saturation and value were within [90, 180], [8, 255] and [103, 255], respectively. All non-overlapping 224 × 244 tiles containing at least 25% tissue by area were collected. Virchow was trained on 2 billion tiles sampled randomly with replacement from approximately 13 billion available tissue tiles. Virchow architecture and training Virchow employs the ViT ‘huge’ architecture (ViT-H/14), a ViT with 632 million parameters that was trained using the DINO v.2 (ref. ) self-supervised learning algorithm, as illustrated in Extended Data Fig. . The ViT is an adaptation of the transformer model for image analysis, treating an image as a sequence of patches. These patches are embedded and processed through a transformer encoder that uses self-attention mechanisms. This approach allows ViT to capture complex spatial relationships across the image. DINO v.2 is based on a student–teacher paradigm: given a student network and a teacher network, each using the same architecture, the student is trained to match the representation of the teacher. The student network is information-limited, as it is trained using noisy variations of input tiles. The teacher network is a slowly updated exponential moving average of past student networks; matching the teacher achieves an effect similar to ensembling over prior student predictions . The student learns a global representation of an image by matching the teacher’s class token, as well as local representations by matching the teacher’s patch tokens. Patch tokens are only matched for a select subset of tokens that are randomly masked out of an input image (for the student), as done in masked image modeling . Additional regularization helps DINO v.2 trained models outperform the earlier DINO variant . The default hyperparameters for training the DINO v.2 model were used for Virchow as detailed in ref. with the following changes: a teacher temperature schedule of 0.04–0.07 in 186,000 iterations and a reciprocal square root learning rate schedule with a warmup of 495,000 iterations (instead of 100,000) and linear cooldown to 0.0 for the last 819,200 iterations . Virchow was trained using AdamW ( β 1 = 0.9, β 2 = 0.999) with float16 precision. Note that with ViT-H, we used 131,072 prototypes (and thus 131,072-dimensional projection heads). During distributed training, each mini-batch was sampled by randomly selecting one WSI per graphics processing unit and 256 foreground tiles per WSI. Pan-cancer detection Specimen-level pan-cancer detection requires a model that aggregates foundation model embeddings from all foreground tiles of all WSIs in a specimen to detect the presence of cancer. All pan-cancer detection models trained in this work use an Agata aggregator model, weakly supervised with multiple-instance learning (see Extended Data Fig. for architecture details). Embedding generation For a 224 × 224 input tile image, a Virchow embedding is defined as the concatenation of the class token and the mean across all 256 of the other predicted tokens. This produces an embedding size of 2,560 (1,280 × 2). For Phikon, only the class token is used, as recommended by ref. . For CTransPath, the mean of all tokens is used as there is no class token. Training data To train the aggregator model, we prepared a subset of the training dataset used for training Virchow (see ‘Million-scale training dataset’ in for details), combined with specimen-level labels (block-level for prostate tissue) indicating the presence or absence of cancer extracted from synoptic and diagnostic reports. The training and validation datasets combined consist of 89,417 slides across 40,402 specimens. See Extended Data Fig. for the training data distribution, stratified by WSI tissue type and cancer status. Aggregator training The Agata aggregator was trained as described in Extended Data Fig. . Because the label is at the level of the specimen, all tiles belonging to the same specimen need to be aggregated during training. Training using embeddings for all tiles of a specimen is prohibitively memory-intensive. We thus select the slide with the highest predicted cancer probability per specimen and backpropagate the gradients only for that slide. As baselines, aggregators using Phikon and CTransPath embeddings were also trained. All aggregators were trained for 25 epochs using the cross-entropy loss and the AdamW optimizer with a base learning rate of 0.0003. During each training run, the checkpoint with the highest validation AUC was selected for evaluation. Testing dataset The pan-cancer detection models are evaluated on a combination of data sourced from MSKCC and external institutions. None of the patients in the evaluation set were seen during training. The dataset contains 22,932 slides from 6,142 specimens across 16 cancer types. We hypothesize that the more data the foundation model is trained on, the better the downstream task performance, especially on data-constrained tasks. To test this hypothesis, we categorize cancer types into common or rare cancer groups. According to the NCI, rare cancers are defined as those occurring in fewer than 15 people out of 100,000 each year in the United States . Based on this definition, common cancer comprises 14,179 slides from 3,547 specimens originating in breast, prostate, lung, colon, skin, bladder, uterus, pancreas and H&N, and rare cancer comprises 8,753 slides from 2,595 specimens originating in liver, stomach, brain, ovary, cervix, testis and bone. Note that each cancer type is determined by its tissue of origin and thus may appear in any tissue (as primary or metastatic cancer). On the other hand, benign specimens for each cancer type were sampled only from the tissue of origin. For example, the liver stratum contains 182 liver specimens with liver cancer (primary), 18 non-liver specimens with liver cancer (metastatic) and 200 benign liver specimens. For each cancer type, Fig. shows the distribution between primary and metastatic cancer, and Extended Data Fig. additionally shows the number of benign specimens. The testing dataset includes 15,622 slides from 3,033 specimens collected at MSKCC (denoted as ‘Internal’ in Fig. ), in addition to 7,310 slides (3109 specimens) sent to MSKCC from institutions around the world (‘External’ in Fig. ). See Extended Data Fig. for the testing data distribution, stratified by cancer type (for specimens with cancer) or by tissue type (for benign specimens). Label extraction To establish the clinical cancer diagnosis at the specimen level, a rule-based natural language processing system was employed. This system decomposes case-level reports to the specimen level and analyzes the associated clinical reports with each specimen, thereby providing a comprehensive understanding of each case. Statistical analysis The performance of the three models is compared using two metrics: AUC and specificity at 95% sensitivity. AUC is a suitable general metric because it does not require selecting a threshold for the model’s probability outputs, something that may need tuning for different data subpopulations. Specificity at 95% sensitivity is informative because a clinical system must be not only sensitive but also specific in practice. For AUC, the pairwise DeLong’s test with Holm’s method for correction is applied to check for statistical significance. For specificity, first Cochran’s Q test is applied, and then McNemar’s test is applied post hoc for all pairs with Holm’s method for correction. The two-sided 95% confidence intervals in Fig. and Extended Data Fig. were calculated using DeLong’s method for AUC and Wilson’s method for specificity. In addition to overall analysis, stratified analysis is also conducted for each cancer type. Clinical evaluation datasets To perform an extensive evaluation of the Virchow-based pan-cancer detection model, we employ seven additional datasets (see Supplementary Table for details). One of these datasets is pan-tissue, and the rest are single-tissue datasets containing tissues for which Paige has clinical products: that is, prostate, breast and lymph node. Pan-tissue product benchmark This dataset contains 2,419 slides across 18 tissue types (Supplementary Table ). Each slide is individually inspected by a pathologist and labeled according to presence of invasive cancer. An important distinction between the testing dataset in ‘Pan-cancer detection’ and this dataset is that the former is stratified according to origin tissue in cancerous specimens, whereas the latter is stratified according to tissue type for all slides, as it is more relevant in a clinical setting. We use this dataset to identify failure modes of the pan-cancer detection model. Prostate product benchmark This dataset contains 2,947 blocks (3,327 slides) of prostate needle core biopsies (Supplementary Table ). Labels for the blocks are extracted from synoptic reports collected at MSKCC. This dataset has been curated to evaluate the standalone performance of Paige Prostate Detect, which is a tissue-specific, clinical-grade model. We use this dataset to compare the pan-cancer detection model to Paige Prostate Detect. Prostate rare variants benchmark This dataset contains 28 slides containing rare variants of prostate cancer (neuroendocrine tumor, atrophic, small lymphocytic lymphoma, foamy cell carcinoma, follicular lymphoma) and 112 benign slides (Supplementary Table ). Cancerous slides are curated and labeled by a pathologist, and are appended with slides from benign blocks determined from synoptic reports collected at MSKCC. Breast product benchmark This dataset contains 190 slides with invasive cancer and 1,501 benign slides, labeled individually by a pathologist according to presence of atypical ductal hyperplasia, atypical lobular hyperplasia, lobular carcinoma in situ, ductal carcinoma in situ, invasive ductal carcinoma, invasive lobular carcinoma and/or other subtypes (Supplementary Table ). This dataset has been curated to evaluate the standalone performance of Paige Breast, which is a tissue-specific, clinical-grade model. We use the subtype information for stratified analysis. Breast rare variants benchmark This dataset contains 23 cases of invasive ductal carcinoma or invasive lobular carcinoma (as control), 75 cases of rare variants (adenoid cystic carcinoma, carcinoma with apocrine differentiation, cribriform carcinoma, invasive micropapillary carcinoma, metaplastic carcinoma (matrix producting subtype, spindle cell and squamous cell), mucinous carcinoma, secretory carcinoma and tubular carcinoma) and 392 benign cases (total 5,031 slides). Cancerous cases are curated by a pathologist, and are appended with benign cases determined from synoptic reports collected at MSKCC. See Supplementary Table for details. BLN This dataset contains 458 lymph node slides with metastasized breast cancer and 295 benign lymph node slides (Supplementary Table ). Each slide has been labeled by a pathologist according to presence of invasive cancer, and the largest tumor on the slide is measured to categorize the tumor into macrometastasis, micrometastasis or infiltrating tumor cells. We use the categories for stratified evaluation. Lymph node rare variants benchmark This dataset contains 48 specimens of rare variants of cancers (diffused large B-cell lymphoma, follicular lymphoma, marginal zone lymphoma, Hodgkin’s lymphoma) selected by a pathologist and 192 benign specimens determined from synoptic reports collected at MSKCC (Supplementary Table ). Biomarker detection We formulated each biomarker prediction task as a binary pathology case classification problem, where a positive label indicates the presence of the biomarker. Each case consists of one or more H&E slides that share the same binary label. We randomly split each dataset into training and testing subsets, ensuring no patient overlap, as shown in Supplementary Table . The clinical importance of each biomarker is described below. Colon-MSI Microsatellite instability (MSI) occurs when DNA regions with short, repeated sequences (microsatellites) are disrupted by single nucleotide mutations, leading to variation in these sequences across cells. Normally, mismatch repair (MMR) genes ( MSH1 , MSH2 , MSH6 , PMS2 ) correct these mutations, maintaining consistency in microsatellites. However, inactivation of any MMR gene (through germline mutation, somatic mutation or epigenetic silencing) results in an increased rate of uncorrected mutations across the genome. MSI is detected using polymerase chain reaction or next-generation sequencing, which identifies a high number of unrepaired mutations in microsatellites, indicative of deficient mismatch repair (dMMR). Microsatellite instability high (MSI-H) suggests dMMR in cells, identifiable via IHC, which shows absent staining for MMR proteins. MSI-H is present in approximately 15% of colorectal cancers (CRCs), often linked to germline mutations that elevate hereditary cancer risk. Consequently, routine MSI or IHC-based dMMR screening is recommended for all primary colorectal carcinoma samples. The Colon-MSI dataset, comprising 2,698 CRC samples with 288 MSI-H/dMMR positive cases, uses both IHC and MSK-IMPACT sequencing for dMMR and MSI-H detection, prioritizing IHC results when both test outcomes are available. Breast-CDH1 The biallelic loss of cadherin 1 ( CDH1 ) gene (encoding E-cadherin) is strongly correlated with lobular breast cancer and a distinct histologic phenotype and biologic behavior . CDH1 inactivating mutations associated with loss of heterozygosity or a second somatic loss-of-function mutation as determined by MSK-IMPACT sequencing test results were considered as ‘ CDH1 biallelic mutations’. The CDH1 dataset comprises a total of 1,077 estrogen receptor-positive (ER+) primary breast cancer samples, in which 139 were positive and 918 were negative. The remaining 20 samples with other types of variants—that is, monoallelic mutations—were excluded. Bladder-FGFR The fibroblast growth factor receptor (FGFR) is encoded by four genes ( FGFR1 , FGFR2 , FGFR3 , FGFR4 ). FGFR gene alterations screening in bladder carcinoma allows the identification of patients targetable by FGFR inhibitors. Anecdotal experience from pathologists suggested there may be a morphological signal for FGFR alterations . The FGFR binary label focuses on FGFR3 p.S249C , p.R248C , p.Y373C , p.G370C mutations, FGFR3 - TACC3 fusions and FGFR2 p.N549H , pN549K , p.N549S , p.N549T mutations based on data from the MSK-IMPACT cohort. From the total of 1,038 samples (1,087 WSIs), 26.2% have FGFR3 alterations. Lung-EGFR The EGFR oncogenic mutation screening in non-small cell lung cancer is essential to determine eligibility for targeted therapies in late stage non-small cell lung cancer . The oncogenic status of EGFR mutation was determined based on OncoKB annotation . EGFR mutations with any oncogenic effect (including predicted/likely oncogenic) were defined as positive label, and EGFR mutation with unknown oncogenic status were excluded. Prostate-AR The AR amplification/overexpression was found in 30%–50% of castration resistant prostate cancers and was associated with resistance to androgen deprivation therapy. In the AR dataset, the copy number amplification of AR was determined by MSK-IMPACT sequencing test, for which the fold change was greater than two. Gastric-HER2 Human epidermal growth factor receptor 2 ( HER2 ) overexpression and/or amplification are much more heterogeneous in gastric cancer compared to breast cancer. Approximate 20% of gastric cancer patients are found to correlate with HER2 overexpression/high-level amplification, and they would be likely to benefit from treatment with an anti-HER2 antibody therapy. Here, a HER2 IHC result of 2+, confirmed positive with fluorescence in situ hybridization (FISH) or an IHC result of 3+ were considered HER2 amplification. Endometrial-PTEN PTEN is the most frequently mutated tumor suppressor gene in endometrial cancer. The presence of PTEN mutation showed to be significantly associated with poorer prognosis in survival and disease recurrence. The oncogenic status of PTEN mutation was determined based on MSK-IMPACT sequencing and OncoKB annotation . The variants associated with any oncogenic effect (including predicted and/or likely oncogenic) were defined as positive label for PTEN mutations, and variants with unknown oncogenic status were excluded. Thyroid-RET RET mutations were highly associated with medullary thyroid cancer, which accounts for about 5–10% of all thyroid cancer. Screening RET oncogenic mutations plays an important role in diagnosis and prognosis of medullary thyroid cancer. The positive label for RET oncogenic mutation was determined by MSK-IMPACT sequencing and OncoKB annotation . Skin-BRAF BRAF is one of the most frequently mutated genes in melanoma, and V600E mutation is the most common variant, which leads to constitutive activation of the BRAF/MEK/ERK signaling pathway. Targeted therapy with BRAF inhibitors showed better survival outcome in patients with BRAF V600-mutated melanoma. Therefore, the detection of BRAF V600 mutations in melanoma helps to determine treatment strategies. In the BRAF dataset, the oncogenic mutation status and the presence of V600E variant were determined based on the MSK-IMPACT cohort and OncoKB annotation . Ovarian-FGA High-grade serous ovarian cancer is characterized by high prevalence of TP53 mutations and genome instability with widespread genetic alteration. The fraction of genome altered (FGA) was determined from MSK-IMPACT sequencing data, where FGA ≥ 30% was treated as a positive label. A cut-off for FGA was established that enriched for TP53 mutations in the distribution of ovarian cancer cases. Aggregator training For weakly supervised biomarker prediction, we used embeddings and Agata , as in ‘Pan-cancer detection’, to transform a set of tiles extracted from WSIs that belong to the same case to the case-level target label. Virchow is used to generate tile-level embeddings on all the evaluated datasets with 224 × 224 resolution at ×20 magnification. To thoroughly compare the quality of the embeddings, we trained an aggregator for learning rates in 1 × 10 −4 , 5 × 10 −5 , 1 × 10 −5 , 5 × 10 −6 , 1 × 10 −6 and report the best observed test AUC scores in Fig. . Due to the small biomarker dataset sizes, the learning rate was not chosen on a validation set to evaluate generalization; rather, this serves as a benchmark across the different types of tile embeddings (Virchow, UNI, Phikon and CTransPath), yielding an estimate of the best possible biomarker performance for each type. Statistical analysis AUC is used to compare models without having to select a threshold on the models’ predicted probability values, which may differ by data subpopulation. The two-sided 95% confidence intervals in Fig. are calculated using DeLong’s method . Tile-level benchmarking For evaluating Virchow on tile-sized images, the linear probing protocol, as well as dataset descriptions and the statistical analysis, are described below. Dataset details, including training, validation, and testing splits, are also summarized in Supplementary Table . Linear probing protocol For each experiment, we trained a linear tile classifier with a batch size of 4,096 using the stochastic gradient descent optimizer with a cosine learning rate schedule, from 0.01 to 0, for 12,500 iterations, on top of embeddings generated by a frozen encoder. The large number of iterations is intended to allow any linear classifier to converge as far as it can at each learning rate step along the learning rate schedule. All embeddings were normalized by Z -scoring before classification. Linear probing experiments did not use data augmentation. For testing set evaluation, the classifier checkpoint that achieved the lowest loss on the validation set was selected. A validation set was used for all tasks. If one was not provided with the public dataset, we randomly split out 10% of the training data to make a validation set. PanMSK For a comprehensive in-distribution benchmark, 3,999 slides across the 17 tissue types in Fig. were held out from the training dataset collected from MSKCC. Of these, 1,456 contained cancer that was either partially or exhaustively annotated with segmentation masks by pathologists. These annotations were used to create a tile-level dataset of cancer versus non-cancer classification, which we refer to as PanMSK. All images in PanMSK are 224 × 224 pixel tiles at 0.5 mpp. See Supplementary Note for further details. CRC The CRC classification public dataset contains 100,000 images for training (from which we randomly selected 10,000 for validation) and 7,180 images for testing (224 × 224 pixels) at ×20 magnification sorted into nine morphological classes. Analysis is performed with both the Macenko-stain-normalized (NCT-CRC-HE-100K) and unnormalized (NCT-CRC-HE-100K-NONORM) variants of the dataset. It should be noted that the training set is normalized in both cases, and only the testing test is unnormalized in the latter variant. Thus, the unnormalized variant of CRC involves a distribution shift from training to testing. WILDS The Camelyon17-WILDS dataset is a public dataset comprising 455,954 images, each with a resolution of 96 × 96 pixels, taken at ×10 magnification and downsampled from ×40. This dataset is derived from the larger Camelyon17 dataset and focuses on lymph node metastases. Each image in the dataset is annotated with a binary label indicating the presence or absence of a tumor within the central 32 × 32 pixel region. Uniquely designed to test OOD generalization, the training set (335,996 images) is composed of data from three different hospitals, whereas the validation subset (34,904 images) and testing subset (85,054 images) each originate from separate hospitals not represented in the training data. MHIST The colorectal polyp classification public dataset (MHIST ) contains 3,152 images (224 × 224 pixels) presenting either hyperplastic polyp or sessile serrated adenoma at ×5 magnification (downsampled from ×40 to increase the field of view). This dataset contains 2,175 images in the training subset (of which we randomly selected 217 for validation) and 977 images in the testing subset. TCGA TIL The TCGA TIL public dataset is composed of 304,097 images (100 × 100 pixels) at ×20 magnification – , split into 247,822 training images, 38,601 validation images and 56,275 testing images. Images are considered positive for tumor-infiltrating lymphocytes if at least two TILs are present and labeled negative otherwise. We upsampled the images to 224 × 224 to use with Virchow. PCam The PatchCamelyon (PCam) public dataset consists of 327,680 images (96 × 96 pixels) at ×10 magnification, downsampled from ×40 to increase the field of view , . The data is split into a training subset (262,144 images), a validation subset (32,768 images), and a testing subset (32,768 images). Images are labeled as either cancer or benign. We upsampled the images to 224 × 224 pixels to use with Virchow. MIDOG The MIDOG public dataset consists of 21,806 mitotic and non-mitotic events labeled on 503 7,000 × 5,000 WSI regions from several tumor, species and scanner types . Data was converted into a binary classification task by expanding each 50 × 50 pixel annotation to 224 × 224 regions and then randomly shifting in the horizontal and vertical regions such that the event is not centered in the tile. All negative instances that overlapped with positive instances were removed from the dataset. The resulting dataset consists of training, validation and testing subsets with 13,107, 4,359 and 4,340 images, respectively (of which 6,720, 2,249 and 2,222 have mitotic events, respectively, and the rest contain confounders that mimic mitotic events). TCGA CRC-MSI The TCGA CRC-MSI classification public dataset consists of 51,918 512 × 512 regions taken at ×20 magnification presenting colorectal adenocarcinoma samples . Samples were extracted and annotated from TCGA. Regions were labeled either as microsatellite-instable or microsatellite-stable. We downsampled regions to 448 × 448 to use with Virchow. Statistical analysis The (weighted) F 1 score is used to compare models as this metric is robust to class imbalance. Accuracy and balanced accuracy are also computed, as described in Supplementary Note . The two-sided 95% confidence intervals in Fig. and Supplementary Table were computed with 1,000 bootstrapping iterations over the metrics on the testing set without retraining the classifier. McNemar’s test was used to determine statistically significant ( P < 0.05) differences between results. Qualitative feature analysis We performed an unsupervised feature analysis similar to the procedure in ref. , using the CoNSeP dataset of H&E stained slides with colorectal adenocarcinoma. CoNSeP provides nuclear annotations of cells in the following seven categories: normal epithelial, malignant/dysplastic epithelial, fibroblast, muscle, inflammatory, endothelial and miscellaneous (including necrotic, mitotic and cells that couldn’t be categorized). Because CoNSeP images are of size 1,000 × 1,000 and Virchow takes in images of size 224 × 224, we resized images to 896 × 896 and divided them into a 4 × 4 grid of non-overlapping 224 × 224 subimages before extracting tile-level features. For a given image, we used principal component analysis (PCA) on all the tile features from the subimages, normalized the first and second principal components to values within [0, 1] and thresholded at 0.5. Figure shows some examples of the unsupervised feature separation achieved in this way. Software For data collection, we used Python (v.3.10.11) along with Pandas (v.2.2.2) for indexing the data and metadata used for pretraining and benchmarking. OpenSlide (v.1.3.1) and Pillow (v.10.0.0) were used for preprocessing the image tiles for the benchmark. Where appropriate, we extracted per-specimen labels from clinical reports using DBT (v.1.5.0). We used Python (v.3.10.11) for all experiments and analyses in the study, which can be replicated using open-source libraries as outlined below. For self-supervised pretraining, we used PyTorch (v.2.0.1) and Torchvision (v.0.15.1). The DINO v.2 code was ported from the official repository ( https://github.com/facebookresearch/dinov2 ) and adapted to PyTorch Lightning (v.1.9.0). All WSI processing during pretraining was performed online and was supported by cucim (v.23.10.0) and torchvision (v.0.16.1). For downstream task benchmarking, we use scikit-learn (v.1.4.2) for logistic regression and metrics computation. Implementations of other pretrained visual encoders benchmarked in the study were obtained from the following links: UNI ( https://huggingface.co/MahmoodLab/UNI ), Phikon ( https://huggingface.co/owkin/phikon ), DINOp=8 ( https://github.com/lunit-io/benchmark-ssl-pathology ), PLIP ( https://huggingface.co/vinid/plip ), CTransPath ( https://github.com/Xiyue-Wang/TransPath ) and the original natural image pretrained DINO v.2 ( https://github.com/facebookresearch/dinov2 ). Reporting summary Further information on research design is available in the linked to this article. Institutional review board review was not applicable for the research described in this study. This research study was conducted retrospectively from deidentified data licensed to Paige.AI, Inc. from MSKCC. The data used in this study were all collected originally for clinical use by MSKCC in the practice setting and are therefore considered secondary data. Only data previously deidentified by MSKCC were utilized in the analysis, and unique patient identifiers were completely removed from the analytical dataset. To the best of our knowledge, MSKCC has not transferred any data for which the applicable patient has not consented to or otherwise agreed to MSKCC’s Notice of Privacy Practices or a substantially similar notice, waiver or consent. The training digital pathology dataset comprises 1,488,550 WSIs derived from 119,629 patients. These WSIs are all stained with H&E, a routine stain that stains the nuclei blue and the extracellular matrix and cytoplasm pink. The WSIs are scanned at ×20 resolution or 0.5 mpp using Leica scanners. Seventeen high-level tissue groups are included, as illustrated in Fig. . WSIs are gigapixels in size and are challenging to use directly during training. Instead, Virchow was trained on tissue tiles that were sampled from foreground tissue in each WSI. To detect foreground, each WSI was downsampled 16× with bilinear interpolation, and every pixel of the downsampled image was evaluated as to whether its hue, saturation and value were within [90, 180], [8, 255] and [103, 255], respectively. All non-overlapping 224 × 244 tiles containing at least 25% tissue by area were collected. Virchow was trained on 2 billion tiles sampled randomly with replacement from approximately 13 billion available tissue tiles. Virchow employs the ViT ‘huge’ architecture (ViT-H/14), a ViT with 632 million parameters that was trained using the DINO v.2 (ref. ) self-supervised learning algorithm, as illustrated in Extended Data Fig. . The ViT is an adaptation of the transformer model for image analysis, treating an image as a sequence of patches. These patches are embedded and processed through a transformer encoder that uses self-attention mechanisms. This approach allows ViT to capture complex spatial relationships across the image. DINO v.2 is based on a student–teacher paradigm: given a student network and a teacher network, each using the same architecture, the student is trained to match the representation of the teacher. The student network is information-limited, as it is trained using noisy variations of input tiles. The teacher network is a slowly updated exponential moving average of past student networks; matching the teacher achieves an effect similar to ensembling over prior student predictions . The student learns a global representation of an image by matching the teacher’s class token, as well as local representations by matching the teacher’s patch tokens. Patch tokens are only matched for a select subset of tokens that are randomly masked out of an input image (for the student), as done in masked image modeling . Additional regularization helps DINO v.2 trained models outperform the earlier DINO variant . The default hyperparameters for training the DINO v.2 model were used for Virchow as detailed in ref. with the following changes: a teacher temperature schedule of 0.04–0.07 in 186,000 iterations and a reciprocal square root learning rate schedule with a warmup of 495,000 iterations (instead of 100,000) and linear cooldown to 0.0 for the last 819,200 iterations . Virchow was trained using AdamW ( β 1 = 0.9, β 2 = 0.999) with float16 precision. Note that with ViT-H, we used 131,072 prototypes (and thus 131,072-dimensional projection heads). During distributed training, each mini-batch was sampled by randomly selecting one WSI per graphics processing unit and 256 foreground tiles per WSI. Specimen-level pan-cancer detection requires a model that aggregates foundation model embeddings from all foreground tiles of all WSIs in a specimen to detect the presence of cancer. All pan-cancer detection models trained in this work use an Agata aggregator model, weakly supervised with multiple-instance learning (see Extended Data Fig. for architecture details). Embedding generation For a 224 × 224 input tile image, a Virchow embedding is defined as the concatenation of the class token and the mean across all 256 of the other predicted tokens. This produces an embedding size of 2,560 (1,280 × 2). For Phikon, only the class token is used, as recommended by ref. . For CTransPath, the mean of all tokens is used as there is no class token. Training data To train the aggregator model, we prepared a subset of the training dataset used for training Virchow (see ‘Million-scale training dataset’ in for details), combined with specimen-level labels (block-level for prostate tissue) indicating the presence or absence of cancer extracted from synoptic and diagnostic reports. The training and validation datasets combined consist of 89,417 slides across 40,402 specimens. See Extended Data Fig. for the training data distribution, stratified by WSI tissue type and cancer status. Aggregator training The Agata aggregator was trained as described in Extended Data Fig. . Because the label is at the level of the specimen, all tiles belonging to the same specimen need to be aggregated during training. Training using embeddings for all tiles of a specimen is prohibitively memory-intensive. We thus select the slide with the highest predicted cancer probability per specimen and backpropagate the gradients only for that slide. As baselines, aggregators using Phikon and CTransPath embeddings were also trained. All aggregators were trained for 25 epochs using the cross-entropy loss and the AdamW optimizer with a base learning rate of 0.0003. During each training run, the checkpoint with the highest validation AUC was selected for evaluation. Testing dataset The pan-cancer detection models are evaluated on a combination of data sourced from MSKCC and external institutions. None of the patients in the evaluation set were seen during training. The dataset contains 22,932 slides from 6,142 specimens across 16 cancer types. We hypothesize that the more data the foundation model is trained on, the better the downstream task performance, especially on data-constrained tasks. To test this hypothesis, we categorize cancer types into common or rare cancer groups. According to the NCI, rare cancers are defined as those occurring in fewer than 15 people out of 100,000 each year in the United States . Based on this definition, common cancer comprises 14,179 slides from 3,547 specimens originating in breast, prostate, lung, colon, skin, bladder, uterus, pancreas and H&N, and rare cancer comprises 8,753 slides from 2,595 specimens originating in liver, stomach, brain, ovary, cervix, testis and bone. Note that each cancer type is determined by its tissue of origin and thus may appear in any tissue (as primary or metastatic cancer). On the other hand, benign specimens for each cancer type were sampled only from the tissue of origin. For example, the liver stratum contains 182 liver specimens with liver cancer (primary), 18 non-liver specimens with liver cancer (metastatic) and 200 benign liver specimens. For each cancer type, Fig. shows the distribution between primary and metastatic cancer, and Extended Data Fig. additionally shows the number of benign specimens. The testing dataset includes 15,622 slides from 3,033 specimens collected at MSKCC (denoted as ‘Internal’ in Fig. ), in addition to 7,310 slides (3109 specimens) sent to MSKCC from institutions around the world (‘External’ in Fig. ). See Extended Data Fig. for the testing data distribution, stratified by cancer type (for specimens with cancer) or by tissue type (for benign specimens). Label extraction To establish the clinical cancer diagnosis at the specimen level, a rule-based natural language processing system was employed. This system decomposes case-level reports to the specimen level and analyzes the associated clinical reports with each specimen, thereby providing a comprehensive understanding of each case. Statistical analysis The performance of the three models is compared using two metrics: AUC and specificity at 95% sensitivity. AUC is a suitable general metric because it does not require selecting a threshold for the model’s probability outputs, something that may need tuning for different data subpopulations. Specificity at 95% sensitivity is informative because a clinical system must be not only sensitive but also specific in practice. For AUC, the pairwise DeLong’s test with Holm’s method for correction is applied to check for statistical significance. For specificity, first Cochran’s Q test is applied, and then McNemar’s test is applied post hoc for all pairs with Holm’s method for correction. The two-sided 95% confidence intervals in Fig. and Extended Data Fig. were calculated using DeLong’s method for AUC and Wilson’s method for specificity. In addition to overall analysis, stratified analysis is also conducted for each cancer type. For a 224 × 224 input tile image, a Virchow embedding is defined as the concatenation of the class token and the mean across all 256 of the other predicted tokens. This produces an embedding size of 2,560 (1,280 × 2). For Phikon, only the class token is used, as recommended by ref. . For CTransPath, the mean of all tokens is used as there is no class token. To train the aggregator model, we prepared a subset of the training dataset used for training Virchow (see ‘Million-scale training dataset’ in for details), combined with specimen-level labels (block-level for prostate tissue) indicating the presence or absence of cancer extracted from synoptic and diagnostic reports. The training and validation datasets combined consist of 89,417 slides across 40,402 specimens. See Extended Data Fig. for the training data distribution, stratified by WSI tissue type and cancer status. The Agata aggregator was trained as described in Extended Data Fig. . Because the label is at the level of the specimen, all tiles belonging to the same specimen need to be aggregated during training. Training using embeddings for all tiles of a specimen is prohibitively memory-intensive. We thus select the slide with the highest predicted cancer probability per specimen and backpropagate the gradients only for that slide. As baselines, aggregators using Phikon and CTransPath embeddings were also trained. All aggregators were trained for 25 epochs using the cross-entropy loss and the AdamW optimizer with a base learning rate of 0.0003. During each training run, the checkpoint with the highest validation AUC was selected for evaluation. The pan-cancer detection models are evaluated on a combination of data sourced from MSKCC and external institutions. None of the patients in the evaluation set were seen during training. The dataset contains 22,932 slides from 6,142 specimens across 16 cancer types. We hypothesize that the more data the foundation model is trained on, the better the downstream task performance, especially on data-constrained tasks. To test this hypothesis, we categorize cancer types into common or rare cancer groups. According to the NCI, rare cancers are defined as those occurring in fewer than 15 people out of 100,000 each year in the United States . Based on this definition, common cancer comprises 14,179 slides from 3,547 specimens originating in breast, prostate, lung, colon, skin, bladder, uterus, pancreas and H&N, and rare cancer comprises 8,753 slides from 2,595 specimens originating in liver, stomach, brain, ovary, cervix, testis and bone. Note that each cancer type is determined by its tissue of origin and thus may appear in any tissue (as primary or metastatic cancer). On the other hand, benign specimens for each cancer type were sampled only from the tissue of origin. For example, the liver stratum contains 182 liver specimens with liver cancer (primary), 18 non-liver specimens with liver cancer (metastatic) and 200 benign liver specimens. For each cancer type, Fig. shows the distribution between primary and metastatic cancer, and Extended Data Fig. additionally shows the number of benign specimens. The testing dataset includes 15,622 slides from 3,033 specimens collected at MSKCC (denoted as ‘Internal’ in Fig. ), in addition to 7,310 slides (3109 specimens) sent to MSKCC from institutions around the world (‘External’ in Fig. ). See Extended Data Fig. for the testing data distribution, stratified by cancer type (for specimens with cancer) or by tissue type (for benign specimens). To establish the clinical cancer diagnosis at the specimen level, a rule-based natural language processing system was employed. This system decomposes case-level reports to the specimen level and analyzes the associated clinical reports with each specimen, thereby providing a comprehensive understanding of each case. The performance of the three models is compared using two metrics: AUC and specificity at 95% sensitivity. AUC is a suitable general metric because it does not require selecting a threshold for the model’s probability outputs, something that may need tuning for different data subpopulations. Specificity at 95% sensitivity is informative because a clinical system must be not only sensitive but also specific in practice. For AUC, the pairwise DeLong’s test with Holm’s method for correction is applied to check for statistical significance. For specificity, first Cochran’s Q test is applied, and then McNemar’s test is applied post hoc for all pairs with Holm’s method for correction. The two-sided 95% confidence intervals in Fig. and Extended Data Fig. were calculated using DeLong’s method for AUC and Wilson’s method for specificity. In addition to overall analysis, stratified analysis is also conducted for each cancer type. To perform an extensive evaluation of the Virchow-based pan-cancer detection model, we employ seven additional datasets (see Supplementary Table for details). One of these datasets is pan-tissue, and the rest are single-tissue datasets containing tissues for which Paige has clinical products: that is, prostate, breast and lymph node. Pan-tissue product benchmark This dataset contains 2,419 slides across 18 tissue types (Supplementary Table ). Each slide is individually inspected by a pathologist and labeled according to presence of invasive cancer. An important distinction between the testing dataset in ‘Pan-cancer detection’ and this dataset is that the former is stratified according to origin tissue in cancerous specimens, whereas the latter is stratified according to tissue type for all slides, as it is more relevant in a clinical setting. We use this dataset to identify failure modes of the pan-cancer detection model. Prostate product benchmark This dataset contains 2,947 blocks (3,327 slides) of prostate needle core biopsies (Supplementary Table ). Labels for the blocks are extracted from synoptic reports collected at MSKCC. This dataset has been curated to evaluate the standalone performance of Paige Prostate Detect, which is a tissue-specific, clinical-grade model. We use this dataset to compare the pan-cancer detection model to Paige Prostate Detect. Prostate rare variants benchmark This dataset contains 28 slides containing rare variants of prostate cancer (neuroendocrine tumor, atrophic, small lymphocytic lymphoma, foamy cell carcinoma, follicular lymphoma) and 112 benign slides (Supplementary Table ). Cancerous slides are curated and labeled by a pathologist, and are appended with slides from benign blocks determined from synoptic reports collected at MSKCC. Breast product benchmark This dataset contains 190 slides with invasive cancer and 1,501 benign slides, labeled individually by a pathologist according to presence of atypical ductal hyperplasia, atypical lobular hyperplasia, lobular carcinoma in situ, ductal carcinoma in situ, invasive ductal carcinoma, invasive lobular carcinoma and/or other subtypes (Supplementary Table ). This dataset has been curated to evaluate the standalone performance of Paige Breast, which is a tissue-specific, clinical-grade model. We use the subtype information for stratified analysis. Breast rare variants benchmark This dataset contains 23 cases of invasive ductal carcinoma or invasive lobular carcinoma (as control), 75 cases of rare variants (adenoid cystic carcinoma, carcinoma with apocrine differentiation, cribriform carcinoma, invasive micropapillary carcinoma, metaplastic carcinoma (matrix producting subtype, spindle cell and squamous cell), mucinous carcinoma, secretory carcinoma and tubular carcinoma) and 392 benign cases (total 5,031 slides). Cancerous cases are curated by a pathologist, and are appended with benign cases determined from synoptic reports collected at MSKCC. See Supplementary Table for details. BLN This dataset contains 458 lymph node slides with metastasized breast cancer and 295 benign lymph node slides (Supplementary Table ). Each slide has been labeled by a pathologist according to presence of invasive cancer, and the largest tumor on the slide is measured to categorize the tumor into macrometastasis, micrometastasis or infiltrating tumor cells. We use the categories for stratified evaluation. Lymph node rare variants benchmark This dataset contains 48 specimens of rare variants of cancers (diffused large B-cell lymphoma, follicular lymphoma, marginal zone lymphoma, Hodgkin’s lymphoma) selected by a pathologist and 192 benign specimens determined from synoptic reports collected at MSKCC (Supplementary Table ). This dataset contains 2,419 slides across 18 tissue types (Supplementary Table ). Each slide is individually inspected by a pathologist and labeled according to presence of invasive cancer. An important distinction between the testing dataset in ‘Pan-cancer detection’ and this dataset is that the former is stratified according to origin tissue in cancerous specimens, whereas the latter is stratified according to tissue type for all slides, as it is more relevant in a clinical setting. We use this dataset to identify failure modes of the pan-cancer detection model. This dataset contains 2,947 blocks (3,327 slides) of prostate needle core biopsies (Supplementary Table ). Labels for the blocks are extracted from synoptic reports collected at MSKCC. This dataset has been curated to evaluate the standalone performance of Paige Prostate Detect, which is a tissue-specific, clinical-grade model. We use this dataset to compare the pan-cancer detection model to Paige Prostate Detect. This dataset contains 28 slides containing rare variants of prostate cancer (neuroendocrine tumor, atrophic, small lymphocytic lymphoma, foamy cell carcinoma, follicular lymphoma) and 112 benign slides (Supplementary Table ). Cancerous slides are curated and labeled by a pathologist, and are appended with slides from benign blocks determined from synoptic reports collected at MSKCC. This dataset contains 190 slides with invasive cancer and 1,501 benign slides, labeled individually by a pathologist according to presence of atypical ductal hyperplasia, atypical lobular hyperplasia, lobular carcinoma in situ, ductal carcinoma in situ, invasive ductal carcinoma, invasive lobular carcinoma and/or other subtypes (Supplementary Table ). This dataset has been curated to evaluate the standalone performance of Paige Breast, which is a tissue-specific, clinical-grade model. We use the subtype information for stratified analysis. This dataset contains 23 cases of invasive ductal carcinoma or invasive lobular carcinoma (as control), 75 cases of rare variants (adenoid cystic carcinoma, carcinoma with apocrine differentiation, cribriform carcinoma, invasive micropapillary carcinoma, metaplastic carcinoma (matrix producting subtype, spindle cell and squamous cell), mucinous carcinoma, secretory carcinoma and tubular carcinoma) and 392 benign cases (total 5,031 slides). Cancerous cases are curated by a pathologist, and are appended with benign cases determined from synoptic reports collected at MSKCC. See Supplementary Table for details. This dataset contains 458 lymph node slides with metastasized breast cancer and 295 benign lymph node slides (Supplementary Table ). Each slide has been labeled by a pathologist according to presence of invasive cancer, and the largest tumor on the slide is measured to categorize the tumor into macrometastasis, micrometastasis or infiltrating tumor cells. We use the categories for stratified evaluation. This dataset contains 48 specimens of rare variants of cancers (diffused large B-cell lymphoma, follicular lymphoma, marginal zone lymphoma, Hodgkin’s lymphoma) selected by a pathologist and 192 benign specimens determined from synoptic reports collected at MSKCC (Supplementary Table ). We formulated each biomarker prediction task as a binary pathology case classification problem, where a positive label indicates the presence of the biomarker. Each case consists of one or more H&E slides that share the same binary label. We randomly split each dataset into training and testing subsets, ensuring no patient overlap, as shown in Supplementary Table . The clinical importance of each biomarker is described below. Colon-MSI Microsatellite instability (MSI) occurs when DNA regions with short, repeated sequences (microsatellites) are disrupted by single nucleotide mutations, leading to variation in these sequences across cells. Normally, mismatch repair (MMR) genes ( MSH1 , MSH2 , MSH6 , PMS2 ) correct these mutations, maintaining consistency in microsatellites. However, inactivation of any MMR gene (through germline mutation, somatic mutation or epigenetic silencing) results in an increased rate of uncorrected mutations across the genome. MSI is detected using polymerase chain reaction or next-generation sequencing, which identifies a high number of unrepaired mutations in microsatellites, indicative of deficient mismatch repair (dMMR). Microsatellite instability high (MSI-H) suggests dMMR in cells, identifiable via IHC, which shows absent staining for MMR proteins. MSI-H is present in approximately 15% of colorectal cancers (CRCs), often linked to germline mutations that elevate hereditary cancer risk. Consequently, routine MSI or IHC-based dMMR screening is recommended for all primary colorectal carcinoma samples. The Colon-MSI dataset, comprising 2,698 CRC samples with 288 MSI-H/dMMR positive cases, uses both IHC and MSK-IMPACT sequencing for dMMR and MSI-H detection, prioritizing IHC results when both test outcomes are available. Breast-CDH1 The biallelic loss of cadherin 1 ( CDH1 ) gene (encoding E-cadherin) is strongly correlated with lobular breast cancer and a distinct histologic phenotype and biologic behavior . CDH1 inactivating mutations associated with loss of heterozygosity or a second somatic loss-of-function mutation as determined by MSK-IMPACT sequencing test results were considered as ‘ CDH1 biallelic mutations’. The CDH1 dataset comprises a total of 1,077 estrogen receptor-positive (ER+) primary breast cancer samples, in which 139 were positive and 918 were negative. The remaining 20 samples with other types of variants—that is, monoallelic mutations—were excluded. Bladder-FGFR The fibroblast growth factor receptor (FGFR) is encoded by four genes ( FGFR1 , FGFR2 , FGFR3 , FGFR4 ). FGFR gene alterations screening in bladder carcinoma allows the identification of patients targetable by FGFR inhibitors. Anecdotal experience from pathologists suggested there may be a morphological signal for FGFR alterations . The FGFR binary label focuses on FGFR3 p.S249C , p.R248C , p.Y373C , p.G370C mutations, FGFR3 - TACC3 fusions and FGFR2 p.N549H , pN549K , p.N549S , p.N549T mutations based on data from the MSK-IMPACT cohort. From the total of 1,038 samples (1,087 WSIs), 26.2% have FGFR3 alterations. Lung-EGFR The EGFR oncogenic mutation screening in non-small cell lung cancer is essential to determine eligibility for targeted therapies in late stage non-small cell lung cancer . The oncogenic status of EGFR mutation was determined based on OncoKB annotation . EGFR mutations with any oncogenic effect (including predicted/likely oncogenic) were defined as positive label, and EGFR mutation with unknown oncogenic status were excluded. Prostate-AR The AR amplification/overexpression was found in 30%–50% of castration resistant prostate cancers and was associated with resistance to androgen deprivation therapy. In the AR dataset, the copy number amplification of AR was determined by MSK-IMPACT sequencing test, for which the fold change was greater than two. Gastric-HER2 Human epidermal growth factor receptor 2 ( HER2 ) overexpression and/or amplification are much more heterogeneous in gastric cancer compared to breast cancer. Approximate 20% of gastric cancer patients are found to correlate with HER2 overexpression/high-level amplification, and they would be likely to benefit from treatment with an anti-HER2 antibody therapy. Here, a HER2 IHC result of 2+, confirmed positive with fluorescence in situ hybridization (FISH) or an IHC result of 3+ were considered HER2 amplification. Endometrial-PTEN PTEN is the most frequently mutated tumor suppressor gene in endometrial cancer. The presence of PTEN mutation showed to be significantly associated with poorer prognosis in survival and disease recurrence. The oncogenic status of PTEN mutation was determined based on MSK-IMPACT sequencing and OncoKB annotation . The variants associated with any oncogenic effect (including predicted and/or likely oncogenic) were defined as positive label for PTEN mutations, and variants with unknown oncogenic status were excluded. Thyroid-RET RET mutations were highly associated with medullary thyroid cancer, which accounts for about 5–10% of all thyroid cancer. Screening RET oncogenic mutations plays an important role in diagnosis and prognosis of medullary thyroid cancer. The positive label for RET oncogenic mutation was determined by MSK-IMPACT sequencing and OncoKB annotation . Skin-BRAF BRAF is one of the most frequently mutated genes in melanoma, and V600E mutation is the most common variant, which leads to constitutive activation of the BRAF/MEK/ERK signaling pathway. Targeted therapy with BRAF inhibitors showed better survival outcome in patients with BRAF V600-mutated melanoma. Therefore, the detection of BRAF V600 mutations in melanoma helps to determine treatment strategies. In the BRAF dataset, the oncogenic mutation status and the presence of V600E variant were determined based on the MSK-IMPACT cohort and OncoKB annotation . Ovarian-FGA High-grade serous ovarian cancer is characterized by high prevalence of TP53 mutations and genome instability with widespread genetic alteration. The fraction of genome altered (FGA) was determined from MSK-IMPACT sequencing data, where FGA ≥ 30% was treated as a positive label. A cut-off for FGA was established that enriched for TP53 mutations in the distribution of ovarian cancer cases. Aggregator training For weakly supervised biomarker prediction, we used embeddings and Agata , as in ‘Pan-cancer detection’, to transform a set of tiles extracted from WSIs that belong to the same case to the case-level target label. Virchow is used to generate tile-level embeddings on all the evaluated datasets with 224 × 224 resolution at ×20 magnification. To thoroughly compare the quality of the embeddings, we trained an aggregator for learning rates in 1 × 10 −4 , 5 × 10 −5 , 1 × 10 −5 , 5 × 10 −6 , 1 × 10 −6 and report the best observed test AUC scores in Fig. . Due to the small biomarker dataset sizes, the learning rate was not chosen on a validation set to evaluate generalization; rather, this serves as a benchmark across the different types of tile embeddings (Virchow, UNI, Phikon and CTransPath), yielding an estimate of the best possible biomarker performance for each type. Statistical analysis AUC is used to compare models without having to select a threshold on the models’ predicted probability values, which may differ by data subpopulation. The two-sided 95% confidence intervals in Fig. are calculated using DeLong’s method . Microsatellite instability (MSI) occurs when DNA regions with short, repeated sequences (microsatellites) are disrupted by single nucleotide mutations, leading to variation in these sequences across cells. Normally, mismatch repair (MMR) genes ( MSH1 , MSH2 , MSH6 , PMS2 ) correct these mutations, maintaining consistency in microsatellites. However, inactivation of any MMR gene (through germline mutation, somatic mutation or epigenetic silencing) results in an increased rate of uncorrected mutations across the genome. MSI is detected using polymerase chain reaction or next-generation sequencing, which identifies a high number of unrepaired mutations in microsatellites, indicative of deficient mismatch repair (dMMR). Microsatellite instability high (MSI-H) suggests dMMR in cells, identifiable via IHC, which shows absent staining for MMR proteins. MSI-H is present in approximately 15% of colorectal cancers (CRCs), often linked to germline mutations that elevate hereditary cancer risk. Consequently, routine MSI or IHC-based dMMR screening is recommended for all primary colorectal carcinoma samples. The Colon-MSI dataset, comprising 2,698 CRC samples with 288 MSI-H/dMMR positive cases, uses both IHC and MSK-IMPACT sequencing for dMMR and MSI-H detection, prioritizing IHC results when both test outcomes are available. The biallelic loss of cadherin 1 ( CDH1 ) gene (encoding E-cadherin) is strongly correlated with lobular breast cancer and a distinct histologic phenotype and biologic behavior . CDH1 inactivating mutations associated with loss of heterozygosity or a second somatic loss-of-function mutation as determined by MSK-IMPACT sequencing test results were considered as ‘ CDH1 biallelic mutations’. The CDH1 dataset comprises a total of 1,077 estrogen receptor-positive (ER+) primary breast cancer samples, in which 139 were positive and 918 were negative. The remaining 20 samples with other types of variants—that is, monoallelic mutations—were excluded. The fibroblast growth factor receptor (FGFR) is encoded by four genes ( FGFR1 , FGFR2 , FGFR3 , FGFR4 ). FGFR gene alterations screening in bladder carcinoma allows the identification of patients targetable by FGFR inhibitors. Anecdotal experience from pathologists suggested there may be a morphological signal for FGFR alterations . The FGFR binary label focuses on FGFR3 p.S249C , p.R248C , p.Y373C , p.G370C mutations, FGFR3 - TACC3 fusions and FGFR2 p.N549H , pN549K , p.N549S , p.N549T mutations based on data from the MSK-IMPACT cohort. From the total of 1,038 samples (1,087 WSIs), 26.2% have FGFR3 alterations. The EGFR oncogenic mutation screening in non-small cell lung cancer is essential to determine eligibility for targeted therapies in late stage non-small cell lung cancer . The oncogenic status of EGFR mutation was determined based on OncoKB annotation . EGFR mutations with any oncogenic effect (including predicted/likely oncogenic) were defined as positive label, and EGFR mutation with unknown oncogenic status were excluded. The AR amplification/overexpression was found in 30%–50% of castration resistant prostate cancers and was associated with resistance to androgen deprivation therapy. In the AR dataset, the copy number amplification of AR was determined by MSK-IMPACT sequencing test, for which the fold change was greater than two. Human epidermal growth factor receptor 2 ( HER2 ) overexpression and/or amplification are much more heterogeneous in gastric cancer compared to breast cancer. Approximate 20% of gastric cancer patients are found to correlate with HER2 overexpression/high-level amplification, and they would be likely to benefit from treatment with an anti-HER2 antibody therapy. Here, a HER2 IHC result of 2+, confirmed positive with fluorescence in situ hybridization (FISH) or an IHC result of 3+ were considered HER2 amplification. PTEN is the most frequently mutated tumor suppressor gene in endometrial cancer. The presence of PTEN mutation showed to be significantly associated with poorer prognosis in survival and disease recurrence. The oncogenic status of PTEN mutation was determined based on MSK-IMPACT sequencing and OncoKB annotation . The variants associated with any oncogenic effect (including predicted and/or likely oncogenic) were defined as positive label for PTEN mutations, and variants with unknown oncogenic status were excluded. RET mutations were highly associated with medullary thyroid cancer, which accounts for about 5–10% of all thyroid cancer. Screening RET oncogenic mutations plays an important role in diagnosis and prognosis of medullary thyroid cancer. The positive label for RET oncogenic mutation was determined by MSK-IMPACT sequencing and OncoKB annotation . BRAF is one of the most frequently mutated genes in melanoma, and V600E mutation is the most common variant, which leads to constitutive activation of the BRAF/MEK/ERK signaling pathway. Targeted therapy with BRAF inhibitors showed better survival outcome in patients with BRAF V600-mutated melanoma. Therefore, the detection of BRAF V600 mutations in melanoma helps to determine treatment strategies. In the BRAF dataset, the oncogenic mutation status and the presence of V600E variant were determined based on the MSK-IMPACT cohort and OncoKB annotation . High-grade serous ovarian cancer is characterized by high prevalence of TP53 mutations and genome instability with widespread genetic alteration. The fraction of genome altered (FGA) was determined from MSK-IMPACT sequencing data, where FGA ≥ 30% was treated as a positive label. A cut-off for FGA was established that enriched for TP53 mutations in the distribution of ovarian cancer cases. For weakly supervised biomarker prediction, we used embeddings and Agata , as in ‘Pan-cancer detection’, to transform a set of tiles extracted from WSIs that belong to the same case to the case-level target label. Virchow is used to generate tile-level embeddings on all the evaluated datasets with 224 × 224 resolution at ×20 magnification. To thoroughly compare the quality of the embeddings, we trained an aggregator for learning rates in 1 × 10 −4 , 5 × 10 −5 , 1 × 10 −5 , 5 × 10 −6 , 1 × 10 −6 and report the best observed test AUC scores in Fig. . Due to the small biomarker dataset sizes, the learning rate was not chosen on a validation set to evaluate generalization; rather, this serves as a benchmark across the different types of tile embeddings (Virchow, UNI, Phikon and CTransPath), yielding an estimate of the best possible biomarker performance for each type. AUC is used to compare models without having to select a threshold on the models’ predicted probability values, which may differ by data subpopulation. The two-sided 95% confidence intervals in Fig. are calculated using DeLong’s method . For evaluating Virchow on tile-sized images, the linear probing protocol, as well as dataset descriptions and the statistical analysis, are described below. Dataset details, including training, validation, and testing splits, are also summarized in Supplementary Table . Linear probing protocol For each experiment, we trained a linear tile classifier with a batch size of 4,096 using the stochastic gradient descent optimizer with a cosine learning rate schedule, from 0.01 to 0, for 12,500 iterations, on top of embeddings generated by a frozen encoder. The large number of iterations is intended to allow any linear classifier to converge as far as it can at each learning rate step along the learning rate schedule. All embeddings were normalized by Z -scoring before classification. Linear probing experiments did not use data augmentation. For testing set evaluation, the classifier checkpoint that achieved the lowest loss on the validation set was selected. A validation set was used for all tasks. If one was not provided with the public dataset, we randomly split out 10% of the training data to make a validation set. PanMSK For a comprehensive in-distribution benchmark, 3,999 slides across the 17 tissue types in Fig. were held out from the training dataset collected from MSKCC. Of these, 1,456 contained cancer that was either partially or exhaustively annotated with segmentation masks by pathologists. These annotations were used to create a tile-level dataset of cancer versus non-cancer classification, which we refer to as PanMSK. All images in PanMSK are 224 × 224 pixel tiles at 0.5 mpp. See Supplementary Note for further details. CRC The CRC classification public dataset contains 100,000 images for training (from which we randomly selected 10,000 for validation) and 7,180 images for testing (224 × 224 pixels) at ×20 magnification sorted into nine morphological classes. Analysis is performed with both the Macenko-stain-normalized (NCT-CRC-HE-100K) and unnormalized (NCT-CRC-HE-100K-NONORM) variants of the dataset. It should be noted that the training set is normalized in both cases, and only the testing test is unnormalized in the latter variant. Thus, the unnormalized variant of CRC involves a distribution shift from training to testing. WILDS The Camelyon17-WILDS dataset is a public dataset comprising 455,954 images, each with a resolution of 96 × 96 pixels, taken at ×10 magnification and downsampled from ×40. This dataset is derived from the larger Camelyon17 dataset and focuses on lymph node metastases. Each image in the dataset is annotated with a binary label indicating the presence or absence of a tumor within the central 32 × 32 pixel region. Uniquely designed to test OOD generalization, the training set (335,996 images) is composed of data from three different hospitals, whereas the validation subset (34,904 images) and testing subset (85,054 images) each originate from separate hospitals not represented in the training data. MHIST The colorectal polyp classification public dataset (MHIST ) contains 3,152 images (224 × 224 pixels) presenting either hyperplastic polyp or sessile serrated adenoma at ×5 magnification (downsampled from ×40 to increase the field of view). This dataset contains 2,175 images in the training subset (of which we randomly selected 217 for validation) and 977 images in the testing subset. TCGA TIL The TCGA TIL public dataset is composed of 304,097 images (100 × 100 pixels) at ×20 magnification – , split into 247,822 training images, 38,601 validation images and 56,275 testing images. Images are considered positive for tumor-infiltrating lymphocytes if at least two TILs are present and labeled negative otherwise. We upsampled the images to 224 × 224 to use with Virchow. PCam The PatchCamelyon (PCam) public dataset consists of 327,680 images (96 × 96 pixels) at ×10 magnification, downsampled from ×40 to increase the field of view , . The data is split into a training subset (262,144 images), a validation subset (32,768 images), and a testing subset (32,768 images). Images are labeled as either cancer or benign. We upsampled the images to 224 × 224 pixels to use with Virchow. MIDOG The MIDOG public dataset consists of 21,806 mitotic and non-mitotic events labeled on 503 7,000 × 5,000 WSI regions from several tumor, species and scanner types . Data was converted into a binary classification task by expanding each 50 × 50 pixel annotation to 224 × 224 regions and then randomly shifting in the horizontal and vertical regions such that the event is not centered in the tile. All negative instances that overlapped with positive instances were removed from the dataset. The resulting dataset consists of training, validation and testing subsets with 13,107, 4,359 and 4,340 images, respectively (of which 6,720, 2,249 and 2,222 have mitotic events, respectively, and the rest contain confounders that mimic mitotic events). TCGA CRC-MSI The TCGA CRC-MSI classification public dataset consists of 51,918 512 × 512 regions taken at ×20 magnification presenting colorectal adenocarcinoma samples . Samples were extracted and annotated from TCGA. Regions were labeled either as microsatellite-instable or microsatellite-stable. We downsampled regions to 448 × 448 to use with Virchow. Statistical analysis The (weighted) F 1 score is used to compare models as this metric is robust to class imbalance. Accuracy and balanced accuracy are also computed, as described in Supplementary Note . The two-sided 95% confidence intervals in Fig. and Supplementary Table were computed with 1,000 bootstrapping iterations over the metrics on the testing set without retraining the classifier. McNemar’s test was used to determine statistically significant ( P < 0.05) differences between results. For each experiment, we trained a linear tile classifier with a batch size of 4,096 using the stochastic gradient descent optimizer with a cosine learning rate schedule, from 0.01 to 0, for 12,500 iterations, on top of embeddings generated by a frozen encoder. The large number of iterations is intended to allow any linear classifier to converge as far as it can at each learning rate step along the learning rate schedule. All embeddings were normalized by Z -scoring before classification. Linear probing experiments did not use data augmentation. For testing set evaluation, the classifier checkpoint that achieved the lowest loss on the validation set was selected. A validation set was used for all tasks. If one was not provided with the public dataset, we randomly split out 10% of the training data to make a validation set. For a comprehensive in-distribution benchmark, 3,999 slides across the 17 tissue types in Fig. were held out from the training dataset collected from MSKCC. Of these, 1,456 contained cancer that was either partially or exhaustively annotated with segmentation masks by pathologists. These annotations were used to create a tile-level dataset of cancer versus non-cancer classification, which we refer to as PanMSK. All images in PanMSK are 224 × 224 pixel tiles at 0.5 mpp. See Supplementary Note for further details. The CRC classification public dataset contains 100,000 images for training (from which we randomly selected 10,000 for validation) and 7,180 images for testing (224 × 224 pixels) at ×20 magnification sorted into nine morphological classes. Analysis is performed with both the Macenko-stain-normalized (NCT-CRC-HE-100K) and unnormalized (NCT-CRC-HE-100K-NONORM) variants of the dataset. It should be noted that the training set is normalized in both cases, and only the testing test is unnormalized in the latter variant. Thus, the unnormalized variant of CRC involves a distribution shift from training to testing. The Camelyon17-WILDS dataset is a public dataset comprising 455,954 images, each with a resolution of 96 × 96 pixels, taken at ×10 magnification and downsampled from ×40. This dataset is derived from the larger Camelyon17 dataset and focuses on lymph node metastases. Each image in the dataset is annotated with a binary label indicating the presence or absence of a tumor within the central 32 × 32 pixel region. Uniquely designed to test OOD generalization, the training set (335,996 images) is composed of data from three different hospitals, whereas the validation subset (34,904 images) and testing subset (85,054 images) each originate from separate hospitals not represented in the training data. The colorectal polyp classification public dataset (MHIST ) contains 3,152 images (224 × 224 pixels) presenting either hyperplastic polyp or sessile serrated adenoma at ×5 magnification (downsampled from ×40 to increase the field of view). This dataset contains 2,175 images in the training subset (of which we randomly selected 217 for validation) and 977 images in the testing subset. The TCGA TIL public dataset is composed of 304,097 images (100 × 100 pixels) at ×20 magnification – , split into 247,822 training images, 38,601 validation images and 56,275 testing images. Images are considered positive for tumor-infiltrating lymphocytes if at least two TILs are present and labeled negative otherwise. We upsampled the images to 224 × 224 to use with Virchow. The PatchCamelyon (PCam) public dataset consists of 327,680 images (96 × 96 pixels) at ×10 magnification, downsampled from ×40 to increase the field of view , . The data is split into a training subset (262,144 images), a validation subset (32,768 images), and a testing subset (32,768 images). Images are labeled as either cancer or benign. We upsampled the images to 224 × 224 pixels to use with Virchow. The MIDOG public dataset consists of 21,806 mitotic and non-mitotic events labeled on 503 7,000 × 5,000 WSI regions from several tumor, species and scanner types . Data was converted into a binary classification task by expanding each 50 × 50 pixel annotation to 224 × 224 regions and then randomly shifting in the horizontal and vertical regions such that the event is not centered in the tile. All negative instances that overlapped with positive instances were removed from the dataset. The resulting dataset consists of training, validation and testing subsets with 13,107, 4,359 and 4,340 images, respectively (of which 6,720, 2,249 and 2,222 have mitotic events, respectively, and the rest contain confounders that mimic mitotic events). The TCGA CRC-MSI classification public dataset consists of 51,918 512 × 512 regions taken at ×20 magnification presenting colorectal adenocarcinoma samples . Samples were extracted and annotated from TCGA. Regions were labeled either as microsatellite-instable or microsatellite-stable. We downsampled regions to 448 × 448 to use with Virchow. The (weighted) F 1 score is used to compare models as this metric is robust to class imbalance. Accuracy and balanced accuracy are also computed, as described in Supplementary Note . The two-sided 95% confidence intervals in Fig. and Supplementary Table were computed with 1,000 bootstrapping iterations over the metrics on the testing set without retraining the classifier. McNemar’s test was used to determine statistically significant ( P < 0.05) differences between results. We performed an unsupervised feature analysis similar to the procedure in ref. , using the CoNSeP dataset of H&E stained slides with colorectal adenocarcinoma. CoNSeP provides nuclear annotations of cells in the following seven categories: normal epithelial, malignant/dysplastic epithelial, fibroblast, muscle, inflammatory, endothelial and miscellaneous (including necrotic, mitotic and cells that couldn’t be categorized). Because CoNSeP images are of size 1,000 × 1,000 and Virchow takes in images of size 224 × 224, we resized images to 896 × 896 and divided them into a 4 × 4 grid of non-overlapping 224 × 224 subimages before extracting tile-level features. For a given image, we used principal component analysis (PCA) on all the tile features from the subimages, normalized the first and second principal components to values within [0, 1] and thresholded at 0.5. Figure shows some examples of the unsupervised feature separation achieved in this way. For data collection, we used Python (v.3.10.11) along with Pandas (v.2.2.2) for indexing the data and metadata used for pretraining and benchmarking. OpenSlide (v.1.3.1) and Pillow (v.10.0.0) were used for preprocessing the image tiles for the benchmark. Where appropriate, we extracted per-specimen labels from clinical reports using DBT (v.1.5.0). We used Python (v.3.10.11) for all experiments and analyses in the study, which can be replicated using open-source libraries as outlined below. For self-supervised pretraining, we used PyTorch (v.2.0.1) and Torchvision (v.0.15.1). The DINO v.2 code was ported from the official repository ( https://github.com/facebookresearch/dinov2 ) and adapted to PyTorch Lightning (v.1.9.0). All WSI processing during pretraining was performed online and was supported by cucim (v.23.10.0) and torchvision (v.0.16.1). For downstream task benchmarking, we use scikit-learn (v.1.4.2) for logistic regression and metrics computation. Implementations of other pretrained visual encoders benchmarked in the study were obtained from the following links: UNI ( https://huggingface.co/MahmoodLab/UNI ), Phikon ( https://huggingface.co/owkin/phikon ), DINOp=8 ( https://github.com/lunit-io/benchmark-ssl-pathology ), PLIP ( https://huggingface.co/vinid/plip ), CTransPath ( https://github.com/Xiyue-Wang/TransPath ) and the original natural image pretrained DINO v.2 ( https://github.com/facebookresearch/dinov2 ). Further information on research design is available in the linked to this article. Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41591-024-03141-0. Supplementary Information The document is split into ‘Supplementary Notes’ sections that may or may not contain descriptive text along with tables and figures. Section 1 ‘Early foundation models in computational pathology’ describes pathology foundation models in the literature. Table 1.1 summarizes these models. Section 2 ‘Clinical Evaluation’, Table 2.1 summarizes datasets used in clinical eval of Virchow (results in Fig. ). Tables 2.2–2.8: per-dataset tables detailing stratified sample counts for the datasets in Table 2.1. Section 3: Table 3.1 describes slide- and case-level sample counts for the biomarker prediction datasets (train/tune/test). Section 4 ‘Tile-level benchmarks’ describes additional metrics used to evaluate tile-level linear probing results. Table 4.1 describes the tile benchmark datasets. Table 4.2: tile-level results (additional metrics). Table 4.3: TCGA CRC-MSI tile-level biomarker prediction results (this task is not in the main figure). Table 4.4: PCAM and WILDS using a pretrained CNN from a breast-specialized model. Section 5 ‘Multitissue PanMSK dataset’ describes the preparation of the internal PanMSK tile-level benchmark dataset in detail. Table 5.1: PanMSK data splits. Fig. 5.1: tile distribution (tissues, cancer/benign). Reporting Summary |
Pediatric advocacy: Advancement in academic institutions | c905aa58-ab57-4ff2-8d50-f06de695d1c6 | 11126388 | Pediatrics[mh] | Children, adolescents, and families face an array of health-related challenges, many shaped by state or federal policies that may create or perpetuate racial inequities. The promising 15-year decline in childhood poverty to 11% in 2019 changed course during the COVID-19 pandemic. In January 2022, the overall child poverty rate rose to 17% with higher rates seen in children who identify as Black (25.4%) and Latino (23.9%). Similarly, it is projected that approximately 5 million children will lose health insurance with the “Medicaid unwinding” at the end of the public health emergency, with a disproportionate impact on Hispanic and Black individuals. , Escalating rates of anxiety and depression coupled with limited access to mental health treatment and scarce inpatient beds led to the 2021 Declaration of National Emergency for Child and Adolescent Mental Health issued by the American Academy of Pediatrics, the American Academy of Child and Adolescent Psychiatry and the Children’s Hospital Association. Firearm-related deaths emerged as the leading cause of mortality in youth while threats to reproductive health care and gender-affirming care have become a harsh reality. These complex and varied threats to child well-being highlight the need for pediatricians to develop community and policy initiatives to elevate research-based best practices and mitigate the impact on children. The need for pediatricians to serve as child advocates has never been more pressing. Unlike physicians who care exclusively for adults, pediatricians are well-positioned to advocate on behalf of children. Pediatricians see and bear witness to the impact of the social determinants of health on children. Pediatricians also enjoy uniquely high levels of trust, particularly because of the longitudinal continuity of care provided. Finally, pediatricians are able to contextualize the evidence-based recommendations when interacting with policymakers. If we want pediatricians of the future to practice advocacy they must encounter it during their training, both in focused educational settings and being role-modeled in AMCs where most training occurs. Given the urgent need to advance the advocacy efforts on behalf of children, how do academic pediatricians align this critical work with AMCs? In an effort to understand the value placed on advocacy, we surveyed pediatric department chairs across the United States. Results indicate academic pediatric chairs felt increasing importance of advocacy in their departments over time, more than 75% felt it was important for faculty overall, 55% thought it was “important” or “very important” for promotion and 62% believed supporting advocacy was “important” or “very important.” Chairs expressed an abundance of barriers to advocacy activities by faculty in their departments, including sustainability for advocacy such as funding and protected time, and uncertainty around their role in advocacy as a pediatric department within larger health systems. Further work is underway to understand how health systems support and dissuade physician advocacy. Pediatric departments can advance advocacy activities through alignment, training, and creating innovative funding opportunities and leadership positions. First, strategic planning should focus on aligning faculty advocacy endeavors with hospital and department missions. The use of shared language in the hospital community benefit report as well as annual Chair reports, allows for alignment between faculty efforts within larger missions, making a clearer case for investment in the work. Second, formal assistance is necessary as many faculty lack curricular training, mentorship, or protected time to engage in advocacy work. AMCs should provide faculty training and resources through professional development curricula, mentorship programs, and internal grant-funded opportunities. Investing in structural support for faculty advocacy activities will lead to sustainable advocacy. Third, once the framework and opportunities are established, advocacy efforts must also be aligned with not only traditional academic values but also traditional academic structures. This includes the development of leadership positions such as Vice Chairs and department advocacy leaders, which when complemented with a scholarly approach to the work, is inclusive of a pathway forward to promotion. Finally, departments can support the translation of advocacy work into faculty academic promotion using an advocacy portfolio (AP). , Modeled after the success of the educator’s portfolio, the AP can provide a clear roadmap by incorporating advocacy work into the promotion domains of scholarship/research, education, and service/clinical. One example involves Duke University School of Medicine Appointments, Promotion and Tenure that, in 2020, debuted a new framework for promotion at the faculty, department, and institutional level that values the traditional areas of primary focus (research, clinical, education) while also valuing the many “expressions of scholarship” of faculty that uphold institutional values. They define advocacy scholarship as “scholarly activity that promotes the social, economic, educational, and political changes that ameliorate threats to human health and advance the well-being of people.” As with traditional scholarship, work cited is required to highlight proof of excellence and evidence of a scholarly approach. Faculty identify advocacy-specific scholarly areas cited in the AP: advocacy engagement, community outreach, knowledge dissemination, advocacy teaching/mentoring, and advocacy leadership/administration. Additional forms of high-impact advocacy work beyond peer-reviewed manuscripts include co-authorship of policy statements/legislative briefs/consensus statements, legislative testimony, development of public health initiatives that become standard of care, participation in local and regional task forces, and establishment of community partnerships. There is a need to develop a scholarly framework and advocacy metrics to standardize how this work is viewed across academic pediatrics to allow the field to evolve in an evidence-based way. Documenting quantifiable evidence of advocacy work enables appointment and promotion committee members to evaluate the scholarly quality of such work for advancement. Nerlinger et al describe the use of a logic model of demonstrating quality and quantity inputs to measure impacts and outcomes as one way to measure the scholarship of advocacy initiatives. A case example of using a logic model for advocacy in improving outcomes for asthma follows. Imagine a 5-year-old female with poorly controlled moderate persistent asthma, who lives in a household with mold and has inadequate access to her medications. This leads to increased utilization of emergency room and urgent care visits for asthma exacerbation with resultant hospitalization for status asthmaticus. The logic model maps inputs and outcomes, including the number of referrals made to community-based organizations, including medical-legal partnerships, to aid with remediating housing conditions for this child. Providing testimony for mold to be added to the list of state housing code violations is an example of high-impact advocacy. Working with managed care health plans to ensure access to evidence-based asthma medications demonstrates a second policy intervention. Additionally, systematically partnering with the community health workforce in asthma remediation is an example of community-engaged advocacy that can measurably improve patient outcomes. Finally, traditional dissemination of these advocacy efforts in abstracts, oral presentations, or publications is also possible. Using this stepwise approach is important because attribution of success in advocacy may take many years and an often-circuitous path, especially for legislative and policy efforts. Ultimately, faculty advocacy activities and scholarship require institutional support for effectiveness and sustainability. As advocacy is an emerging component of the academic health system model, long-term institutional buy-in will require both champions in senior leadership and grassroots efforts (Table ). Medical education is increasingly embracing the integral role of health advocacy and the closely related importance of community collaboration as essential components in improving health systems and patient outcomes. In 2021, leaders from the Association of American Medical Colleges (AAMC) wrote “Embracing community collaborations as academic medicine’s fourth mission provides an opportunity to reimagine what optimal health can be—together….and make meaningful progress toward achieving health justice for all. Now is our time to act.” In order to promote changes that allow faculty to thrive as health advocates, institutions need to implement substantive changes and investments and/or expand their perspective on promotion, career development, and community engagement. Institutional leaders at all levels may face barriers to achieving this vision and need to rely on various theories and styles of change management. One particularly relevant approach, adaptive leadership, is designed to address complex and longstanding challenges and is centered on building capacity across all organizational levels. This approach leverages past successes while at the same time prioritizing innovation and diversity of thought, staff, and expertise; it recognizes that any successful “adaptation” preserves what is essential, reorganizes what no longer best meets an organization’s needs, and creates new opportunities to allow an organization or individual to thrive. Using this framework, several examples follow of activities that may be particularly impactful in advancing a sustainable commitment to advocacy within academic institutions, though there are many other potentially effective activities and strategies. Preserving what is essential: First and foremost, it is important to identify and articulate the overarching sense of purpose and how day-to-day activities are influenced by these guiding principles. Ensuring that their team is reminded of and centered on this sense of purpose will prepare advocates for the potentially hard work ahead and ensure they stay true to your mission. It is critical for advocates to understand and convey that successful advocacy results in systemic change to benefit children and families, their community, and potentially their institution or practice, including new or expanded funding. Better meeting an organization’s needs with existing resources: By focusing on three key areas—increased relationship building with elected officials and partner advocates, enhancing communication (including with media), and aligning existing resources – leaders can strengthen existing infrastructure and provide more focus on desired activities and outcomes. To achieve success in these areas, it is useful to think about who potential champions are and/or who is likely to be influential in achieving their goals work to better understand their priorities and how they may align with the advocate’s vision for change. Sometimes relatively minor adjustments within an organization can have a large impact. Creating new opportunities : The cultivation of supplemental funding sources can provide opportunities for innovation and be an important lever for achieving buy-in and leveraging additional institutional support. Some funding opportunities may be solely focused on health advocacy and/or community engagement, but it is also useful to look for opportunities to partner with others and enhance their work. Successful advocacy is demanding and complex, as are the multitude of problems we are challenged to address, from gun violence to mental health. Pediatrics must always maintain a priority focus on what is best for children and families and be ready to adopt innovative approaches to advance child well-being. Leveraging institutional leadership support, providing a framework and pathway for faculty advancement through a scholarly approach, and strengthening partnerships, funding, and communications will continue to advance successful advocacy efforts. |
Short-term outcomes of laparoscopic central hepatectomy: a comparison with open surgery | 7e816dbb-81cc-4b22-ad5b-1e87c62b3181 | 11842538 | Surgical Procedures, Operative[mh] | Laparoscopic liver resection (LLR) was first reported in 1991 by Reich et al. . With the rapid advancement of laparoscopic techniques and equipment, the rate of LLR has been increasing worldwide, and the indications for LLR have expanded . Increasing numbers of authors have reported the efficacy of LLR, which is associated with reduced blood loss, lower transfusion requirements, and shorter length of hospital stay . Centrally located tumors (in segments 4, 5, and 8) require central hepatectomy (CH), which involves central bisectionectomy (CBS) or right anterior sectionectomy (AS). Although CH has the advantage of preserving sufficient remnant liver volume, it involves two liver dissection planes, and the surgical procedure is therefore complicated. Thus, CH has been reported as a surgical procedure associated with a high risk of bile leakage due to exposure of major Glissonean structures and the extensive liver dissection planes . There have been some recent reports on laparoscopic CH (LCH) , although LCH carries a high degree of difficulty . LCH can only be achieved with a high level of expertise in laparoscopic liver surgery. Further, the safety and feasibility of this procedure have not been established yet. Therefore, the aim of this study was to retrospectively evaluate the intra- and short-term outcomes of laparoscopic vs. open CH cases in our department and assess the safety and efficacy of LCH. Patients and methods CH was defined as CBS and AS requiring resection of the right anterior Glissonean pedicle (excluding cases with biliary reconstruction). In total, 38 patients underwent CH at the Department of Gastroenterological and Pediatric Surgery, Oita University Faculty of Medicine between January 2010 to November 2023. At our institution, we perform approximately 60 liver resections annually, with laparoscopic procedures accounting for about 80% of them. Our team includes board-certified surgeons recognized by the Japanese Society of Hepato-Biliary-Pancreatic Surgery for advanced skills and by the Japan Society for Endoscopic Surgery for expertise in laparoscopic liver resection, thus ensuring high-quality surgical outcomes. We collected information on age, sex, body mass index (BMI), tumor characteristics, tumor proximity (< 1 cm) to the hepatic vein or the root of the right anterior Glissonean pedicle, clinical data, histopathology, surgical methods, operative time, blood loss, length of postoperative hospital stay, and postoperative complications from the patients’ medical records. We retrospectively compared these data between the open CH (OCH) and LCH groups to assess the safety and efficacy of laparoscopic surgery. This study was approved by the Ethics Committee of Oita University Faculty of Medicine (No. 1601). Informed consent was obtained in the form of opt-out on the hospital web site. Surgical procedure A video of our standardized laparoscopic CBS procedure is presented in Video 1. In all laparoscopic cases, the operations were started with the patient in the left half-lateral decubitus position, and the first port was inserted through a supraumbilical incision by the open method. Four other ports were inserted in the right upper quadrant of the abdomen. At our institution, liver resection is performed with the intermittent Pringle maneuver in all cases. First, after cholecystectomy, the right anterior Glissonean pedicle is secured, and a clamp test is performed to confirm the demarcation line. For AS, the cranioventral approach is used to dissect the liver along the middle hepatic vein from its root toward the root of the right anterior Glissonean pedicle. For CBS, the left transection plane between segment 4 and the left lateral lobe is dissected along the falciform ligament on the surface of the liver. Then, branches of segment 4b and segment 4a are ligated and gradually transected. Dissection is continued toward the top of the left plane, and the roots of the left hepatic and middle hepatic veins are exposed. At the same time, at the hepatic hilum, dissection between segment 4 and the caudate lobes is advanced along the hepatic hilar plate toward the root of the right anterior Glissonean pedicle. After dissection around the root of the right anterior Glissonean pedicle is fully developed, the right anterior Glissonean pedicle is dissected with a stapling device. Finally, the liver is dissected in one direction toward the right side so as to connect the left wall of the right hepatic vein and the demarcation line on the surface, which then completes the dissection. Statistical analysis Continuous data are expressed as the median value and range (lowest and highest), and categorical data are expressed as counts, with the associated percentile value calculated. The Wilcoxon rank sum test was used to compare continuous data, and Pearson’s χ 2 test was used for categorical data. A multivariate analysis was performed using a Cox proportional hazard model to identify independent risk factors of bile leakage after CH. A p-value < 0.05 was considered to indicate statistical significance. Multivariate analysis was performed for factors with a p-value < 0.10 on univariate analysis. All statistical analyses were performed with JMP software version 17.2 for MAC (SAS Institute, Cary, NC, USA). CH was defined as CBS and AS requiring resection of the right anterior Glissonean pedicle (excluding cases with biliary reconstruction). In total, 38 patients underwent CH at the Department of Gastroenterological and Pediatric Surgery, Oita University Faculty of Medicine between January 2010 to November 2023. At our institution, we perform approximately 60 liver resections annually, with laparoscopic procedures accounting for about 80% of them. Our team includes board-certified surgeons recognized by the Japanese Society of Hepato-Biliary-Pancreatic Surgery for advanced skills and by the Japan Society for Endoscopic Surgery for expertise in laparoscopic liver resection, thus ensuring high-quality surgical outcomes. We collected information on age, sex, body mass index (BMI), tumor characteristics, tumor proximity (< 1 cm) to the hepatic vein or the root of the right anterior Glissonean pedicle, clinical data, histopathology, surgical methods, operative time, blood loss, length of postoperative hospital stay, and postoperative complications from the patients’ medical records. We retrospectively compared these data between the open CH (OCH) and LCH groups to assess the safety and efficacy of laparoscopic surgery. This study was approved by the Ethics Committee of Oita University Faculty of Medicine (No. 1601). Informed consent was obtained in the form of opt-out on the hospital web site. A video of our standardized laparoscopic CBS procedure is presented in Video 1. In all laparoscopic cases, the operations were started with the patient in the left half-lateral decubitus position, and the first port was inserted through a supraumbilical incision by the open method. Four other ports were inserted in the right upper quadrant of the abdomen. At our institution, liver resection is performed with the intermittent Pringle maneuver in all cases. First, after cholecystectomy, the right anterior Glissonean pedicle is secured, and a clamp test is performed to confirm the demarcation line. For AS, the cranioventral approach is used to dissect the liver along the middle hepatic vein from its root toward the root of the right anterior Glissonean pedicle. For CBS, the left transection plane between segment 4 and the left lateral lobe is dissected along the falciform ligament on the surface of the liver. Then, branches of segment 4b and segment 4a are ligated and gradually transected. Dissection is continued toward the top of the left plane, and the roots of the left hepatic and middle hepatic veins are exposed. At the same time, at the hepatic hilum, dissection between segment 4 and the caudate lobes is advanced along the hepatic hilar plate toward the root of the right anterior Glissonean pedicle. After dissection around the root of the right anterior Glissonean pedicle is fully developed, the right anterior Glissonean pedicle is dissected with a stapling device. Finally, the liver is dissected in one direction toward the right side so as to connect the left wall of the right hepatic vein and the demarcation line on the surface, which then completes the dissection. Continuous data are expressed as the median value and range (lowest and highest), and categorical data are expressed as counts, with the associated percentile value calculated. The Wilcoxon rank sum test was used to compare continuous data, and Pearson’s χ 2 test was used for categorical data. A multivariate analysis was performed using a Cox proportional hazard model to identify independent risk factors of bile leakage after CH. A p-value < 0.05 was considered to indicate statistical significance. Multivariate analysis was performed for factors with a p-value < 0.10 on univariate analysis. All statistical analyses were performed with JMP software version 17.2 for MAC (SAS Institute, Cary, NC, USA). Comparison of patient clinical characteristics Clinical characteristics of the patients can be compared between the OCH and LCH groups in Table . There were no differences in tumor condition and liver function between the two groups, although there were a greater number of females in the LCH group. Comparison of surgical outcomes Surgical outcomes between the LCH and OCH groups can be compared in Table . There was one case of open conversion in the LCH group. There was no difference in operative time between the two groups, but the LCH group had significantly less blood loss and lower transfusion rates. The postoperative outcomes are shown in Table . There was no difference in the incidence of bile leakage between the two groups (33% vs. 42%; p = 0.42). The incidences of ascites (0% vs. 17%; p = 0.047) and surgical site infection (SSI) (0% vs. 21%; p = 0.02) were significantly lower in the LCH group. There were no significant differences in other complications. Postoperative hospital stay was significantly shorter in the LCH group (14 vs. 30 days; p = 0.005). Risk factors for bile leakage in central hepatectomy CH is a procedure associated with a high incidence of bile leakage, and thus the risk factors for bile leakage were investigated. Univariate and multivariate analysis both revealed proximity of the tumor to the right anterior Glissonean pedicle to be the significant risk factor for bile leakage (odds ratio, 6.84; 95% confidence interval, 1.67–32.7; p = 0.01)(Table ). Clinical characteristics of the patients can be compared between the OCH and LCH groups in Table . There were no differences in tumor condition and liver function between the two groups, although there were a greater number of females in the LCH group. Surgical outcomes between the LCH and OCH groups can be compared in Table . There was one case of open conversion in the LCH group. There was no difference in operative time between the two groups, but the LCH group had significantly less blood loss and lower transfusion rates. The postoperative outcomes are shown in Table . There was no difference in the incidence of bile leakage between the two groups (33% vs. 42%; p = 0.42). The incidences of ascites (0% vs. 17%; p = 0.047) and surgical site infection (SSI) (0% vs. 21%; p = 0.02) were significantly lower in the LCH group. There were no significant differences in other complications. Postoperative hospital stay was significantly shorter in the LCH group (14 vs. 30 days; p = 0.005). CH is a procedure associated with a high incidence of bile leakage, and thus the risk factors for bile leakage were investigated. Univariate and multivariate analysis both revealed proximity of the tumor to the right anterior Glissonean pedicle to be the significant risk factor for bile leakage (odds ratio, 6.84; 95% confidence interval, 1.67–32.7; p = 0.01)(Table ). In this study, we retrospectively reviewed cases of LCH and OCH performed in our department to assess the safety and efficacy of laparoscopic surgery. Compared to the OCH group, the LCH group showed significantly less blood loss, lower transfusion rate, and no difference in operative time. Among postoperative complications, the rate of bile leakage was 33% in the LCH group and 42% in the OCH group, but the difference was not significant. The incidences of ascites and SSI were significantly lower in the LCH group, and postoperative hospital stay was significantly shorter in the LCH group. Univariate and multivariate analysis of risk factors for bile leakage after CH showed tumor proximity of < 1 cm to the right anterior Glissonean pedicle to be the significant risk factor. LCH is a highly complex procedure that requires extensive training . At our institution, we establish surgeon qualification criteria based on a scoring system for LLR . We begin with partial resections for low-difficulty anterolateral lesions and progressively advance to more complex liver resections in a step-by-step manner. To perform major hepatectomy such as LCH, we believe that, as reported by Kuemmerli et al. and Hasegawa et al. , experience with at least 60 cases of LLR is essential. While studies have investigated international benchmarks for the outcomes of LLR , no such benchmarks have been established for LCH. Other previous studies have reported the outcomes of LCH compared to OCH . Among the short-term surgical outcomes, there was no difference in blood loss and the incidence of postoperative complications between the LCH and OCH groups, whereas the LCH group had a longer operative time and shorter postoperative hospital stay . Cho et al. reported an overall postoperative morbidity rate of 30% and a bile leakage rate of 15% in their LCH group . In our study as well, the LCH group experienced less blood loss, lower incidences of ascites and SSI, and a shorter postoperative hospital stay, consistent with these previous reports. However, the incidence of postoperative bile leakage in our patients was higher than that previously reported, so we investigated the risk factors for bile leakage. To our knowledge, there have been no reports investigating the causes of bile leakage in LCH, making the present findings novel in this respect. In general, LLR is reported to be associated with fewer complications compared to open hepatectomy, but the incidence of bile leakage has been reported to be comparable when limited to CH . CH is considered to be a surgical procedure with a high incidence of bile leakage . Nanashima et al. reported an incidence rate of bile leakage of 37% , whereas Ueno et al. reported a rate of 44.8% . Several reports have examined the risk factors for bile leakage in CH . Ueno et al. examined the biliary complications in CH and reported that the tumor proximity to the right anterior Glissonean pedicle and a longer right hepatic duct were risk factors for biliary complications . In our study as well, proximity of the tumor to the right anterior Glissonean pedicle was found to be the risk factor for bile leakage. This factor could be the potential cause for stenosis of the right posterior bile duct (RPBD) branch. Yoon et al. investigated the factors contributing to the occurrence of the RPBD stenosis in CH . Approximately 80% of the RPBD is located supraportally distal to the bifurcation of the anterior and posterior Glissonean sheaths. When the anterior Glissonean sheath is ligated, injuries such as stricture and ischemic insult of the RPBD could occur. They reported that handling the Glissonean pedicle close to the right anterior Glissonean pedicle root could lead to stenosis of the RPBD. This stenosis increases the intraductal pressure in the RPBD, potentially causing bile leakage from the liver dissection plane. Additionally, in cases in which the tumor is closer to the hepatic hilum, there is a higher possibility of incising the caudate lobe and damaging its bile ducts. This is consistent with the findings of Nanashima et al., who reported that segment 1 resection and inferior vena cava compression were risk factors for bile leakage with CBS . These may be the reasons why a tumor close to the root of the right anterior Glissonean pedicle were detected as risk factors for bile leakage. Recently, the robotic approach has been increasingly adopted in liver surgery. Robotic-assisted liver resection (RLR) offers several advantages over LLR, including enhanced precision, reduced blood loss, lower conversion rates, and higher R0 resection rates . These benefits stem from the robotic system’s superior visualization, increased instrument dexterity, and improved maneuverability. However, the widespread adoption of RLR remains limited due to its higher costs, longer learning curve, and the lack of significant differences in long-term oncological outcomes compared to LLR . While its utility in complex liver resections, such as CH, appears promising, further studies are required to validate its advantages. The present study has several limitations. First, this is a single-center study with a small number of cases. Second, the operator was different in each case although the surgical team included expert board-certified surgeons. Third, long-term prognoses were not comparable because of the variety of diseases in the patients included in this study. Compared to OCH, LCH reduced blood loss and could be performed safely. Although the frequency of postoperative bile leakage was not different between these two procedures, there was less postoperative ascites and fewer SSIs with LCH, suggesting the potential for a shortened postoperative hospital stay. However, proximity of the tumor to the root of the right anterior Glissonean pedicle was considered a risk factor for postoperative bile leakage, and caution is required in such cases. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Spiritual Care Experiences of Nurses Working in Gynecology Clinic: A Qualitative Research | 743c7d28-9c69-4558-af30-b200599e8a2b | 11831960 | Gynaecology[mh] | Introduction Pregnancy and childbirth are among the most important events in a woman's reproductive life. During this period, women experience complex emotions such as happiness, faith, hope, anxiety, and fear at the same time. In this process with full of uncertainties, obstetrics and gynecology nurses who provide primary care to women play a key role . Considering that pregnancy and birth are some of the most stressful life events in every woman's life, the nurse needs to assess the factors required to cope with the woman's stress and make it an enjoyable event. In recent years, it has been emphasized that pregnancy and birth are considered as a spiritual experience, and spiritual care is necessary to prepare women for birth . Spiritual care is defined as the type of care provided to meet the patient's religious and existential needs regarding the meaning and purpose of life . In 1998, the World Health Organization recognized spiritual health as the fourth dimension of health after physical, psychological, and social health to meet the needs of patients . Moreover, the importance of meeting the spiritual care needs of patients has been emphasized internationally and included in various care guidelines . Spiritual care is an important part of holistic care and an indicator of quality of care. Unmet spiritual needs lead to spiritual distress such as loneliness, decreased quality of life, hopelessness, decreased sense of spiritual peace, and depression . Patient‐centered spiritual care should promote human contact through a compassionate relationship that encourages hope and comfort expressed through beliefs, values, traditions, and practices . Nurses, who play an important role in patients' spiritual care, positively impact patients in spiritual care to improve their psychological coping skills, satisfaction, inner spiritual strength, and quality of life . v The American Nursing Association emphasized that the spiritual care competencies of nurses should be evaluated . Nurses' spiritual care competencies are affected by many factors, such as age and education. Internationally, nurses show a lack of training and mentoring in spiritual care, express their inadequacy in assessing and addressing the spiritual domain, and express uncertainty about whether patients' spiritual care needs are met . In a cross‐sectional study conducted by Kiaei et al. (2015), nurses stated that their knowledge about spiritual care was limited, they felt inadequate in providing spiritual care, and they did not receive adequate training . In particular, gynecology nurses, who play an essential role in maternal and neonatal health, need to have the necessary skills and professional competencies to include spirituality in nursing practices effectively and safely. In the literature, there is a lack of research on the provision of care for the spiritual needs that gynecology nurses may encounter and how they provide care for the spiritual needs of service users from the perspective of gynecology nurses. In light of this information, this study aimed to investigate the spiritual care experiences of nurses working in the gynecology clinic. Methods 2.1 Research Type, Location and Time This qualitative research was conducted between April 2 and May 20, 2024, with nurses with at least 1 year of clinical experience in gynecology actively working in hospitals in different provinces of Turkey. The study was carried out using the descriptive phenomenology type of phenomenological design, one of the qualitative research methods. Phenomenological design is a research method that enables people to express their understanding, feelings, perspectives, and perceptions about a particular phenomenon or concept. It describes how they experience this phenomenon . The primary purpose of phenomenology is to understand the essence of the object by reducing individual experiences about an event to a universal explanation . Creswell (2016) discusses phenomenological studies in two different groups: (1) hermeneutic (interpretive) and (2) empirical (experimental, descriptive, unbiased). Descriptive phenomenology “emphasizes the description of experiences by the researcher rather than interpreting lived experiences.” For this purpose, qualitative researchers define the phenomenon ‐with phenomenology,” and emphasize the descriptive type of phenomenology. In descriptive phenomenology studies, rather than the participants' experiences, the researcher defines these experiences . This phenomenological study was conducted as suggested by Miller (2003) , and the spiritual care experiences of the nurses working in the gynecology clinic were descriptively handled holistically by questioning the nurses' experiences. The Consolidated Criteria for Reporting Qualitative Research (COREQ) checklist guided the study. 2.2 Population and Sample The study population consisted of gynecology nurses working in government hospitals in Turkey. The study's sample size was determined by taking into account the qualitative research design, the selected sample's diversity, and the participants' statements. The inclusion criteria for the participants were voluntarily agreeing to participate in the study, working in the gynecology clinic for at least 1 year, and accepting the online interview. The sample was determined using the criterion sampling method, one of the purposeful sampling methods. The purposive sampling method used in qualitative research enables the selecting of people who are suitable for the study and who can best answer the research questions. In this method, people who meet the necessary criteria or have specific characteristics are preferred . Criterion sampling is based on examining the sample that meets the criteria determined by the researcher . Without calculating the sample size, interviews continued until data saturation was reached, and the sample group was finalized with 12 nurses in this process. 2.3 Data Collection The data collection process was conducted online between April 2 and May 20, 2024, by a researcher with a doctoral degree in obstetrics and gynecology nursing and the Registered Nurse/Assistant Professor (FB) title. The researcher has received training and courses on qualitative research methods and is experienced in qualitative approaches. The data consisted of a first part with seven questions to determine the socio‐demographic characteristics of the nurses and a second part with eight open‐ended questions about the spiritual care experiences of the nurses prepared in accordance with the scientific literature. Before collecting the data, a pilot test was conducted with two nurses, and these data were not included in the study. When the research reached 12 women nurses, the purpose of the study was achieved, and the data collection process was completed. The interviews were conducted online and recorded at the convenience and lonely of the nurses. Verbal consent was obtained from the participants at the beginning of the interviews. Each interview lasted approximately 30–35 min. On the day of the interviews, the audio recordings were analyzed and recorded by giving code numbers to the participants without their names and identity information. 2.4 Data Analysis After the data obtained from the interviews were transcribed by the researchers (FB, MŞ, HS) through the Transcriptor program ( https://app.transkriptor.com/dashboard ), each researcher checked the compatibility of the texts with the audio recordings, and missing or incorrect data were edited at this time. The data were analyzed using the descriptive analysis method in MAXQDA 22, a computer‐aided qualitative data analysis software. Descriptive analysis presents the obtained data to the reader in a format organized and interpreted according to the previously determined conceptual framework or themes. The data analysis methods of content analysis and descriptive analysis are the same. Data can be organized according to the themes revealed by the research questions or presented by considering the questions or dimensions used in the interview and observation processes. In descriptive analysis, direct quotes are used more to reflect the statements of the individuals interviewed or observed strikingly . The data uploaded to the program's system were read multiple times to holistically understand the participants' experiences, and the researchers independently coded them (FB, MŞ, HS). The differences between the researchers were discussed, and the arrangements continued until a consensus was reached on the themes. The coded data were categorized, and the first codes were formed and then turned into themes. After obtaining an expert's opinion with qualitative research experience in the nursing field, the themes were edited until a common opinion was reached. The results were tabulated and presented as themes, sub‐themes, and codes (Table ). In the finalized texts, participant numbers N1‐N12 were used instead of the real names of the nurses. 2.5 Validity and Reliability Validity and reliability in qualitative research are ensured through internal validity, external validity, internal reliability, and external reliability methods . For the internal validity of the research, expert review was utilized, the research was shared with an expert academician who had previously conducted qualitative research and was asked to evaluate the research as a whole in terms of its conceptual dimension, objectives, problem, method, data collection tool, data analysis, and reporting. The research was finalized by considering the expert review's suggestions. For the external validity of the research, the research process, what was done in this process, the research model, the study group, the data collection tool, the data collection process, and the analysis and interpretation of the data were given in detail. For the internal reliability of the study, direct quotations were made from the opinions of the nurses, coding was done separately by the researchers (FB, MŞ, HS), and the percentage of inter‐coder agreement was calculated. The researchers read the data transcripts separately, coded the participants' statements, and created a code list. The coding reliability was calculated using the inter‐coder similarity formula proposed by Miles and Huberman and determined to be 83% . For the external reliability of the study, descriptive data of the nurses were presented, how the interviews and data were recorded, how the analysis was performed, and how the results were combined and presented were explained in detail. 2.6 Ethical Considerations Ethics committee permission of Hakkari University Scientific Research and Publication Ethics Committee dated 18.03.2024 and numbered 2024/45 was obtained for the study. The nurses participating in the study were informed about the aims and design of the study, and their verbal consent was obtained. The informed consent of the participants was obtained online, and the study was conducted in accordance with the Declaration of Helsinki. Research Type, Location and Time This qualitative research was conducted between April 2 and May 20, 2024, with nurses with at least 1 year of clinical experience in gynecology actively working in hospitals in different provinces of Turkey. The study was carried out using the descriptive phenomenology type of phenomenological design, one of the qualitative research methods. Phenomenological design is a research method that enables people to express their understanding, feelings, perspectives, and perceptions about a particular phenomenon or concept. It describes how they experience this phenomenon . The primary purpose of phenomenology is to understand the essence of the object by reducing individual experiences about an event to a universal explanation . Creswell (2016) discusses phenomenological studies in two different groups: (1) hermeneutic (interpretive) and (2) empirical (experimental, descriptive, unbiased). Descriptive phenomenology “emphasizes the description of experiences by the researcher rather than interpreting lived experiences.” For this purpose, qualitative researchers define the phenomenon ‐with phenomenology,” and emphasize the descriptive type of phenomenology. In descriptive phenomenology studies, rather than the participants' experiences, the researcher defines these experiences . This phenomenological study was conducted as suggested by Miller (2003) , and the spiritual care experiences of the nurses working in the gynecology clinic were descriptively handled holistically by questioning the nurses' experiences. The Consolidated Criteria for Reporting Qualitative Research (COREQ) checklist guided the study. Population and Sample The study population consisted of gynecology nurses working in government hospitals in Turkey. The study's sample size was determined by taking into account the qualitative research design, the selected sample's diversity, and the participants' statements. The inclusion criteria for the participants were voluntarily agreeing to participate in the study, working in the gynecology clinic for at least 1 year, and accepting the online interview. The sample was determined using the criterion sampling method, one of the purposeful sampling methods. The purposive sampling method used in qualitative research enables the selecting of people who are suitable for the study and who can best answer the research questions. In this method, people who meet the necessary criteria or have specific characteristics are preferred . Criterion sampling is based on examining the sample that meets the criteria determined by the researcher . Without calculating the sample size, interviews continued until data saturation was reached, and the sample group was finalized with 12 nurses in this process. Data Collection The data collection process was conducted online between April 2 and May 20, 2024, by a researcher with a doctoral degree in obstetrics and gynecology nursing and the Registered Nurse/Assistant Professor (FB) title. The researcher has received training and courses on qualitative research methods and is experienced in qualitative approaches. The data consisted of a first part with seven questions to determine the socio‐demographic characteristics of the nurses and a second part with eight open‐ended questions about the spiritual care experiences of the nurses prepared in accordance with the scientific literature. Before collecting the data, a pilot test was conducted with two nurses, and these data were not included in the study. When the research reached 12 women nurses, the purpose of the study was achieved, and the data collection process was completed. The interviews were conducted online and recorded at the convenience and lonely of the nurses. Verbal consent was obtained from the participants at the beginning of the interviews. Each interview lasted approximately 30–35 min. On the day of the interviews, the audio recordings were analyzed and recorded by giving code numbers to the participants without their names and identity information. Data Analysis After the data obtained from the interviews were transcribed by the researchers (FB, MŞ, HS) through the Transcriptor program ( https://app.transkriptor.com/dashboard ), each researcher checked the compatibility of the texts with the audio recordings, and missing or incorrect data were edited at this time. The data were analyzed using the descriptive analysis method in MAXQDA 22, a computer‐aided qualitative data analysis software. Descriptive analysis presents the obtained data to the reader in a format organized and interpreted according to the previously determined conceptual framework or themes. The data analysis methods of content analysis and descriptive analysis are the same. Data can be organized according to the themes revealed by the research questions or presented by considering the questions or dimensions used in the interview and observation processes. In descriptive analysis, direct quotes are used more to reflect the statements of the individuals interviewed or observed strikingly . The data uploaded to the program's system were read multiple times to holistically understand the participants' experiences, and the researchers independently coded them (FB, MŞ, HS). The differences between the researchers were discussed, and the arrangements continued until a consensus was reached on the themes. The coded data were categorized, and the first codes were formed and then turned into themes. After obtaining an expert's opinion with qualitative research experience in the nursing field, the themes were edited until a common opinion was reached. The results were tabulated and presented as themes, sub‐themes, and codes (Table ). In the finalized texts, participant numbers N1‐N12 were used instead of the real names of the nurses. Validity and Reliability Validity and reliability in qualitative research are ensured through internal validity, external validity, internal reliability, and external reliability methods . For the internal validity of the research, expert review was utilized, the research was shared with an expert academician who had previously conducted qualitative research and was asked to evaluate the research as a whole in terms of its conceptual dimension, objectives, problem, method, data collection tool, data analysis, and reporting. The research was finalized by considering the expert review's suggestions. For the external validity of the research, the research process, what was done in this process, the research model, the study group, the data collection tool, the data collection process, and the analysis and interpretation of the data were given in detail. For the internal reliability of the study, direct quotations were made from the opinions of the nurses, coding was done separately by the researchers (FB, MŞ, HS), and the percentage of inter‐coder agreement was calculated. The researchers read the data transcripts separately, coded the participants' statements, and created a code list. The coding reliability was calculated using the inter‐coder similarity formula proposed by Miles and Huberman and determined to be 83% . For the external reliability of the study, descriptive data of the nurses were presented, how the interviews and data were recorded, how the analysis was performed, and how the results were combined and presented were explained in detail. Ethical Considerations Ethics committee permission of Hakkari University Scientific Research and Publication Ethics Committee dated 18.03.2024 and numbered 2024/45 was obtained for the study. The nurses participating in the study were informed about the aims and design of the study, and their verbal consent was obtained. The informed consent of the participants was obtained online, and the study was conducted in accordance with the Declaration of Helsinki. Results All of the nurses who participated in our study were female and aged 25 years and over. More than half of the nurses (58.3%) were married and had children. The majority of the nurses (83.3%) had Bachelor's degrees. The professional experience of the nurses ranged between 2 and 35 years. Moreover, the majority of the nurses (83.3%) evaluated their economic status as good, while 66.6% evaluated their religious beliefs as moderate (Table ). The spiritual care experiences of the nurses participating in the present study were determined to have five main themes (Table ). 3.1 Theme 1: The Concept of Spiritual Care Gynecology nurses expressed the theme of spiritual care concept in a single Subtheme as the meaning attributed to spiritual care. 3.1.1 Subtheme 1: The Meaning Attributed to Spiritual Care It was stated that nurses attributed very special and beautiful meanings to spiritual care and played an essential role in establishing more robust communication with the patient. It was noted that the meanings attributed have vital aspects such as psychological support, establishing strong communication, and reducing anxiety and fear. The statements of some of the participants are as follows: When I think of spiritual care, I think of the communication between me and the patient. I think of being able to trust each other. (N5) When I think of spiritual care, psychological support comes to mind. Since patients are in pain, it is part of our job to give them psychological support, to explain that this situation can happen, to try to reduce their anxiety and fear, and to comfort them by telling them that this situation is normal. This is what we do… (N8) 3.2 Theme 2: Spiritual Care Practices Nurses expressed the theme of spiritual care practices in three sub‐themes: psychological support, communication, and social support. It is observed that nurses mostly exhibit spiritual care practices with behaviors such as talking, listening, being friendly, and making them feel that they are with them to psychologically comfort them by trying to empathize with them patients. 3.2.1 Subtheme 1: Psychological Support Almost all of the nurses stated that they provided psychological support to the patients by ‘being with’ ( n = 10), using an ‘appropriate approach’ ( n = 7), and performing spiritual care practices in this way. The statement of one of the participants is as follows: In addition to treating patients, that is, medical support, we provide spiritual care by being with them. I don't know; we provide appropriate care to make them feel good and peaceful. The energy we get when we first meet them is also very important. I think these affect our spiritual care practices, and everything becomes easier. No matter how tired we are, we are with the patient more, and therefore, the time we allocate to the patient increases… (H3) 3.2.2 Subtheme 2: Communication Nurses stated that communication is a strong factor in spiritual care practices and that they communicate mostly through speaking, listening, giving information, and being friendly. The sample participant statement is quoted below as follows: First of all, while giving spiritual care, I see them as individuals and treat them as such. I don't know their ideas, where they lack, don't know, or are afraid; I don't know how much knowledge they have about the birth process. I start by measuring this first. I chat with them, talk to them, listen to them. Then I inform them. Thus, I help them to realize that birth is not something to be afraid of. Either way, after they have already taken care of the communication steps, they become brave enough to ask questions after a while… (H4) 3.2.3 Subtheme 3. Social Support Nurses stated that talking with family members is a practical spiritual practice in providing patients with social support. The statement of one of the participants is as follows: We provide spiritual care. My practices are more effective, especially in the postnatal period when I am postpartum. Very young pregnant women come here frequently. You know, women and children, and they cannot express themselves. I ensure that they receive family support and inform family members so they do not feel more alone. And that they may be at ease… (H1) 3.3 Theme 3: Factors Complicating Spiritual Care Nurses expressed the theme of complicating factors while providing spiritual care in three sub‐themes: factors related to the patient, environmental factors and factors related to the patient's relatives. 3.3.1 Subtheme 1: Patient‐Related Factors Nurses stated that the cultural differences of the patients, their inability to communicate, agitation, and the emergence of violence factor were the complicating factors related to the patients in spiritual care practices. The statements of some of the participants are as follows: Communication is sometimes tricky; there are complex patients. They push us very hard in every sense. We can also be subjected to insults. As if being subjected to verbal violence is not enough, it can also lead to physical violence. It can actually lead to physical violence. We have seen many pregnant women biting, kicking, shouting, and verbally abusing us on the delivery table. These are the problematic side of the job, but they are also negative…. (N2) Pregnant women are not all the same. They do not all come from the same culture. Turkey is, unfortunately, very undeveloped in this regard. We work devotedly, but sometimes it is not enough… (N4) It is difficult if the patient is not open to communication. I think it is difficult for them to express themselves. They cannot say some things open‐heartedly. Sometimes, we try hard, but it doesn't work. I think this is the most difficult part of the gynecology department. It is not something to be ashamed of. In fact, this is the most important thing for them, so why should they be ashamed? We understand and try to explain, but we definitely cannot communicate. (N7) 3.3.2 Subtheme 2: Environmental Factors Nurses stated that having problems with team members in spiritual care practices and not providing a comfort environment are environmental factors that make spiritual care difficult. Examples of nurse quotations related to this are as follows: There are times when I have a hard time on duty. There is no gynecologist or pediatrician outside working hours. Since many things are instantaneous, it can be a problem. It is a baby's life. It can take time for doctors to come from home to the hospital. A lot of time is lost. We get very stressed. During these periods, we cannot explain to some doctors the seriousness of the event, and we have problems… (N2) The physical conditions are really bad. The pregnant woman should be alone. There should not be another patient with her, and her comfort should be ensured. Physical conditions need to be improved. I think this is also a problem for state hospitals in Turkey. The physical conditions and circumstances are always inadequate, always partially met, etc… (N6) 3.3.3 Subtheme 3: Factors Related to the Patient's Relatives It was stated that the lack of knowledge of the patient's relatives, not respecting the privacy of the woman, and domestic problems were the complicating factors related to the patient's relatives experienced by the nurses in spiritual care practices. The statements of some of the participants are as follows: The patient comes with her mother‐in‐law. She looks her mother‐in‐law in the eye to do something. She even has her confirm what we say. Mothers‐in‐law have traditional knowledge, and we cannot tell them anything. In conclusion, we are training the mother in vain. I ask myself if this is what I worked for. We need to train the mothers‐in‐law first, but that is difficult. (N1) In‐laws are one of the factors that hinder us. If the patient comes with her mother‐in‐law, it is too hard. That process ruins us. It isn't nice if the mother‐in‐law interferes when we try to teach them. We don't want them to come. They interfere too much. Because the sense of privacy is also very important for us, our pregnant woman cannot feel comfortable with her mother‐in‐law or sister‐in‐law at that moment. (N3) 3.4 Theme 3: Factors Facilitating Spiritual Care Nurses expressed the theme of factors facilitating spiritual care in two sub‐themes: individual and occupational. 3.4.1 Subtheme 1: Individual Factors Nurses stated that empathy and sociocultural factors are individual factors that facilitate spiritual care. They said that approaching according to the sociocultural level of the patient and being in the female gender is a facilitating factor in empathizing. The statements of some of the participants are as follows: First of all, because we are women and because we are women, the essential thing in our profession is to empathize and put ourselves in their place. If I were like this, how would I want to be treated? This is spiritual care. Unfortunately, because we empathize, it is based on our conscience… (N10) It is crucial for me to stay calm. I need to remain calm and patiently treat patients according to their education level. Some patients understand, and patients do not. In this situation, selecting our words and expressions may be necessary. We need to be with them to understand their language, sometimes to integrate with their dialects, sometimes with the words they say. Because we are not all from the same part of the country… (N11) 3.4.2 Subtheme 2: Occupational Factors It was stated that reducing the pain, fear, or anxiety experienced by women during childbirth or gynecological practices by providing education and information can prevent the negative behaviors that nurses are exposed to by women in these clinics, facilitate spiritual care practices, and that effective communication and being friendly are essential factors. The statements of some of the participants are as follows: There are mums like that who attack here and there in labor; they want to throw themselves out. If we say whatever we want, it won't work. During the pandemic period, there were even people who spat in our faces. Amniotic fluid and saliva are coming. What have we seen? I'm sorry, but we work in filth. In fact, yes, we help in childbirth, but they need to be aware of this, so education is critical here. The education of the patients is also vital. We are university graduates, and everything is much easier when we are educated and provide training. Unfortunately, not every birth is the same, and not every pregnant woman's educational status is the same. Every birth is entirely different. (N1) …Spiritual care is a bit more humane. A humane approach is important for me because the other person is a human being. I care about them. I say hello or say how are you, and I approach them with a smiling face. (N12) 3.5 Theme 4: Emotional Effects of Spiritual Care Nurses expressed the theme of emotional effects of spiritual care in two sub‐themes: effects on nurses and effects on patients. 3.5.1 Subtheme 1: Effects on Nurses Nurses felt happy, competent, and peaceful and provided professional satisfaction with spiritual care practices. It is thought that spiritual care practices provide multidimensional positive emotions in nurses and that they perform their profession more fondly. The statements of some of the participants are as follows: We can say that I am conscientiously comfortable and peaceful. I feel satisfied. (N9) Motherhood always makes you feel inadequate no matter what you do. Therefore, I need to provide the best education and support to pregnant women as much as possible. So don't feel inadequate. I feel better, I feel more competent, I think I fulfill my profession adequately, I feel happier when I get positive patient feedback. (N2) 3.5.2 Subtheme 2: Effect on Patients Nurses stated that the spiritual care they applied had many positive emotional effects on patients, such as satisfaction, trust, relaxation, and motivation. It is seen in the quotations that these positive emotional effects on patients also positively affect nurses. The statements of some of the participants are as follows: Once we have established a sense of trust, our patients and mothers can come here and say anything. They can ask everything they want to ask and learn everything they want to learn. They relax and communicate. Thus, the satisfaction of pregnant women and mothers increases. (N1) Mothers and pregnant women are motivated. I feel so peaceful. I am happy. God willing, with their prayers, we will perform this profession for many more years. (N4) 3.6 Theme 5: Recommendations on the Best Way to Provide Spiritual Care Nurses expressed the theme of recommendations for providing spiritual care in the best way in three sub‐themes: recommendations for healthcare professionals, patients, and hospital administration. 3.6.1 Subtheme 1: Recommendations for Health Workers It was stated that training on spiritual care practices, smiling, and strong communication and counseling skills can be effective practices for healthcare professionals in providing spiritual care practices. The statements of some of the participants are as follows: I think we should be given updated training to have counseling skills. We should also be trained to improve ourselves. Communication and counseling skills are important. It should be applied to all staff. (N5) …I think the most important thing is to smile. If we want to support the other person, we need to smile. So, we need to walk around with our mouths in our ears. Smiling is really contagious. (N7) 3.6.2 Subtheme 2: Recommendations for Patients It was stated that educating patients and their relatives, informing them through pregnancy schools, and having sufficient social support for the woman in this process would significantly contribute to the best delivery of spiritual care practices. The statements of some of the participants are as follows: It should be ensured that women come and receive conscious education before the birth process. Apart from that, as I said, they can visit pregnancy schools and receive pregnancy training there… (N3) We can also interview their spouses or people in the family environment because they don't get support from their partners. They can be provided with social support. They hear a lot of negative things from their environment. Mothers or mothers‐in‐law criticize them about baby nutrition and physical conditions. These situations continue during the puerperium. Mothers‐in‐law and mothers can also be trained if the conditions are appropriate… (N2) 3.6.3 Subtheme 3: Recommendations for Hospital Administration Nurses stated that hospital management has essential duties in addition to nurses, patients, and their relatives to ensure that spiritual care is provided best. It was said that providing a suitable physical environment, increasing the number of experienced midwives/nurses, improving salaries, and creating single delivery rooms by considering privacy would increase the quality of spiritual care. The statements of some of the participants are as follows: …Our physical conditions should be good. Our physical conditions are not good. It is the problem with state hospitals. As such, everything is more restricted. We cannot ensure the privacy of the baby and the mother. This is a very comfortable thing at the same time. As I said, single delivery rooms should be available in all state hospitals. It needs to be fulfilled… (N12) The number of experienced nurses and midwives should be increased. Now, it's up to our administrators to fix it. (N4) Theme 1: The Concept of Spiritual Care Gynecology nurses expressed the theme of spiritual care concept in a single Subtheme as the meaning attributed to spiritual care. 3.1.1 Subtheme 1: The Meaning Attributed to Spiritual Care It was stated that nurses attributed very special and beautiful meanings to spiritual care and played an essential role in establishing more robust communication with the patient. It was noted that the meanings attributed have vital aspects such as psychological support, establishing strong communication, and reducing anxiety and fear. The statements of some of the participants are as follows: When I think of spiritual care, I think of the communication between me and the patient. I think of being able to trust each other. (N5) When I think of spiritual care, psychological support comes to mind. Since patients are in pain, it is part of our job to give them psychological support, to explain that this situation can happen, to try to reduce their anxiety and fear, and to comfort them by telling them that this situation is normal. This is what we do… (N8) Subtheme 1: The Meaning Attributed to Spiritual Care It was stated that nurses attributed very special and beautiful meanings to spiritual care and played an essential role in establishing more robust communication with the patient. It was noted that the meanings attributed have vital aspects such as psychological support, establishing strong communication, and reducing anxiety and fear. The statements of some of the participants are as follows: When I think of spiritual care, I think of the communication between me and the patient. I think of being able to trust each other. (N5) When I think of spiritual care, psychological support comes to mind. Since patients are in pain, it is part of our job to give them psychological support, to explain that this situation can happen, to try to reduce their anxiety and fear, and to comfort them by telling them that this situation is normal. This is what we do… (N8) Theme 2: Spiritual Care Practices Nurses expressed the theme of spiritual care practices in three sub‐themes: psychological support, communication, and social support. It is observed that nurses mostly exhibit spiritual care practices with behaviors such as talking, listening, being friendly, and making them feel that they are with them to psychologically comfort them by trying to empathize with them patients. 3.2.1 Subtheme 1: Psychological Support Almost all of the nurses stated that they provided psychological support to the patients by ‘being with’ ( n = 10), using an ‘appropriate approach’ ( n = 7), and performing spiritual care practices in this way. The statement of one of the participants is as follows: In addition to treating patients, that is, medical support, we provide spiritual care by being with them. I don't know; we provide appropriate care to make them feel good and peaceful. The energy we get when we first meet them is also very important. I think these affect our spiritual care practices, and everything becomes easier. No matter how tired we are, we are with the patient more, and therefore, the time we allocate to the patient increases… (H3) 3.2.2 Subtheme 2: Communication Nurses stated that communication is a strong factor in spiritual care practices and that they communicate mostly through speaking, listening, giving information, and being friendly. The sample participant statement is quoted below as follows: First of all, while giving spiritual care, I see them as individuals and treat them as such. I don't know their ideas, where they lack, don't know, or are afraid; I don't know how much knowledge they have about the birth process. I start by measuring this first. I chat with them, talk to them, listen to them. Then I inform them. Thus, I help them to realize that birth is not something to be afraid of. Either way, after they have already taken care of the communication steps, they become brave enough to ask questions after a while… (H4) 3.2.3 Subtheme 3. Social Support Nurses stated that talking with family members is a practical spiritual practice in providing patients with social support. The statement of one of the participants is as follows: We provide spiritual care. My practices are more effective, especially in the postnatal period when I am postpartum. Very young pregnant women come here frequently. You know, women and children, and they cannot express themselves. I ensure that they receive family support and inform family members so they do not feel more alone. And that they may be at ease… (H1) Subtheme 1: Psychological Support Almost all of the nurses stated that they provided psychological support to the patients by ‘being with’ ( n = 10), using an ‘appropriate approach’ ( n = 7), and performing spiritual care practices in this way. The statement of one of the participants is as follows: In addition to treating patients, that is, medical support, we provide spiritual care by being with them. I don't know; we provide appropriate care to make them feel good and peaceful. The energy we get when we first meet them is also very important. I think these affect our spiritual care practices, and everything becomes easier. No matter how tired we are, we are with the patient more, and therefore, the time we allocate to the patient increases… (H3) Subtheme 2: Communication Nurses stated that communication is a strong factor in spiritual care practices and that they communicate mostly through speaking, listening, giving information, and being friendly. The sample participant statement is quoted below as follows: First of all, while giving spiritual care, I see them as individuals and treat them as such. I don't know their ideas, where they lack, don't know, or are afraid; I don't know how much knowledge they have about the birth process. I start by measuring this first. I chat with them, talk to them, listen to them. Then I inform them. Thus, I help them to realize that birth is not something to be afraid of. Either way, after they have already taken care of the communication steps, they become brave enough to ask questions after a while… (H4) Subtheme 3. Social Support Nurses stated that talking with family members is a practical spiritual practice in providing patients with social support. The statement of one of the participants is as follows: We provide spiritual care. My practices are more effective, especially in the postnatal period when I am postpartum. Very young pregnant women come here frequently. You know, women and children, and they cannot express themselves. I ensure that they receive family support and inform family members so they do not feel more alone. And that they may be at ease… (H1) Theme 3: Factors Complicating Spiritual Care Nurses expressed the theme of complicating factors while providing spiritual care in three sub‐themes: factors related to the patient, environmental factors and factors related to the patient's relatives. 3.3.1 Subtheme 1: Patient‐Related Factors Nurses stated that the cultural differences of the patients, their inability to communicate, agitation, and the emergence of violence factor were the complicating factors related to the patients in spiritual care practices. The statements of some of the participants are as follows: Communication is sometimes tricky; there are complex patients. They push us very hard in every sense. We can also be subjected to insults. As if being subjected to verbal violence is not enough, it can also lead to physical violence. It can actually lead to physical violence. We have seen many pregnant women biting, kicking, shouting, and verbally abusing us on the delivery table. These are the problematic side of the job, but they are also negative…. (N2) Pregnant women are not all the same. They do not all come from the same culture. Turkey is, unfortunately, very undeveloped in this regard. We work devotedly, but sometimes it is not enough… (N4) It is difficult if the patient is not open to communication. I think it is difficult for them to express themselves. They cannot say some things open‐heartedly. Sometimes, we try hard, but it doesn't work. I think this is the most difficult part of the gynecology department. It is not something to be ashamed of. In fact, this is the most important thing for them, so why should they be ashamed? We understand and try to explain, but we definitely cannot communicate. (N7) 3.3.2 Subtheme 2: Environmental Factors Nurses stated that having problems with team members in spiritual care practices and not providing a comfort environment are environmental factors that make spiritual care difficult. Examples of nurse quotations related to this are as follows: There are times when I have a hard time on duty. There is no gynecologist or pediatrician outside working hours. Since many things are instantaneous, it can be a problem. It is a baby's life. It can take time for doctors to come from home to the hospital. A lot of time is lost. We get very stressed. During these periods, we cannot explain to some doctors the seriousness of the event, and we have problems… (N2) The physical conditions are really bad. The pregnant woman should be alone. There should not be another patient with her, and her comfort should be ensured. Physical conditions need to be improved. I think this is also a problem for state hospitals in Turkey. The physical conditions and circumstances are always inadequate, always partially met, etc… (N6) 3.3.3 Subtheme 3: Factors Related to the Patient's Relatives It was stated that the lack of knowledge of the patient's relatives, not respecting the privacy of the woman, and domestic problems were the complicating factors related to the patient's relatives experienced by the nurses in spiritual care practices. The statements of some of the participants are as follows: The patient comes with her mother‐in‐law. She looks her mother‐in‐law in the eye to do something. She even has her confirm what we say. Mothers‐in‐law have traditional knowledge, and we cannot tell them anything. In conclusion, we are training the mother in vain. I ask myself if this is what I worked for. We need to train the mothers‐in‐law first, but that is difficult. (N1) In‐laws are one of the factors that hinder us. If the patient comes with her mother‐in‐law, it is too hard. That process ruins us. It isn't nice if the mother‐in‐law interferes when we try to teach them. We don't want them to come. They interfere too much. Because the sense of privacy is also very important for us, our pregnant woman cannot feel comfortable with her mother‐in‐law or sister‐in‐law at that moment. (N3) Subtheme 1: Patient‐Related Factors Nurses stated that the cultural differences of the patients, their inability to communicate, agitation, and the emergence of violence factor were the complicating factors related to the patients in spiritual care practices. The statements of some of the participants are as follows: Communication is sometimes tricky; there are complex patients. They push us very hard in every sense. We can also be subjected to insults. As if being subjected to verbal violence is not enough, it can also lead to physical violence. It can actually lead to physical violence. We have seen many pregnant women biting, kicking, shouting, and verbally abusing us on the delivery table. These are the problematic side of the job, but they are also negative…. (N2) Pregnant women are not all the same. They do not all come from the same culture. Turkey is, unfortunately, very undeveloped in this regard. We work devotedly, but sometimes it is not enough… (N4) It is difficult if the patient is not open to communication. I think it is difficult for them to express themselves. They cannot say some things open‐heartedly. Sometimes, we try hard, but it doesn't work. I think this is the most difficult part of the gynecology department. It is not something to be ashamed of. In fact, this is the most important thing for them, so why should they be ashamed? We understand and try to explain, but we definitely cannot communicate. (N7) Subtheme 2: Environmental Factors Nurses stated that having problems with team members in spiritual care practices and not providing a comfort environment are environmental factors that make spiritual care difficult. Examples of nurse quotations related to this are as follows: There are times when I have a hard time on duty. There is no gynecologist or pediatrician outside working hours. Since many things are instantaneous, it can be a problem. It is a baby's life. It can take time for doctors to come from home to the hospital. A lot of time is lost. We get very stressed. During these periods, we cannot explain to some doctors the seriousness of the event, and we have problems… (N2) The physical conditions are really bad. The pregnant woman should be alone. There should not be another patient with her, and her comfort should be ensured. Physical conditions need to be improved. I think this is also a problem for state hospitals in Turkey. The physical conditions and circumstances are always inadequate, always partially met, etc… (N6) Subtheme 3: Factors Related to the Patient's Relatives It was stated that the lack of knowledge of the patient's relatives, not respecting the privacy of the woman, and domestic problems were the complicating factors related to the patient's relatives experienced by the nurses in spiritual care practices. The statements of some of the participants are as follows: The patient comes with her mother‐in‐law. She looks her mother‐in‐law in the eye to do something. She even has her confirm what we say. Mothers‐in‐law have traditional knowledge, and we cannot tell them anything. In conclusion, we are training the mother in vain. I ask myself if this is what I worked for. We need to train the mothers‐in‐law first, but that is difficult. (N1) In‐laws are one of the factors that hinder us. If the patient comes with her mother‐in‐law, it is too hard. That process ruins us. It isn't nice if the mother‐in‐law interferes when we try to teach them. We don't want them to come. They interfere too much. Because the sense of privacy is also very important for us, our pregnant woman cannot feel comfortable with her mother‐in‐law or sister‐in‐law at that moment. (N3) Theme 3: Factors Facilitating Spiritual Care Nurses expressed the theme of factors facilitating spiritual care in two sub‐themes: individual and occupational. 3.4.1 Subtheme 1: Individual Factors Nurses stated that empathy and sociocultural factors are individual factors that facilitate spiritual care. They said that approaching according to the sociocultural level of the patient and being in the female gender is a facilitating factor in empathizing. The statements of some of the participants are as follows: First of all, because we are women and because we are women, the essential thing in our profession is to empathize and put ourselves in their place. If I were like this, how would I want to be treated? This is spiritual care. Unfortunately, because we empathize, it is based on our conscience… (N10) It is crucial for me to stay calm. I need to remain calm and patiently treat patients according to their education level. Some patients understand, and patients do not. In this situation, selecting our words and expressions may be necessary. We need to be with them to understand their language, sometimes to integrate with their dialects, sometimes with the words they say. Because we are not all from the same part of the country… (N11) 3.4.2 Subtheme 2: Occupational Factors It was stated that reducing the pain, fear, or anxiety experienced by women during childbirth or gynecological practices by providing education and information can prevent the negative behaviors that nurses are exposed to by women in these clinics, facilitate spiritual care practices, and that effective communication and being friendly are essential factors. The statements of some of the participants are as follows: There are mums like that who attack here and there in labor; they want to throw themselves out. If we say whatever we want, it won't work. During the pandemic period, there were even people who spat in our faces. Amniotic fluid and saliva are coming. What have we seen? I'm sorry, but we work in filth. In fact, yes, we help in childbirth, but they need to be aware of this, so education is critical here. The education of the patients is also vital. We are university graduates, and everything is much easier when we are educated and provide training. Unfortunately, not every birth is the same, and not every pregnant woman's educational status is the same. Every birth is entirely different. (N1) …Spiritual care is a bit more humane. A humane approach is important for me because the other person is a human being. I care about them. I say hello or say how are you, and I approach them with a smiling face. (N12) Subtheme 1: Individual Factors Nurses stated that empathy and sociocultural factors are individual factors that facilitate spiritual care. They said that approaching according to the sociocultural level of the patient and being in the female gender is a facilitating factor in empathizing. The statements of some of the participants are as follows: First of all, because we are women and because we are women, the essential thing in our profession is to empathize and put ourselves in their place. If I were like this, how would I want to be treated? This is spiritual care. Unfortunately, because we empathize, it is based on our conscience… (N10) It is crucial for me to stay calm. I need to remain calm and patiently treat patients according to their education level. Some patients understand, and patients do not. In this situation, selecting our words and expressions may be necessary. We need to be with them to understand their language, sometimes to integrate with their dialects, sometimes with the words they say. Because we are not all from the same part of the country… (N11) Subtheme 2: Occupational Factors It was stated that reducing the pain, fear, or anxiety experienced by women during childbirth or gynecological practices by providing education and information can prevent the negative behaviors that nurses are exposed to by women in these clinics, facilitate spiritual care practices, and that effective communication and being friendly are essential factors. The statements of some of the participants are as follows: There are mums like that who attack here and there in labor; they want to throw themselves out. If we say whatever we want, it won't work. During the pandemic period, there were even people who spat in our faces. Amniotic fluid and saliva are coming. What have we seen? I'm sorry, but we work in filth. In fact, yes, we help in childbirth, but they need to be aware of this, so education is critical here. The education of the patients is also vital. We are university graduates, and everything is much easier when we are educated and provide training. Unfortunately, not every birth is the same, and not every pregnant woman's educational status is the same. Every birth is entirely different. (N1) …Spiritual care is a bit more humane. A humane approach is important for me because the other person is a human being. I care about them. I say hello or say how are you, and I approach them with a smiling face. (N12) Theme 4: Emotional Effects of Spiritual Care Nurses expressed the theme of emotional effects of spiritual care in two sub‐themes: effects on nurses and effects on patients. 3.5.1 Subtheme 1: Effects on Nurses Nurses felt happy, competent, and peaceful and provided professional satisfaction with spiritual care practices. It is thought that spiritual care practices provide multidimensional positive emotions in nurses and that they perform their profession more fondly. The statements of some of the participants are as follows: We can say that I am conscientiously comfortable and peaceful. I feel satisfied. (N9) Motherhood always makes you feel inadequate no matter what you do. Therefore, I need to provide the best education and support to pregnant women as much as possible. So don't feel inadequate. I feel better, I feel more competent, I think I fulfill my profession adequately, I feel happier when I get positive patient feedback. (N2) 3.5.2 Subtheme 2: Effect on Patients Nurses stated that the spiritual care they applied had many positive emotional effects on patients, such as satisfaction, trust, relaxation, and motivation. It is seen in the quotations that these positive emotional effects on patients also positively affect nurses. The statements of some of the participants are as follows: Once we have established a sense of trust, our patients and mothers can come here and say anything. They can ask everything they want to ask and learn everything they want to learn. They relax and communicate. Thus, the satisfaction of pregnant women and mothers increases. (N1) Mothers and pregnant women are motivated. I feel so peaceful. I am happy. God willing, with their prayers, we will perform this profession for many more years. (N4) Subtheme 1: Effects on Nurses Nurses felt happy, competent, and peaceful and provided professional satisfaction with spiritual care practices. It is thought that spiritual care practices provide multidimensional positive emotions in nurses and that they perform their profession more fondly. The statements of some of the participants are as follows: We can say that I am conscientiously comfortable and peaceful. I feel satisfied. (N9) Motherhood always makes you feel inadequate no matter what you do. Therefore, I need to provide the best education and support to pregnant women as much as possible. So don't feel inadequate. I feel better, I feel more competent, I think I fulfill my profession adequately, I feel happier when I get positive patient feedback. (N2) Subtheme 2: Effect on Patients Nurses stated that the spiritual care they applied had many positive emotional effects on patients, such as satisfaction, trust, relaxation, and motivation. It is seen in the quotations that these positive emotional effects on patients also positively affect nurses. The statements of some of the participants are as follows: Once we have established a sense of trust, our patients and mothers can come here and say anything. They can ask everything they want to ask and learn everything they want to learn. They relax and communicate. Thus, the satisfaction of pregnant women and mothers increases. (N1) Mothers and pregnant women are motivated. I feel so peaceful. I am happy. God willing, with their prayers, we will perform this profession for many more years. (N4) Theme 5: Recommendations on the Best Way to Provide Spiritual Care Nurses expressed the theme of recommendations for providing spiritual care in the best way in three sub‐themes: recommendations for healthcare professionals, patients, and hospital administration. 3.6.1 Subtheme 1: Recommendations for Health Workers It was stated that training on spiritual care practices, smiling, and strong communication and counseling skills can be effective practices for healthcare professionals in providing spiritual care practices. The statements of some of the participants are as follows: I think we should be given updated training to have counseling skills. We should also be trained to improve ourselves. Communication and counseling skills are important. It should be applied to all staff. (N5) …I think the most important thing is to smile. If we want to support the other person, we need to smile. So, we need to walk around with our mouths in our ears. Smiling is really contagious. (N7) 3.6.2 Subtheme 2: Recommendations for Patients It was stated that educating patients and their relatives, informing them through pregnancy schools, and having sufficient social support for the woman in this process would significantly contribute to the best delivery of spiritual care practices. The statements of some of the participants are as follows: It should be ensured that women come and receive conscious education before the birth process. Apart from that, as I said, they can visit pregnancy schools and receive pregnancy training there… (N3) We can also interview their spouses or people in the family environment because they don't get support from their partners. They can be provided with social support. They hear a lot of negative things from their environment. Mothers or mothers‐in‐law criticize them about baby nutrition and physical conditions. These situations continue during the puerperium. Mothers‐in‐law and mothers can also be trained if the conditions are appropriate… (N2) 3.6.3 Subtheme 3: Recommendations for Hospital Administration Nurses stated that hospital management has essential duties in addition to nurses, patients, and their relatives to ensure that spiritual care is provided best. It was said that providing a suitable physical environment, increasing the number of experienced midwives/nurses, improving salaries, and creating single delivery rooms by considering privacy would increase the quality of spiritual care. The statements of some of the participants are as follows: …Our physical conditions should be good. Our physical conditions are not good. It is the problem with state hospitals. As such, everything is more restricted. We cannot ensure the privacy of the baby and the mother. This is a very comfortable thing at the same time. As I said, single delivery rooms should be available in all state hospitals. It needs to be fulfilled… (N12) The number of experienced nurses and midwives should be increased. Now, it's up to our administrators to fix it. (N4) Subtheme 1: Recommendations for Health Workers It was stated that training on spiritual care practices, smiling, and strong communication and counseling skills can be effective practices for healthcare professionals in providing spiritual care practices. The statements of some of the participants are as follows: I think we should be given updated training to have counseling skills. We should also be trained to improve ourselves. Communication and counseling skills are important. It should be applied to all staff. (N5) …I think the most important thing is to smile. If we want to support the other person, we need to smile. So, we need to walk around with our mouths in our ears. Smiling is really contagious. (N7) Subtheme 2: Recommendations for Patients It was stated that educating patients and their relatives, informing them through pregnancy schools, and having sufficient social support for the woman in this process would significantly contribute to the best delivery of spiritual care practices. The statements of some of the participants are as follows: It should be ensured that women come and receive conscious education before the birth process. Apart from that, as I said, they can visit pregnancy schools and receive pregnancy training there… (N3) We can also interview their spouses or people in the family environment because they don't get support from their partners. They can be provided with social support. They hear a lot of negative things from their environment. Mothers or mothers‐in‐law criticize them about baby nutrition and physical conditions. These situations continue during the puerperium. Mothers‐in‐law and mothers can also be trained if the conditions are appropriate… (N2) Subtheme 3: Recommendations for Hospital Administration Nurses stated that hospital management has essential duties in addition to nurses, patients, and their relatives to ensure that spiritual care is provided best. It was said that providing a suitable physical environment, increasing the number of experienced midwives/nurses, improving salaries, and creating single delivery rooms by considering privacy would increase the quality of spiritual care. The statements of some of the participants are as follows: …Our physical conditions should be good. Our physical conditions are not good. It is the problem with state hospitals. As such, everything is more restricted. We cannot ensure the privacy of the baby and the mother. This is a very comfortable thing at the same time. As I said, single delivery rooms should be available in all state hospitals. It needs to be fulfilled… (N12) The number of experienced nurses and midwives should be increased. Now, it's up to our administrators to fix it. (N4) Discussion This study examined the spiritual care experiences of nurses working in a gynecology clinic. As a result of the analysis of the interviews, five main themes were identified: “the concept of spiritual care”, “factors complicating spiritual care “, “factors facilitating spiritual care”, “emotional effects of spiritual care”, and “recommendations on the best way to provide spiritual care”. The only subtheme of the main theme of the concept of spiritual care was determined as “the meaning attributed to spiritual care”. Participants attributed many positive meanings (such as psychological support, communication, reducing fear and anxiety, and compassion) to spiritual care. In a mixed‐method study conducted with 282 nurses in South Korea, researchers expressed the meaning nurses attributed to spiritual care as four themes (helping to prepare for an honorable death with religious support, providing comfort and empathy, supporting spiritual satisfaction by finding meaning, and providing comprehensive care to the patient and family) . In a quasi‐experimental study conducted with nurses, Riahi et al. (2018) stated that improving spiritual care provided by nurses can result in various outcomes such as increased patient satisfaction, decreased symptoms of anxiety and depression during hospitalization, shorter hospital stays, and overall improved quality of life . As a result of studies conducted in societies and cultures with different religious beliefs, it has been concluded that nurses attribute positive meanings to spiritual care, and it benefits patients and their relatives. The sub‐themes of the main theme of factors that make spiritual care difficult were determined as “patient‐related”, “environmental” and “patient relatives‐related” factors. It was determined that nurses had difficulty in communicating with patients, and factors such as cultural differences, lack of knowledge, and privacy made spiritual care difficult. Similar to the findings of previous studies, it has been determined that the most frequently mentioned barriers to providing spiritual care were lack of time/intense work tempo, insufficient knowledge and training on spiritual care, inability to communicate, diversity of spiritual needs of patients, and lack of privacy, respectively . It is thought that the excessive workload of nurses is an essential obstacle in providing spiritual care to patients, and nurses are inadequate in meeting the spiritual needs of patients because they cannot access sufficient information and training on this subject. Patient and nurse satisfaction can be increased by reducing the workload of nurses and organizing spiritual care training programs for health personnel. The absence of spiritual care units in our country and inadequate training are also essential factors. Establishing comprehensive spiritual care units in hospitals and integrating these units into other services over time by increasing the quality and number of training provided will positively contribute to both nurses and patients. The sub‐themes of the main theme of factors facilitating spiritual care were determined as “individual” and “professional” factors. Factors such as empathy, training, effective communication, and a friendly approach were determined to facilitate spiritual care. In a cross‐sectional study conducted by Han et al. (2023) with nurses, it was determined that there was a significant relationship between nurses' personality characteristics and spiritual care behaviors . It was reported that nurses' spiritual care training, long‐term working experience, and high level of education were effective factors in spiritual care competence and experience in providing spiritual care It is estimated that education is an important factor in spiritual care, positively affects the personality traits of nurses, and contributes positively to patient communication and satisfaction. The sub‐themes of the main theme of emotional effects of spiritual care were determined as the effects on “nurses” and “patients”. It was determined that spiritual care contributed to professional satisfaction, qualification, happiness, and peacefulness in nurses; and satisfaction, trust, relaxation, and motivation in patients. In a study conducted by Güner and Akyüz (2023), it was stated that there was a significant positive relationship between nurses' spiritual levels and quality of care behavior and life satisfaction, and between spiritual care behaviors and life satisfaction . There was a lack of studies on the emotional effects of spiritual care in different cultures and countries with religious beliefs. However, our findings are similar to the results of the study conducted in Turkey, a Muslim country, and it was concluded that spiritual care provides intrinsic motivation in both nurses and patients. The sub‐themes of the main theme of suggestions for the best provision of spiritual care were determined as suggestions for “healthcare workers”, “patients” and “hospital administration”. The recommendations were determined as the healthcare professionals should be educated and friendly, the educational and social support of the patients should be increased, and the administration should increase the number of nurses, improve salaries, organize maternity rooms, and improve physical conditions. The results of previous studies are similar to our findings . However, it is noteworthy that no previous study has been conducted directly with gynecology nurses. Although the workload of nurses is high, the low number of nurses and the lack of a satisfactory salary are factors that reduce satisfaction for both patients and nurses and reduce the quality of care. Conclusion and Suggestions In the present study, the spiritual care experiences of nurses working in the gynecology clinic were expressed as the meaning they attributed to spiritual care, facilitating and complicating factors, emotional effects, and suggestions. It was observed that nurses attributed positive meanings to spiritual care. However, they had high workloads, a lack of training and knowledge about spiritual care, problems with both patient and administration, and could not provide spiritual care completely. It is noteworthy that gynecology nurses, who play a key role for both mother and newborn, have an important gap in meeting patients' spiritual needs in nursing practices despite the interest in spiritual care in the clinic. This finding may guide nursing clinicians, educators, and policymakers to more effectively approach spiritual care as a useful component of holistic care. It is recommended to integrate spiritual content into educational programs and reducing the workload of nurses to enable more effective clinical delivery. Furthermore, it is thought that it would be useful to apply different cultural assessments to obtain more benefits from spiritual care practices. 5.1 Limitations The data of this study are limited to the opinions of nurses working in gynecology clinics of state hospitals, and the results cannot be generalized. Future research should broaden the scope to understand different hospital staff better. 5.2 Implications to Practice The study's findings provide a broad perspective on the problems experienced in gynecology clinics. By revealing the experiences of nurses who provide care and treatment to patients, it may be beneficial in eliminating problems, errors and difficulties in the knowledge, attitudes and skills of nurses working as health professionals regarding the care and treatment practices of individuals. Thus, it is thought to effectively protect the physical and mental health of nurses working professionally. Limitations The data of this study are limited to the opinions of nurses working in gynecology clinics of state hospitals, and the results cannot be generalized. Future research should broaden the scope to understand different hospital staff better. Implications to Practice The study's findings provide a broad perspective on the problems experienced in gynecology clinics. By revealing the experiences of nurses who provide care and treatment to patients, it may be beneficial in eliminating problems, errors and difficulties in the knowledge, attitudes and skills of nurses working as health professionals regarding the care and treatment practices of individuals. Thus, it is thought to effectively protect the physical and mental health of nurses working professionally. Study design: Fatma Başaran, Merve Şahin and Hava Salik. Data collection: Fatma Başaran. Data analysis: Fatma Başaran, Merve Şahin and Hava Salık. Study supervision: Fatma Başaran. Manuscript writing: Fatma Başaran, Merve Şahin and Hava Salık. Critical revisions for important intellectual content: Fatma Başaran, Merve Şahin and Hava Salık. The authors declare no conflicts of interest. |
Mental Practice to Maintain Procedural Competency of Faculty with Decreased Opportunities | b873df89-8fbc-467a-8ebf-1247bb4e6d62 | 11734678 | Pediatrics[mh] | To address our research question, we used a qualitative methodology using semistructured interviews as our data source. A qualitative approach was believed to be the most appropriate method to address the aims of the study. Individual in-depth interviews were assessed to afford the best context for eliciting faculty perceptions of the mental practices used to maintain competency. The study took place in 2023 in a tertiary PCCM department. The institution’s Research Ethics Board approved the conduct of this study. Informed written consent was obtained from each participant. All faculty in PCCM for whom maintenance of competence in procedural skills is a requirement were invited to participate in this study. A semistructured interview lasting approximately 30 minutes was conducted by both authors, both with experience in interviewing and qualitative methodology. The interview explored different aspects of MP as a strategy for maintenance of skill competence ( see Appendix E1 in the data supplement). Although the interview protocol served as an initial guide, it was modified after data collection and analysis to explore new themes that were identified through constant comparative analysis. Audio recordings of the interviews were transcribed and deidentified. The data from interview transcripts were analyzed and coded inductively as well as deductively using Guillot and Collet’s MIIM, exploring the distinct outcomes of performance, motivation, and problem solving . The completeness of the definition and its focus on outcomes made it the preferred lens for analyzing the data to answer our question. Analysis and coding themes were performed by a faculty member and a trainee (T.K.K.W. and B.M.). Data were coded manually using an interpretive approach to thematic analysis, establishing patterns and relations among themes. Our aim of exploring MP though the MIIM influenced our decision to use thematic analysis. Differences were resolved through discussion. Both authors actively identified and openly discussed personal experiences and perspectives of MP to keep their data interpretations in check. An approximation of 20 participants to be interviewed was planned on the basis of our study aims, our sample specificity, theoretical perspectives on MP, and the potential richness of data of the interviews (predicated on our interviewers’ experience and participants’ expressivity). The adequacy of the final sample size was constantly evaluated during data collection and analysis. We achieved data and thematic saturation as the data collection and analysis occurred concurrently. We concluded that we had achieved data saturation during our data collection as the new interviews repeated what was previously expressed. We also achieved thematic saturation during our data analysis, as there was no emergence of new codes or themes. Member checking was performed by presenting the results to participants, who confirmed that the themes resonated with them. To ensure that our analysis was rich, robust, and comprehensive and to enhance its quality and credibility, we used different types of triangulations: 1 ) review by inquiry participants (presenting the data to participants) or member check-in and 2 ) researcher or investigator (having a diverse group of researchers analyze the data: trainees and faculty members). We obtained consent from 18 faculty members in PCCM; however, after 13 interviews, we reached thematic saturation and decided not to conduct the remaining 5 interviews. The interviewed faculty members included six men and seven women. Their years in practice ranged from 2 to 30, with some faculty having practiced for at least 15 years. They all practice in a large training program that enrolls 25–30 PCCM subspecialty trainees per year. The themes presented here emerged from using Guillot and Collet’s MIIM framework to guide our data analysis. We found that all faculty used MP to rehearse strategies to anticipate and troubleshoot problems that might arise during procedures. These problems and strategies were accumulated through their own practice or in discussions with others and frequently rehearsed before procedures. These mental images of what might happen and potential alternatives were believed to help create various plans that facilitated making immediate and best decisions during the performance event. Would I rehearse the steps before an individual procedure? Probably not. If I was thinking about the procedure, I would certainly go through a range of possibilities, I have a mental repertoire. “If this doesn’t work out, what would I do?” (F3) Is it like Michael Jordan standing, imagining how you’re going to perform and what it’s going to look like and then you deliver that type of performance? I don’t think I do it quite like that. I try to think what is going to be slightly tricky and what are my options? (F6) We found that fewer faculty used MP for task purposes, to rehearse the steps of a procedure, to remember its execution, or to maintain their technical aspects of the performance. When this was done, the form of imagery used was visualization or a cognitive rehearsal of the steps. I’ve never been a visual thinker, to be honest, if that’s what you were asking me about....I may deliberately think through the steps rather than how I move my hand. (F11) I do visualise prior to starting a procedure, the steps involved in a procedure. Particularly if it’s something that I haven’t done so...and especially if it’s something which requires a number of different pieces of equipment. Just kind of going through it sequentially in your mind, and then you kind of check off in your mind. (F5) The use of MP for rehearsing strategies or task execution consequently increased self-confidence and reduced anxiety in the new landscape of fewer direct opportunities that came with a supervisory role. Because I think that the secret of an experienced intensivist is having a plan B, C and D and that’s what gets you out of trouble. It’s not knowing that plan A always looks like plan A....It is about what happens if this happens. (F6) Because if you haven’t done it in a while, it’s important to make sure you remember the steps all the way to completion. (F2) Additional results are presented in , , and Appendix E2. Faculty in academic centers need to maintain their procedural expertise while facing decreased clinical opportunities and direct involvement. In this shifting procedural landscape, new strategies are needed, and we investigated MP as a potential solution. We found that faculty used MP the most as a cognitive tool for preparation and problem solving. As a result, MP also helped in anxiety reduction and self-confidence. Faculty used MP less often to improve motor performance. The potential mechanisms of how MP affects performance have been widely investigated. The ability of MP to help the formation and consolidation of mental representations of motor tasks has been linked to improved performance . Skillful coordination during a performance occurs when appropriate mental representations of the motor task and action goals are constructed. Faculty, as experienced practitioners, have already consolidated mental representations of the tasks and might not need MP for this purpose. In addition, faculty are exposed to mental task representation during the supervision of trainees performing the procedure. All faculty used MP for procedural preparation and problem solving. This entailed reviewing contextual factors such as team composition, patient factors, and potential complications. Faculty visualized the images of potential scenarios and formulated prospective solutions. These findings are in line with the results of experiments done with elite athletes and musicians . As with any concept, many definitions exist, and in this study, we applied the most inclusive one. This study has helped map out the use of MP by experienced physicians. The current landscape in PCCM is that of fewer overall procedures’ being performed by intensivists. With this decreased availability, there is a need to maximize each opportunity. MP provides a technique to help improve performance and increase skill proficiency. Another benefit is the low resource commitment required to perform MP compared with other educational options such as simulation . The lack of need for additional ancillary equipment further promotes its flexibility. MP can be used at the patient’s bedside or in the classroom. In addition, these processes can become more explicit and shared with more junior learners, who could benefit from being shown their usefulness. The findings from this study at a pediatric tertiary-level center show that MP is used by staff pediatric intensivists in performing and maintaining procedural competency. These may aid the maintenance of procedural competency by improving motor performance of skills, improving the emotional state of readiness, and creating a platform with which to troubleshoot potential complications. The low-resource nature of MP could make it a useful adjunct when considering how to maintain procedural competency. 10.34197/ats-scholar.2023-0150IN Data Supplement |
Current status of precision oncology in adult glioblastoma | c133cd75-9953-4f2a-833a-58d9ef7c2323 | 11619805 | Internal Medicine[mh] | Introduction In recent years, there has been a rapidly expanding amount of information on the molecular vulnerabilities of cancer cells, informing the development and application of targeted drugs. Actionable targets may comprise the genomic, genetic, epigenetic, transcriptional, proteomic, and metabolomic properties of tumor cells and may vary from one individual tumor to the next. Early studies of targeted treatment determined target engagement and therapeutic efficacy in post hoc subgroup analyses of larger non‐selective and target‐agnostic trials. Precision oncology goes in the opposite way, analyzing an individual tumor to inform treatment tailored to its biological characteristics . Target detection is a prerequisite, but the decision to apply a drug to an individual patient also requires antitumor efficacy based on – at minimum – biological plausibility, key insights from preclinical research, or clinical data from other tumor types or optimally the same tumor entity harboring the targeted vulnerability. Several classifications formalize the grading of clinical significance and actionability of molecular targets . Recent clinical trials, including molecularly targeted agents and immunotherapies, have achieved unprecedented survival in solid cancers enriched with specific molecular alterations. The cancer types that have benefited the most are molecularly defined subgroups of melanoma, non‐small cell lung cancer (NSCLC), and breast cancer. In melanoma, high response rates (> 60%) and prolonged progression‐free survival (PFS) and overall survival (OS) were achieved . In NSCLC, targeted treatment based on defined molecular alterations became the standard of care for first‐line therapy . In breast cancer, targeted treatment is also well established, and several compounds are available and sequentially applied . Targeted treatment is increasingly established for other cancer entities . Of note, genetic tumor cell vulnerabilities are not restricted to a single cancer entity but may occur in many different entities, albeit at a low percentage. This opens the door for the application of drugs registered for one entity to be used for tumors from other cancer entities that harbor the actionable alteration, e.g. in NSCLC and gliomas . Accordingly, entity‐agnostic clinical trials and drug registrations focusing on the presence of a particular alteration have been initiated . However, cross‐entity comparisons illustrate that the response to a certain compound may vary significantly, owing to the differential activation of concurrent pathways, the occurrence of different mutations, and the differing tissue permeability for the respective drug . This review explores the current status and perspective of precision oncology in isocitrate dehydrogenase (IDH) wildtype glioblastoma (GBM), the most common and aggressive malignant CNS tumor in adults. GBM therapy: current status and potential therapeutic targets 2.1 Current standard of care in the newly diagnosed and progressive setting Maximum‐safe tumor resection , radiotherapy of the tumor region , and concomitant and adjuvant temozolomide (TMZ) , an alkylating agent with high CNS penetrance and bioavailability, represent the current standard of care for first‐line GBM treatment . Tumor treating fields applying alternating electrical fields through the scalp are an additional treatment option . Furthermore, nitrosoureas such as lomustine (CCNU), also an alkylating agent with similarly favorable CNS penetrance, are available for first‐ or further‐line treatment . In some countries, the anti‐vascular endothelial growth factor (VEGF) A antibody bevacizumab is used as a non‐selective second‐line therapy, although it prolongs only PFS but not OS . The portfolio of further line therapies is enriched only by re‐resection , re‐irradiation , and possibly the multi‐kinase inhibitor regorafenib (see Section ), while the aggressive course of disease limits the time for further treatment lines. Thus, GBM patients need new and effective treatment options that should ideally be guided by predictive molecular characteristics to ensure that medications with a high likelihood of effectiveness are applied. 2.2 Molecular predictors of therapeutic benefit from alkylating chemotherapy O 6 ‐Methylguanine‐DNA‐methytransferase (MGMT) is a DNA repair enzyme that confers resistance to alkylating chemotherapy. Its expression is mainly regulated by epigenetic modification, and the MGMT gene promoter methylation status was introduced more than 15 years ago as a prognostic and predictive factor strongly associated with benefit from TMZ and CCNU . Patients with an MGMT promoter‐methylated (MGMT‐methylated) GBM receiving TMZ had a significantly longer median OS of up to 31.4 months compared to approximately 17 months in patients without MGMT promoter methylation ( MGMT ‐unmethylated) . As some benefit from TMZ cannot be excluded in patients with an unmethylated MGMT promoter , it is accepted that the MGMT promoter methylation status is not a prerequisite for applying TMZ in GBM first‐line treatment. This notion has some notable exceptions, indicating the first step toward a molecularly informed treatment of GBM. First, in patients > 65 years with an MGMT ‐unmethylated GBM, TMZ showed just under no significant benefit ( P = 0.055), and therefore, radiotherapy alone is a valid therapeutic option in this subgroup . The low activity of TMZ in patients with MGMT ‐unmethylated GBM prompted selective clinical trials for this subgroup, allowing the comparison of experimental treatment arms to placebo without TMZ in the standard arm . Second, the CeTeG/NOA‐09 phase III trial demonstrated an increased median OS of 48.1 months in patients with MGMT ‐methylated GBM receiving CCNU/TMZ, rendering this combined chemotherapy a treatment option selectively for patients belonging to this molecularly defined subgroup . Beyond MGMT gene promoter methylation analysis, the subclassification of brain tumors in general and GBM in particular into DNA methylation‐based subgroups has spawned hopes to identify further treatment‐guiding predictive patterns . The RTK II subgroup is enriched for epidermal growth factor receptor (EGFR) amplification and chromosome 10 loss, which corresponds to the ‘classic’ gene expression subtype described by Verhaak et al. . The RTK I subgroup is characterized by platelet‐derived growth factor (PDGFR) A amplification, corresponding to the ‘proneural’ expression subtype. The MES subgroup (‘mesenchymal’ expression subtype) frequently bears NF1 and PTEN alterations . Comparing radiotherapy versus TMZ in elderly GBM patients, the NOA‐08 trial found the prognostic impact of MGMT promoter methylation status was limited to GBM of the RTK II subgroup and was absent in the RTK I and MES subgroups. In addition, a biosimulation study predicted individual differential responses to CCNU/TMZ and TMZ treatments . These results require prospective validation and are met with skepticism because, despite a common genetic background, GBM cell states show considerable plasticity . 2.3 Genetic vulnerabilities and potential treatment targets in GBM The landscape of genetic alterations in GBM is well known, and clinically annotated expression and mutation data are readily available from TCGA and other data repositories . As shown in Fig. , genetic alterations in GBM frequently involve: Alterations in growth factor receptors/receptor tyrosine kinases (RTK). The most frequent example is EGFR, altered by approximately 60% . This includes amplification in about 40% , frequently associated with other alterations such as EGFR mutations or deletions, the most important being the EGFR variant III (EGFRvIII). While PDGFR is also frequently altered (10–15%), further alterations, including fibroblast growth factor receptor (FGFR, 2–5%), anaplastic lymphoma kinase (ALK), ROS‐1, RET, c‐Met, and neurotrophic tropomyosin receptor kinase (NTRK) 1–3 alterations, are rare . In the case of c‐Met (1–4%; ), FGFR3 (3%; ), and NTRK1‐3 (1–2%; ), alterations occur mostly in the form of gene fusions, which are also found for EGFR in 6–13% of patients . Downstream signal transduction cascades induced by RTK activation also frequently bear alterations. This particularly applies to the PI3K/PTEN/AKT/mTOR (phosphatidylinositol 3‐kinase/phosphatase and tensin homolog/AKT/mammalian target of rapamycin) pathway. Taking RTK, PI3K (25–30%, mainly PIK3CA or PIK3R1 alterations), and PTEN alterations (40%) together, at least one of these alterations is found in 90% of GBM . Neurofibromin 1 (NF1) mutations activating the PI3K pathway by reducing its RAS‐inhibiting effect have been found in 10% of cases . Genes encoding cell cycle proteins are frequently altered, such as cyclin‐dependent kinase inhibitor 2A/B (CDKN2A/B) deletion (60%) controlling cyclin‐dependent kinase 4/6 (CDK4/6), p53 mutation (20–25%), p53‐inhibiting amplifications of mouse double minute 2 homolog (MDM2) and MDM4 (15%), and RB1 mutation or deletion (8%; mutually exclusive with CDKN2A deletion). At least one of these genes is altered in 90% of GBM . Further alterations include DNA repair mechanisms such as mismatch repair deficiency in about 10% of progressive GBM, mostly due to MSH6 loss, and alterations of homologous repair deficiency, DNA checkpoint, and base excision repair . Illustrating the growing drug development pipeline in GBM, all of the aforementioned targets or their associated pathways are currently being investigated as potential treatment options . Despite this well characterized landscape of actionable treatment targets, the identification of successful targeted treatments for GBM remains challenging. In contrast to IDH mutant astrocytoma, where IDH mutation is thought to occur early in gliomagenesis, there is no known early – and thus major – single‐driver alteration in GBM . Further, GBM displays significant cellular and spatial heterogeneity, and potential targets may not be present in most tumor cells . The possibility of longitudinal heterogeneity represents another challenge; it was suggested that molecular targets profoundly change between the newly diagnosed and progressive disease , whereas prominent publications reported no substantial longitudinal changes in genetic alteration profiles . Besides the genetic profile, expression patterns of GBM (e.g. EGFR expression) and, in particular, cells of the tumor microenvironment may substantially change over time and have a vast influence on the composition of the tumor tissue and amenability to therapy, further introducing complexity . Current standard of care in the newly diagnosed and progressive setting Maximum‐safe tumor resection , radiotherapy of the tumor region , and concomitant and adjuvant temozolomide (TMZ) , an alkylating agent with high CNS penetrance and bioavailability, represent the current standard of care for first‐line GBM treatment . Tumor treating fields applying alternating electrical fields through the scalp are an additional treatment option . Furthermore, nitrosoureas such as lomustine (CCNU), also an alkylating agent with similarly favorable CNS penetrance, are available for first‐ or further‐line treatment . In some countries, the anti‐vascular endothelial growth factor (VEGF) A antibody bevacizumab is used as a non‐selective second‐line therapy, although it prolongs only PFS but not OS . The portfolio of further line therapies is enriched only by re‐resection , re‐irradiation , and possibly the multi‐kinase inhibitor regorafenib (see Section ), while the aggressive course of disease limits the time for further treatment lines. Thus, GBM patients need new and effective treatment options that should ideally be guided by predictive molecular characteristics to ensure that medications with a high likelihood of effectiveness are applied. Molecular predictors of therapeutic benefit from alkylating chemotherapy O 6 ‐Methylguanine‐DNA‐methytransferase (MGMT) is a DNA repair enzyme that confers resistance to alkylating chemotherapy. Its expression is mainly regulated by epigenetic modification, and the MGMT gene promoter methylation status was introduced more than 15 years ago as a prognostic and predictive factor strongly associated with benefit from TMZ and CCNU . Patients with an MGMT promoter‐methylated (MGMT‐methylated) GBM receiving TMZ had a significantly longer median OS of up to 31.4 months compared to approximately 17 months in patients without MGMT promoter methylation ( MGMT ‐unmethylated) . As some benefit from TMZ cannot be excluded in patients with an unmethylated MGMT promoter , it is accepted that the MGMT promoter methylation status is not a prerequisite for applying TMZ in GBM first‐line treatment. This notion has some notable exceptions, indicating the first step toward a molecularly informed treatment of GBM. First, in patients > 65 years with an MGMT ‐unmethylated GBM, TMZ showed just under no significant benefit ( P = 0.055), and therefore, radiotherapy alone is a valid therapeutic option in this subgroup . The low activity of TMZ in patients with MGMT ‐unmethylated GBM prompted selective clinical trials for this subgroup, allowing the comparison of experimental treatment arms to placebo without TMZ in the standard arm . Second, the CeTeG/NOA‐09 phase III trial demonstrated an increased median OS of 48.1 months in patients with MGMT ‐methylated GBM receiving CCNU/TMZ, rendering this combined chemotherapy a treatment option selectively for patients belonging to this molecularly defined subgroup . Beyond MGMT gene promoter methylation analysis, the subclassification of brain tumors in general and GBM in particular into DNA methylation‐based subgroups has spawned hopes to identify further treatment‐guiding predictive patterns . The RTK II subgroup is enriched for epidermal growth factor receptor (EGFR) amplification and chromosome 10 loss, which corresponds to the ‘classic’ gene expression subtype described by Verhaak et al. . The RTK I subgroup is characterized by platelet‐derived growth factor (PDGFR) A amplification, corresponding to the ‘proneural’ expression subtype. The MES subgroup (‘mesenchymal’ expression subtype) frequently bears NF1 and PTEN alterations . Comparing radiotherapy versus TMZ in elderly GBM patients, the NOA‐08 trial found the prognostic impact of MGMT promoter methylation status was limited to GBM of the RTK II subgroup and was absent in the RTK I and MES subgroups. In addition, a biosimulation study predicted individual differential responses to CCNU/TMZ and TMZ treatments . These results require prospective validation and are met with skepticism because, despite a common genetic background, GBM cell states show considerable plasticity . Genetic vulnerabilities and potential treatment targets in GBM The landscape of genetic alterations in GBM is well known, and clinically annotated expression and mutation data are readily available from TCGA and other data repositories . As shown in Fig. , genetic alterations in GBM frequently involve: Alterations in growth factor receptors/receptor tyrosine kinases (RTK). The most frequent example is EGFR, altered by approximately 60% . This includes amplification in about 40% , frequently associated with other alterations such as EGFR mutations or deletions, the most important being the EGFR variant III (EGFRvIII). While PDGFR is also frequently altered (10–15%), further alterations, including fibroblast growth factor receptor (FGFR, 2–5%), anaplastic lymphoma kinase (ALK), ROS‐1, RET, c‐Met, and neurotrophic tropomyosin receptor kinase (NTRK) 1–3 alterations, are rare . In the case of c‐Met (1–4%; ), FGFR3 (3%; ), and NTRK1‐3 (1–2%; ), alterations occur mostly in the form of gene fusions, which are also found for EGFR in 6–13% of patients . Downstream signal transduction cascades induced by RTK activation also frequently bear alterations. This particularly applies to the PI3K/PTEN/AKT/mTOR (phosphatidylinositol 3‐kinase/phosphatase and tensin homolog/AKT/mammalian target of rapamycin) pathway. Taking RTK, PI3K (25–30%, mainly PIK3CA or PIK3R1 alterations), and PTEN alterations (40%) together, at least one of these alterations is found in 90% of GBM . Neurofibromin 1 (NF1) mutations activating the PI3K pathway by reducing its RAS‐inhibiting effect have been found in 10% of cases . Genes encoding cell cycle proteins are frequently altered, such as cyclin‐dependent kinase inhibitor 2A/B (CDKN2A/B) deletion (60%) controlling cyclin‐dependent kinase 4/6 (CDK4/6), p53 mutation (20–25%), p53‐inhibiting amplifications of mouse double minute 2 homolog (MDM2) and MDM4 (15%), and RB1 mutation or deletion (8%; mutually exclusive with CDKN2A deletion). At least one of these genes is altered in 90% of GBM . Further alterations include DNA repair mechanisms such as mismatch repair deficiency in about 10% of progressive GBM, mostly due to MSH6 loss, and alterations of homologous repair deficiency, DNA checkpoint, and base excision repair . Illustrating the growing drug development pipeline in GBM, all of the aforementioned targets or their associated pathways are currently being investigated as potential treatment options . Despite this well characterized landscape of actionable treatment targets, the identification of successful targeted treatments for GBM remains challenging. In contrast to IDH mutant astrocytoma, where IDH mutation is thought to occur early in gliomagenesis, there is no known early – and thus major – single‐driver alteration in GBM . Further, GBM displays significant cellular and spatial heterogeneity, and potential targets may not be present in most tumor cells . The possibility of longitudinal heterogeneity represents another challenge; it was suggested that molecular targets profoundly change between the newly diagnosed and progressive disease , whereas prominent publications reported no substantial longitudinal changes in genetic alteration profiles . Besides the genetic profile, expression patterns of GBM (e.g. EGFR expression) and, in particular, cells of the tumor microenvironment may substantially change over time and have a vast influence on the composition of the tumor tissue and amenability to therapy, further introducing complexity . Targeted treatment in untargeted study populations 3.1 GBM trials with targeted agents In the last 20 years, targeted drugs have been applied to GBM patients without individual prior target verification, and they have been mostly explored in patients with progressive or recurring tumors following one or more treatment lines. Table provides an overview of drugs, targets, and observed outcomes, focusing on GBM hallmark alterations such as EGFR, PDGFR, FGFR, c‐MET, and the PIK3CA/Akt/mTOR pathway (see also Fig. ). Beyond this, many drugs directed at less GBM‐specific targets involved in tumor cell growth and/or homeostasis have been evaluated, e.g. transforming growth factor β (TGFβ)‐directed galunisertib , CD95 ligand (CD95L)‐directed APG101 , Src‐directed dasatinib , phosphorylated signal transducer and activator of transcription 3 (pSTAT3)‐directed therapy , hepatocyte growth factor/scatter factor (HGF/SF)‐directed rilotumumab , and proteasome inhibitors such as marizomib . In addition, there have been several approaches to target receptors thought to be mostly expressed in tumor cells (without individual confirmation beforehand) with locally applied ligand‐toxin fusion proteins targeting, e.g. interleukin 13‐receptor (IL13R), interleukin 4‐receptor (IL4R), or transferrin receptors . In summary, these approaches failed to achieve convincing results in adult malignant glioma cohorts. Still, some trials performed post‐hoc secondary explorative analyses to identify biomarkers for treatment benefit , thus justifying the use of the mTOR inhibitor temsirolimus in the ongoing N2M2 trial and indicating a PFS‐prolonging effect of the CD95L‐inhibitor APG101 . In retrospect, the failure of trials without prior target verification, and therefore without enrichment of tumors harboring the targeted alteration, comes as no surprise. Thus, targeted agents must be tested in cohorts preselected for the presence of the targeted genetic alterations. The only exceptions are multi‐kinase inhibitors with a broad spectrum of therapeutic targets, which enable studies in unselected cohorts. Regorafenib, a multi‐kinase inhibitor targeting VEGFR1‐3, TIE2, PDGFR‐β, FGFR, KIT, RET, and RAF, increased OS in the progressive setting in the randomized phase II REGOMA trial . While questions remain as to the extent to which these positive results rely on the VEGF‐directed antiangiogenic effect and the results have recently been challenged , some markers of therapeutic benefit have emerged in explorative analyses. More specifically, the occurrence of a hand‐foot reaction, a common side effect observed in approx. 30% of patients receiving regorafenib, was associated with an increased OS of 6.7 versus 2.6 months in a small retrospective bicentric cohort of patients with progressive glioblastoma receiving regorafenib, and a biomarker analysis of the REGOMA trial described the expression levels of several mRNAs and miRNAs to be associated with survival . GBM trials with targeted agents In the last 20 years, targeted drugs have been applied to GBM patients without individual prior target verification, and they have been mostly explored in patients with progressive or recurring tumors following one or more treatment lines. Table provides an overview of drugs, targets, and observed outcomes, focusing on GBM hallmark alterations such as EGFR, PDGFR, FGFR, c‐MET, and the PIK3CA/Akt/mTOR pathway (see also Fig. ). Beyond this, many drugs directed at less GBM‐specific targets involved in tumor cell growth and/or homeostasis have been evaluated, e.g. transforming growth factor β (TGFβ)‐directed galunisertib , CD95 ligand (CD95L)‐directed APG101 , Src‐directed dasatinib , phosphorylated signal transducer and activator of transcription 3 (pSTAT3)‐directed therapy , hepatocyte growth factor/scatter factor (HGF/SF)‐directed rilotumumab , and proteasome inhibitors such as marizomib . In addition, there have been several approaches to target receptors thought to be mostly expressed in tumor cells (without individual confirmation beforehand) with locally applied ligand‐toxin fusion proteins targeting, e.g. interleukin 13‐receptor (IL13R), interleukin 4‐receptor (IL4R), or transferrin receptors . In summary, these approaches failed to achieve convincing results in adult malignant glioma cohorts. Still, some trials performed post‐hoc secondary explorative analyses to identify biomarkers for treatment benefit , thus justifying the use of the mTOR inhibitor temsirolimus in the ongoing N2M2 trial and indicating a PFS‐prolonging effect of the CD95L‐inhibitor APG101 . In retrospect, the failure of trials without prior target verification, and therefore without enrichment of tumors harboring the targeted alteration, comes as no surprise. Thus, targeted agents must be tested in cohorts preselected for the presence of the targeted genetic alterations. The only exceptions are multi‐kinase inhibitors with a broad spectrum of therapeutic targets, which enable studies in unselected cohorts. Regorafenib, a multi‐kinase inhibitor targeting VEGFR1‐3, TIE2, PDGFR‐β, FGFR, KIT, RET, and RAF, increased OS in the progressive setting in the randomized phase II REGOMA trial . While questions remain as to the extent to which these positive results rely on the VEGF‐directed antiangiogenic effect and the results have recently been challenged , some markers of therapeutic benefit have emerged in explorative analyses. More specifically, the occurrence of a hand‐foot reaction, a common side effect observed in approx. 30% of patients receiving regorafenib, was associated with an increased OS of 6.7 versus 2.6 months in a small retrospective bicentric cohort of patients with progressive glioblastoma receiving regorafenib, and a biomarker analysis of the REGOMA trial described the expression levels of several mRNAs and miRNAs to be associated with survival . Targeted treatment with previous target verification 4.1 Successful trials with molecularly matched drugs in glioma patients There are a few success stories emphasizing that targeted therapy may show efficacy in molecularly selected glioma subpopulations. In patients with tuberous sclerosis, treatment with the mTOR inhibitor everolimus for subependymal giant cell astrocytomas with alterations in the mTOR pathway is well established and leads to tumor reduction of ≥ 30% in 75% of patients . More recently, the IDH inhibitor vorasidenib increased PFS from 11.1 to 27.7 months in IDH‐mutant grade 2 glioma and allowed for significantly delayed further interventions (likelihood of next treatment intervention or death by 24 months, 16.6% vs. 73%) . In GBM, the only successful molecularly matched treatment to date is combined BRAF/MEK inhibition in patients with a constitutively activated MAPK pathway due to a BRAF V600E mutation. An interim analysis of the single‐arm phase 2 ROAR basket trial exploring this approach in BRAF V600E‐mutated progressive GBM with the BRAF inhibitor dabrafenib and the MEK1/2 (the downstream target of BRAF) inhibitor trametinib showed an objective response rate (ORR, complete or partial response according to RANO criteria) of 32% and a PFS of 2.8 and an OS of 13.7 months . Consequently, EANO guidelines conclude that the clinical benefit in patients with BRAF V600E mutant progressive CNS tumors is sufficiently well established to consider it part of the standard of care . 4.2 Entity‐agnostic drug registrations for patients with NTRK gene fusions or microsatellite instability/mismatch repair deficiency While biomarker‐specific drugs are usually marketed for specific cancer types, two drugs received an entity‐agnostic registration: larotrectinib, a tropomyosin kinase receptor inhibitor for tumors bearing NTRK alterations, and the programmed cell death protein 1 (PD‐1) inhibitor pembrolizumab for tumors with microsatellite instability/mismatch repair deficiency. Data are accumulating that in childhood gliomas, larotrectinib addressing NTRK alterations (mostly NTRK fusion transcripts) may induce a high rate of responses , but reports on adult GBM patients remain anecdotal, with the largest series reporting disease stabilization (> 6 months) in 4 of 6 patients (Table ) . A larger series will hopefully provide more reliable information on the efficacy of larotrectinib in this setting. The case is even more challenging for pembrolizumab in patients with microsatellite instability/mismatch repair‐deficient tumors. There are currently no strong data supporting this concept in GBM, as the clinical trial supporting pembrolizumab treatment did not include any GBM patients . Of note, untargeted PD‐1 immune checkpoint inhibition with nivolumab has been extensively studied in newly diagnosed glioblastoma, but failed to prolong survival in large phase 3 trials in MGMT ‐unmethylated as well as MGMT ‐methylated newly diagnosed GBM patients . Also, the trial investigating nivolumab at the first relapse of GBM did not show any survival prolongation . 4.3 Lessons learned from unsuccessful trials with molecularly matched drugs Further trials evaluating targeted treatments in GBM patients with target verification have shown no convincing efficacy data so far (Table ). This applies to negative data in randomized phase 3 trials investigating depatuxizumab mafodotin, an antibody‐drug conjugate composed of an anti‐EGFR antibody conjugated to a tubulin inhibitor, which found no OS improvement in newly diagnosed GBM with confirmed EGFR‐amplification (18.9 vs. 18.7 months, PFS 8.0 vs. 6.3 months), or rindopepimut, an EGFRvIII‐specific peptide vaccine, which showed no benefit in newly diagnosed EGFRvIII‐positive GBM (OS 20.1 vs. 20.0 months, PFS 7.1 vs. 5.6 months) . For single‐arm trials, an ORR of 25%, which translates to a median OS of 15 months , or surpassing a 6‐month PFS rate of 16–20% as observed with CCNU , is generally expected for an effective second‐ or later‐line treatment. Examples of single‐arm trials not reaching this threshold are given in Table and include targeting of EGFR, CDK4/6, and amplified c‐MET. The multifaceted problem of achieving efficacy in trials with molecularly matched treatments is highlighted by the largely unsuccessful therapy of EGFR‐altered GBM despite previous target identification (EGFR amplification in 54%; ), which has taught many lessons in this regard: Successful EGFR‐directed therapy in NSCLC is applied in the context of mutations in the tyrosine kinase domain, which activate the receptor. In contrast, EGFR alterations in GBM mainly affect the extracellular domain, and multiple different oncogenic EGFR variants (mostly deletions and missense mutations) typically coexist and are not homogenously distributed . Mechanistically, EGFR alterations in GBM seem to alter ligand discrimination , suggesting alternative mechanisms in response to EGFR‐directed therapy in GBM compared to NSCLC (for review, see ). EGFR alterations in GBM may not confer oncologic addiction, as they are considered late events in gliomagenesis and are subclonal rather than clonal . This implies that the tumor is a mosaic of cells with different RTK alterations that may cooperate synergistically , increasing cellular fitness and resistance to therapy . The resulting spatial heterogeneity renders biopsy sampling less reliable. Further, it implies temporal heterogeneity during further tumor growth, where relapsed GBM may show a loss of mutated targets . This would require target verification in the progressive tumor rather than based on tissue from the primary surgery. While EGFR amplification is usually conserved , some trials have shown that staining of the EGFR extracellular domain is changed or lost upon therapy and is not associated with a clinically relevant survival prolongation . The treatment of GBM with at least some part of the tumor behind an intact blood–brain barrier represents a pharmacokinetic challenge, as many compounds may not sufficiently penetrate CNS tumors and have a reduced bioavailability, even in the case of in principle sufficient CNS penetration of, e.g., the EGFR inhibitors erlotinib or osimertinib . This contributes to the observation that tumor tissue obtained after EGFR‐directed therapy does not show sufficient target engagement and pathway alterations . The difficulties of EGFR‐directed therapies have been so substantial that new approaches for EGFR targeting, such as EGFR‐directed chimeric antigen receptor (CAR) T cell therapy, are met with caution. The increasing number of available targeted drugs leads to the question of how to effectively scan for potentially successful target/drug combinations in GBM. Adaptive phase II trials in newly diagnosed MGMT ‐unmethylated GBM are promising tools to identify drugs for further analysis in confirmatory phase III trials. Examples of such trials are N2M2, INSIGhT, and the Adaptive Global Innovative Learning Environment for Glioblastoma (GBM AGILE), which investigate several targeted drugs in parallel with obligatory molecular testing and a common temozolomide standard arm . These trials are even more intriguing because their results are continuously monitored, and a Bayesian approach is applied to guide the allocation of patients to more successful trial arms. The first results of the INSIGhT trial have already been published, showing superior PFS but similar OS for the CDK4/6 inhibitor abemaciclib (OS 15.3 vs. 14.8 months, PFS 6.2. vs. 4.7 months) and the EGFR/HER2 inhibitor neratinib (OS 14.2 vs. 14.8 months, PFS 6.0 vs. 4.7 months), both in addition to standard radiochemotherapy . However, preliminary results of GBM AGILE challenge the benefit of regorafenib, as mentioned above . Successful trials with molecularly matched drugs in glioma patients There are a few success stories emphasizing that targeted therapy may show efficacy in molecularly selected glioma subpopulations. In patients with tuberous sclerosis, treatment with the mTOR inhibitor everolimus for subependymal giant cell astrocytomas with alterations in the mTOR pathway is well established and leads to tumor reduction of ≥ 30% in 75% of patients . More recently, the IDH inhibitor vorasidenib increased PFS from 11.1 to 27.7 months in IDH‐mutant grade 2 glioma and allowed for significantly delayed further interventions (likelihood of next treatment intervention or death by 24 months, 16.6% vs. 73%) . In GBM, the only successful molecularly matched treatment to date is combined BRAF/MEK inhibition in patients with a constitutively activated MAPK pathway due to a BRAF V600E mutation. An interim analysis of the single‐arm phase 2 ROAR basket trial exploring this approach in BRAF V600E‐mutated progressive GBM with the BRAF inhibitor dabrafenib and the MEK1/2 (the downstream target of BRAF) inhibitor trametinib showed an objective response rate (ORR, complete or partial response according to RANO criteria) of 32% and a PFS of 2.8 and an OS of 13.7 months . Consequently, EANO guidelines conclude that the clinical benefit in patients with BRAF V600E mutant progressive CNS tumors is sufficiently well established to consider it part of the standard of care . Entity‐agnostic drug registrations for patients with NTRK gene fusions or microsatellite instability/mismatch repair deficiency While biomarker‐specific drugs are usually marketed for specific cancer types, two drugs received an entity‐agnostic registration: larotrectinib, a tropomyosin kinase receptor inhibitor for tumors bearing NTRK alterations, and the programmed cell death protein 1 (PD‐1) inhibitor pembrolizumab for tumors with microsatellite instability/mismatch repair deficiency. Data are accumulating that in childhood gliomas, larotrectinib addressing NTRK alterations (mostly NTRK fusion transcripts) may induce a high rate of responses , but reports on adult GBM patients remain anecdotal, with the largest series reporting disease stabilization (> 6 months) in 4 of 6 patients (Table ) . A larger series will hopefully provide more reliable information on the efficacy of larotrectinib in this setting. The case is even more challenging for pembrolizumab in patients with microsatellite instability/mismatch repair‐deficient tumors. There are currently no strong data supporting this concept in GBM, as the clinical trial supporting pembrolizumab treatment did not include any GBM patients . Of note, untargeted PD‐1 immune checkpoint inhibition with nivolumab has been extensively studied in newly diagnosed glioblastoma, but failed to prolong survival in large phase 3 trials in MGMT ‐unmethylated as well as MGMT ‐methylated newly diagnosed GBM patients . Also, the trial investigating nivolumab at the first relapse of GBM did not show any survival prolongation . Lessons learned from unsuccessful trials with molecularly matched drugs Further trials evaluating targeted treatments in GBM patients with target verification have shown no convincing efficacy data so far (Table ). This applies to negative data in randomized phase 3 trials investigating depatuxizumab mafodotin, an antibody‐drug conjugate composed of an anti‐EGFR antibody conjugated to a tubulin inhibitor, which found no OS improvement in newly diagnosed GBM with confirmed EGFR‐amplification (18.9 vs. 18.7 months, PFS 8.0 vs. 6.3 months), or rindopepimut, an EGFRvIII‐specific peptide vaccine, which showed no benefit in newly diagnosed EGFRvIII‐positive GBM (OS 20.1 vs. 20.0 months, PFS 7.1 vs. 5.6 months) . For single‐arm trials, an ORR of 25%, which translates to a median OS of 15 months , or surpassing a 6‐month PFS rate of 16–20% as observed with CCNU , is generally expected for an effective second‐ or later‐line treatment. Examples of single‐arm trials not reaching this threshold are given in Table and include targeting of EGFR, CDK4/6, and amplified c‐MET. The multifaceted problem of achieving efficacy in trials with molecularly matched treatments is highlighted by the largely unsuccessful therapy of EGFR‐altered GBM despite previous target identification (EGFR amplification in 54%; ), which has taught many lessons in this regard: Successful EGFR‐directed therapy in NSCLC is applied in the context of mutations in the tyrosine kinase domain, which activate the receptor. In contrast, EGFR alterations in GBM mainly affect the extracellular domain, and multiple different oncogenic EGFR variants (mostly deletions and missense mutations) typically coexist and are not homogenously distributed . Mechanistically, EGFR alterations in GBM seem to alter ligand discrimination , suggesting alternative mechanisms in response to EGFR‐directed therapy in GBM compared to NSCLC (for review, see ). EGFR alterations in GBM may not confer oncologic addiction, as they are considered late events in gliomagenesis and are subclonal rather than clonal . This implies that the tumor is a mosaic of cells with different RTK alterations that may cooperate synergistically , increasing cellular fitness and resistance to therapy . The resulting spatial heterogeneity renders biopsy sampling less reliable. Further, it implies temporal heterogeneity during further tumor growth, where relapsed GBM may show a loss of mutated targets . This would require target verification in the progressive tumor rather than based on tissue from the primary surgery. While EGFR amplification is usually conserved , some trials have shown that staining of the EGFR extracellular domain is changed or lost upon therapy and is not associated with a clinically relevant survival prolongation . The treatment of GBM with at least some part of the tumor behind an intact blood–brain barrier represents a pharmacokinetic challenge, as many compounds may not sufficiently penetrate CNS tumors and have a reduced bioavailability, even in the case of in principle sufficient CNS penetration of, e.g., the EGFR inhibitors erlotinib or osimertinib . This contributes to the observation that tumor tissue obtained after EGFR‐directed therapy does not show sufficient target engagement and pathway alterations . The difficulties of EGFR‐directed therapies have been so substantial that new approaches for EGFR targeting, such as EGFR‐directed chimeric antigen receptor (CAR) T cell therapy, are met with caution. The increasing number of available targeted drugs leads to the question of how to effectively scan for potentially successful target/drug combinations in GBM. Adaptive phase II trials in newly diagnosed MGMT ‐unmethylated GBM are promising tools to identify drugs for further analysis in confirmatory phase III trials. Examples of such trials are N2M2, INSIGhT, and the Adaptive Global Innovative Learning Environment for Glioblastoma (GBM AGILE), which investigate several targeted drugs in parallel with obligatory molecular testing and a common temozolomide standard arm . These trials are even more intriguing because their results are continuously monitored, and a Bayesian approach is applied to guide the allocation of patients to more successful trial arms. The first results of the INSIGhT trial have already been published, showing superior PFS but similar OS for the CDK4/6 inhibitor abemaciclib (OS 15.3 vs. 14.8 months, PFS 6.2. vs. 4.7 months) and the EGFR/HER2 inhibitor neratinib (OS 14.2 vs. 14.8 months, PFS 6.0 vs. 4.7 months), both in addition to standard radiochemotherapy . However, preliminary results of GBM AGILE challenge the benefit of regorafenib, as mentioned above . Experience with NGS screening and matched targeted therapy 5.1 Important points to consider for NGS‐based individual GBM therapy Instead of testing single or few drugs in cohorts selected by screening for a single molecular alteration, next‐generation sequencing (NGS) yields an array of genetic alterations for each patient, which may allow the selection of the most promising genetic alteration/targeted drug combination (matched therapy). This approach is an attractive way to evaluate the concept of precision oncology in GBM therapy. In cancer entity‐agnostic case series of patients with metastatic cancer (without GBM patients) such as IMPACT , the ORR, 6‐month stabilization rate, median PFS and OS, and 10‐year survival rate of patients receiving matched therapy tended to be higher than those of patients receiving nonmatched therapies. The first steps are made to implement the approach for primary brain tumors in general and gliomas in particular (Table ). To extend this approach to GBM, three major problems have to be addressed: Target identification. In GBM, it is not trivial to infer the most promising target constellation from a list of genetic alterations and whether single or combined alterations represent the best target. For example, it is unknown whether CDK4/6 inhibitors such as palbociclib or abemaciclib can be employed for RB1‐proficient GBM or if CDKN2A/B and CDK4/6 status also need to be considered . Dysregulation of the CDK4/6‐p16‐RB1 pathway is a hallmark of glioblastoma . While CDK4/6 activation inhibits the tumor suppressor protein RB1, allowing cell cycle progression, CDK4/6 inhibitors cause reduced RB1 phosphorylation and apoptosis. Homozygous deletion of CDKN2A/B, encoding the CDK4/6 inhibitor p16, leads to CDK4/6 disinhibition, which might be required for sensitivity to pharmacological CDK4/6 inhibition . CDK4 alterations or RB1 mutations were associated with resistance to CDK4/6 inhibition in patient‐derived GBM xenografts . Similarly, should application of EGFR block in patients with EGFR amplification or activating mutations be given on the base PTEN alterations, which are frequent in GBM and linked to reduced responsiveness to EGFR inhibitors ? Clear guidelines for these decisions are lacking. Of note, combined target selection inevitably narrows down treatment options for individual patients. Treatment selection for precision oncology is challenging, irrespective of cancer type. In GBM, this problem may be accentuated as there has only been one successful molecularly guided trial thus far . Therefore, treatment selection mostly has to rely on results from other tumor entities, e.g. breast cancer for DNA damage repair alterations, cholangiocarcinoma and urothelial carcinoma for FGFR alterations, and NSCLC for EGFR alterations . However, applicability may be reduced due to differences in mutation sites, activation of compensatory pathways, and tissue penetration. Treatment selection based on preclinical in vitro / in vivo results or on biological rationale is even less convincing and leads to a lower strength of recommendation according to current grading guidance . In summary, the paucity of data requires the adoption of treatment strategies based on other tumor entities or preclinical data, both with reduced applicability for GBM. Efficacy assessment. The assessment of treatment success by standard metrics such as ORR, PFS (both largely based on imaging parameters), or OS rates is limited in heterogeneous cohorts of GBM patients receiving precision oncology treatment. Further, interindividual comparison is impaired by the potential prognostic effect of the targeted molecular alterations and the treatment of patients at differing stages of their illness. Intraindividual comparison of PFS until first progression to PFS under matched therapy (PFS2/PFS1) is an elegant alternative, and a PFS2/PFS1 ratio of 1.3 or higher has been accepted as a marker of effective matched therapy in systemic cancers . Considering the progression time scale with a median PFS of 7 months in newly diagnosed GBM and 2 months in progressive GBM, a PFS2/PFS1 ratio of 1.3 translates to a significant PFS increment if the evaluated treatment is initiated at first recurrence and to a numerically small PFS increment if initiated at further recurrence . Accordingly, modifications have been suggested to adjust for very short PFS1 < 2 months to prevent overcalling and for significant PFS2 > 6 months to prevent undercalling of treatment responses in brain tumors, and the concept has been applied in brain tumors . Further approaches include an adaptation of the ESMO Magnitude in Clinical Benefit Scale to brain tumors, combining PFS with imaging response duration (Neuro‐MCBS, see Ref. ). 5.2 Case series with NGS‐based molecular‐guided GBM therapy in clinical routine An increasing number of neuro‐oncology centers offer NGS screening for actionable mutations to patients with progressive glioblastoma with no further standard treatment options. At these institutions, multidisciplinary molecular tumor boards are established, providing personalized recommendations for targeted therapies based on individual NGS results. So far, five publications report on mono‐ or oligoinstitutional experience with NGS‐informed personalized therapy : Blumenthal et al. reported the first retrospective cohort from five tertiary hospitals comprising 43 glioma patients (34 GBM), where a NGS panel detected actionable alterations in 95%, leading to targeted treatment in 30% (10/34), but without any treatment response. Byron et al. performed a prospective monocentric trial to evaluate the feasibility of whole exome sequencing‐informed treatment recommendations within 35 days of surgery. Among 16 GBM patients, a recommendation was possible in 94%, 44% (7/16) received a targeted treatment, and one patient (6%) achieved a treatment response. Lazaridis et al. reported a retrospective monocentric cohort of 41 glioma (32 GBM) patients. Following NGS and further methods of genomic profiling, actionable targets were identified in 76% (24/32) and 16 GBM patients receiving targeted treatment achieved an increased PFS (3.8 vs. 2.0 months) and OS (13 vs. 4 months) compared to 16 GBM patients with unmatched empiric treatment. Renovanz et al. reported on their experience from the Center for Personalized Medicine Tübingen, which has an established certified clinical workflow for personalized medicine in the clinical routine for cancer patients without options for trial participation or further registered treatments. This heterogeneous and heavily pretreated cohort included 262 GBM . Following comprehensive molecular profiling, molecularly instructed treatment recommendations were made in 93% (243/262) and 41 GBM patients were treated accordingly, resulting in a PFS2/PFS1 > 1.3 in 36% (13 of 36 evaluable patients). Padovan et al. reported a retrospective, monocentric cohort of 417 GBM patients receiving NGS screening. While actionable targets were identified in 82%, 36 patients (8.6%) received a targeted treatment, of which 20% (7/36) achieved a PFS2/PFS1 > 1.3. The five series are described in more detail in Table . Of note, only two of these studies were prospectively documented , and the number of evaluable patients with targeted therapy per cohort remains low, ranging from < 20 to 36 patients . The largest studies to date also highlight the current problem of precision therapy. As reported by Renovanz et al. , among 262 GBM patients receiving NGS screening, only 41 actually started matched therapy (about 14% of patients tested). While a high percentage of these (88%, n = 36) could be evaluated for efficacy and demonstrate the determination of the authors, the low rate of initiated therapies emphasizes the many obstacles for molecularly matched therapy, such as the rapid deterioration of patients with progressive GBM and the lack of reimbursement by health insurance companies attributable to insufficient GBM‐specific evidence of efficacy . Padovan et al. reported a similar experience, where only 36 (8.6%) of 417 GBM patients receiving NGS screening were able to initiate matched therapy. No treatment response was observed by Blumenthal et al., while Byron et al. report a single patient receiving a potentially successful treatment of olaparib/trametinib/carboplatin with a PFS2/PFS1 ratio of > 3 . The larger studies provide some data on the efficacy of matched therapy in GBM . Lazaridis et al. , also considering CNS drug penetration in the decision‐making process, presented 16 patients receiving matched therapy compared to 16 patients with unmatched therapy (Table ), potentially introducing selection bias. The encouraging PFS results of 3.8 and 2.0 months with versus without matched therapy were mainly driven by the efficacy of BRAF V600E‐directed therapy with dabrafenib/trametinib and c‐Met‐directed therapy with the tyrosine kinase inhibitor (TKI) cabozantinib. While the first observation is in line with the interim results from the ROAR trial discussed in Section , the cabozantinib results warrant critical discussion as the drug targets not only c‐Met but also VEGFR. Inhibition of the VEGF pathway, e.g. with the VEGF‐A antibody bevacizumab, is well known to prolong PFS but not OS . The intermingling of VEGF‐directed therapy with therapy directed at other targets was also present in the study by Renovanz et al. . As mentioned, 14% of GBM patients started matched therapy, and 13/36 evaluable patients (5% of screened patients or 36% of patients receiving treatment and being evaluable) were successfully treated according to a PFS2/PFS1 ratio > 1.3, while the median PFS was 2.3 months . Here, three of the four patients receiving regorafenib based on FMS‐like tyrosine kinase (FLT) or EGFR alterations did benefit (PFS2/PFS1 > 1.3). As discussed above, regorafenib targets several tyrosine kinases, including VEGFR, raising the possibility that the observed PFS benefit might at least partially be attributable to VEGFR targeting. Apart from this, a signal for treatment benefit was only seen for everolimus in tumors with PTEN or PIK3CA alterations (3/6 with PFS2/PFS1 > 1.3), which is in line with results from the mTOR inhibitor temsirolimus trial in newly diagnosed MGMT ‐unmethylated GBM , and for tumors with FGFR fusion transcripts treated with the FGFR inhibitor erdafitinib (3/4 with PFS2/PFS1 > 1.3). Two patients had a positive PFS2/PFS1 signal with the EGFR antibody‐toxin conjugate depatuxizumab mafodotin in the context of an early‐access program, illustrating its PFS prolongation in a previous trial , which did not translate into a longer OS in large phase III trials (Table ). Padovan et al. confirm these observations: among the 36/417 patients receiving targeted treatment, 19% were treated successfully with a PFS2/PFS1 ratio > 1.3, including three objective responses, again mainly driven by dabrafenib/trametinib in BRAF V600E‐altered GBM and erdafitinib in FGFR3‐altered GBM, while the median PFS was 2.1 months. Other publications report the results of NGS screening and/or matched therapy allocation without including a substantial number of GBM patients evaluable for treatment efficacy . Using a methodologically different approach, Luger et al. retrospectively analyzed a cohort of 351 patients treated with off‐label therapy and identified 15 patients with high‐grade glioma (8 GBM) who received matched therapy. This series was dominated by the observation that 3/6 patients treated with BRAF V600E‐directed therapy had disease stability for 5+ months. In summary, these reports highlight that few NGS‐screened GBM patients receive targeted treatment, and even fewer may benefit from it. Important points to consider for NGS‐based individual GBM therapy Instead of testing single or few drugs in cohorts selected by screening for a single molecular alteration, next‐generation sequencing (NGS) yields an array of genetic alterations for each patient, which may allow the selection of the most promising genetic alteration/targeted drug combination (matched therapy). This approach is an attractive way to evaluate the concept of precision oncology in GBM therapy. In cancer entity‐agnostic case series of patients with metastatic cancer (without GBM patients) such as IMPACT , the ORR, 6‐month stabilization rate, median PFS and OS, and 10‐year survival rate of patients receiving matched therapy tended to be higher than those of patients receiving nonmatched therapies. The first steps are made to implement the approach for primary brain tumors in general and gliomas in particular (Table ). To extend this approach to GBM, three major problems have to be addressed: Target identification. In GBM, it is not trivial to infer the most promising target constellation from a list of genetic alterations and whether single or combined alterations represent the best target. For example, it is unknown whether CDK4/6 inhibitors such as palbociclib or abemaciclib can be employed for RB1‐proficient GBM or if CDKN2A/B and CDK4/6 status also need to be considered . Dysregulation of the CDK4/6‐p16‐RB1 pathway is a hallmark of glioblastoma . While CDK4/6 activation inhibits the tumor suppressor protein RB1, allowing cell cycle progression, CDK4/6 inhibitors cause reduced RB1 phosphorylation and apoptosis. Homozygous deletion of CDKN2A/B, encoding the CDK4/6 inhibitor p16, leads to CDK4/6 disinhibition, which might be required for sensitivity to pharmacological CDK4/6 inhibition . CDK4 alterations or RB1 mutations were associated with resistance to CDK4/6 inhibition in patient‐derived GBM xenografts . Similarly, should application of EGFR block in patients with EGFR amplification or activating mutations be given on the base PTEN alterations, which are frequent in GBM and linked to reduced responsiveness to EGFR inhibitors ? Clear guidelines for these decisions are lacking. Of note, combined target selection inevitably narrows down treatment options for individual patients. Treatment selection for precision oncology is challenging, irrespective of cancer type. In GBM, this problem may be accentuated as there has only been one successful molecularly guided trial thus far . Therefore, treatment selection mostly has to rely on results from other tumor entities, e.g. breast cancer for DNA damage repair alterations, cholangiocarcinoma and urothelial carcinoma for FGFR alterations, and NSCLC for EGFR alterations . However, applicability may be reduced due to differences in mutation sites, activation of compensatory pathways, and tissue penetration. Treatment selection based on preclinical in vitro / in vivo results or on biological rationale is even less convincing and leads to a lower strength of recommendation according to current grading guidance . In summary, the paucity of data requires the adoption of treatment strategies based on other tumor entities or preclinical data, both with reduced applicability for GBM. Efficacy assessment. The assessment of treatment success by standard metrics such as ORR, PFS (both largely based on imaging parameters), or OS rates is limited in heterogeneous cohorts of GBM patients receiving precision oncology treatment. Further, interindividual comparison is impaired by the potential prognostic effect of the targeted molecular alterations and the treatment of patients at differing stages of their illness. Intraindividual comparison of PFS until first progression to PFS under matched therapy (PFS2/PFS1) is an elegant alternative, and a PFS2/PFS1 ratio of 1.3 or higher has been accepted as a marker of effective matched therapy in systemic cancers . Considering the progression time scale with a median PFS of 7 months in newly diagnosed GBM and 2 months in progressive GBM, a PFS2/PFS1 ratio of 1.3 translates to a significant PFS increment if the evaluated treatment is initiated at first recurrence and to a numerically small PFS increment if initiated at further recurrence . Accordingly, modifications have been suggested to adjust for very short PFS1 < 2 months to prevent overcalling and for significant PFS2 > 6 months to prevent undercalling of treatment responses in brain tumors, and the concept has been applied in brain tumors . Further approaches include an adaptation of the ESMO Magnitude in Clinical Benefit Scale to brain tumors, combining PFS with imaging response duration (Neuro‐MCBS, see Ref. ). Case series with NGS‐based molecular‐guided GBM therapy in clinical routine An increasing number of neuro‐oncology centers offer NGS screening for actionable mutations to patients with progressive glioblastoma with no further standard treatment options. At these institutions, multidisciplinary molecular tumor boards are established, providing personalized recommendations for targeted therapies based on individual NGS results. So far, five publications report on mono‐ or oligoinstitutional experience with NGS‐informed personalized therapy : Blumenthal et al. reported the first retrospective cohort from five tertiary hospitals comprising 43 glioma patients (34 GBM), where a NGS panel detected actionable alterations in 95%, leading to targeted treatment in 30% (10/34), but without any treatment response. Byron et al. performed a prospective monocentric trial to evaluate the feasibility of whole exome sequencing‐informed treatment recommendations within 35 days of surgery. Among 16 GBM patients, a recommendation was possible in 94%, 44% (7/16) received a targeted treatment, and one patient (6%) achieved a treatment response. Lazaridis et al. reported a retrospective monocentric cohort of 41 glioma (32 GBM) patients. Following NGS and further methods of genomic profiling, actionable targets were identified in 76% (24/32) and 16 GBM patients receiving targeted treatment achieved an increased PFS (3.8 vs. 2.0 months) and OS (13 vs. 4 months) compared to 16 GBM patients with unmatched empiric treatment. Renovanz et al. reported on their experience from the Center for Personalized Medicine Tübingen, which has an established certified clinical workflow for personalized medicine in the clinical routine for cancer patients without options for trial participation or further registered treatments. This heterogeneous and heavily pretreated cohort included 262 GBM . Following comprehensive molecular profiling, molecularly instructed treatment recommendations were made in 93% (243/262) and 41 GBM patients were treated accordingly, resulting in a PFS2/PFS1 > 1.3 in 36% (13 of 36 evaluable patients). Padovan et al. reported a retrospective, monocentric cohort of 417 GBM patients receiving NGS screening. While actionable targets were identified in 82%, 36 patients (8.6%) received a targeted treatment, of which 20% (7/36) achieved a PFS2/PFS1 > 1.3. The five series are described in more detail in Table . Of note, only two of these studies were prospectively documented , and the number of evaluable patients with targeted therapy per cohort remains low, ranging from < 20 to 36 patients . The largest studies to date also highlight the current problem of precision therapy. As reported by Renovanz et al. , among 262 GBM patients receiving NGS screening, only 41 actually started matched therapy (about 14% of patients tested). While a high percentage of these (88%, n = 36) could be evaluated for efficacy and demonstrate the determination of the authors, the low rate of initiated therapies emphasizes the many obstacles for molecularly matched therapy, such as the rapid deterioration of patients with progressive GBM and the lack of reimbursement by health insurance companies attributable to insufficient GBM‐specific evidence of efficacy . Padovan et al. reported a similar experience, where only 36 (8.6%) of 417 GBM patients receiving NGS screening were able to initiate matched therapy. No treatment response was observed by Blumenthal et al., while Byron et al. report a single patient receiving a potentially successful treatment of olaparib/trametinib/carboplatin with a PFS2/PFS1 ratio of > 3 . The larger studies provide some data on the efficacy of matched therapy in GBM . Lazaridis et al. , also considering CNS drug penetration in the decision‐making process, presented 16 patients receiving matched therapy compared to 16 patients with unmatched therapy (Table ), potentially introducing selection bias. The encouraging PFS results of 3.8 and 2.0 months with versus without matched therapy were mainly driven by the efficacy of BRAF V600E‐directed therapy with dabrafenib/trametinib and c‐Met‐directed therapy with the tyrosine kinase inhibitor (TKI) cabozantinib. While the first observation is in line with the interim results from the ROAR trial discussed in Section , the cabozantinib results warrant critical discussion as the drug targets not only c‐Met but also VEGFR. Inhibition of the VEGF pathway, e.g. with the VEGF‐A antibody bevacizumab, is well known to prolong PFS but not OS . The intermingling of VEGF‐directed therapy with therapy directed at other targets was also present in the study by Renovanz et al. . As mentioned, 14% of GBM patients started matched therapy, and 13/36 evaluable patients (5% of screened patients or 36% of patients receiving treatment and being evaluable) were successfully treated according to a PFS2/PFS1 ratio > 1.3, while the median PFS was 2.3 months . Here, three of the four patients receiving regorafenib based on FMS‐like tyrosine kinase (FLT) or EGFR alterations did benefit (PFS2/PFS1 > 1.3). As discussed above, regorafenib targets several tyrosine kinases, including VEGFR, raising the possibility that the observed PFS benefit might at least partially be attributable to VEGFR targeting. Apart from this, a signal for treatment benefit was only seen for everolimus in tumors with PTEN or PIK3CA alterations (3/6 with PFS2/PFS1 > 1.3), which is in line with results from the mTOR inhibitor temsirolimus trial in newly diagnosed MGMT ‐unmethylated GBM , and for tumors with FGFR fusion transcripts treated with the FGFR inhibitor erdafitinib (3/4 with PFS2/PFS1 > 1.3). Two patients had a positive PFS2/PFS1 signal with the EGFR antibody‐toxin conjugate depatuxizumab mafodotin in the context of an early‐access program, illustrating its PFS prolongation in a previous trial , which did not translate into a longer OS in large phase III trials (Table ). Padovan et al. confirm these observations: among the 36/417 patients receiving targeted treatment, 19% were treated successfully with a PFS2/PFS1 ratio > 1.3, including three objective responses, again mainly driven by dabrafenib/trametinib in BRAF V600E‐altered GBM and erdafitinib in FGFR3‐altered GBM, while the median PFS was 2.1 months. Other publications report the results of NGS screening and/or matched therapy allocation without including a substantial number of GBM patients evaluable for treatment efficacy . Using a methodologically different approach, Luger et al. retrospectively analyzed a cohort of 351 patients treated with off‐label therapy and identified 15 patients with high‐grade glioma (8 GBM) who received matched therapy. This series was dominated by the observation that 3/6 patients treated with BRAF V600E‐directed therapy had disease stability for 5+ months. In summary, these reports highlight that few NGS‐screened GBM patients receive targeted treatment, and even fewer may benefit from it. Advancing GBM precision oncology: beyond tumor cell targets While precision oncology heralds potentially great benefit in GBM, the clinical results achieved with targeted drugs remain underwhelming. There are several areas of ongoing research to overcome this. This includes (a) the identification of treatment targets beyond DNA sequencing, such as multi‐omics‐based exploitation of altered pathways, immunophenotyping, epigenetic profiling, metabolomics, and single‐cell analyses . Another area of current research focuses on (b) optimization of the method of target engagement, e.g. including CAR T‐ or NK‐cells, tumor vaccination, oncolytic viruses, and antibody‐drug conjugates (see for review), despite the low frequency of currently addressable targets . (c) The optimization of (sequential) target verification, e.g. via radiomic and liquid biopsy strategies , could also improve the extent of the therapeutic benefit/treatment response. (d) Understanding and overcoming molecular mechanisms of acquired therapy resistance is key to developing more potent therapeutic modalities , and (e) the optimization of clinical trial conductance could better inform future studies . Here, we focus on discussing the advancement of precision oncology in GBM toward novel treatment targets beyond tumor cell‐intrinsic targets, thus including targets in the tumor‐associated microenvironment (TME). The TME of GBM is increasingly well characterized , and changes in the TME may be crucial for tumor progression . Interactions of tumor cells with vascular structures, immune cells, neurons, glial cells, and each other may provide new targets for therapy (Fig. ). In this context, extracellular vesicles (EV) are notable for their involvement at both a diagnostic and therapeutic level. EVs are membrane‐bound vesicles secreted into the extracellular space that can cross the blood–brain barrier and carry a broad range of cargos, including nucleic acids, lipids, and proteins, together with markers reflecting their biogenesis . GBM‐derived EVs purified from blood or cerebrospinal fluid allow for tumor diagnosis and noninvasive longitudinal sampling for detection of tumor progression, treatment targets, and treatment response . Further, EVs may be taken up by neighboring and distant cells in the TME as well as GBM cells, thus representing both an important means of GBM‐TME communication and a possible therapeutic approach for targeted drug delivery . 6.1 Antiangiogenic therapy The tumor‐vascularization axis has been targeted with the application of the VEGF‐A inhibitor bevacizumab, leading to prolonged PFS (potentially also due to antiedematous effects) but not OS, both in newly diagnosed and progressive GBM . Other anti‐angiogenic drugs, such as cediranib, a VEGFR inhibitor, had a similar PFS benefit in newly diagnosed GBM, but again, there was no OS benefit both in the newly diagnosed and progressive settings . Cilengitide, an integrin inhibitor targeting angiogenesis, did not find its way into clinical application following a PFS‐ and OS‐negative phase III trial . However, antiangiogenic therapy was applied in unselected cohorts without pretesting for the respective targeted angiogenic factors, and thus its potential might be higher in a precision oncology approach. Retrospective analyses aimed to identify molecular subgroups with an OS benefit from bevacizumab . In a biomarker analysis of AVAglio, the ‘proneural’ gene expression subtype was associated with a significant OS advantage (17.1 vs. 12.8 months) in newly diagnosed GBM receiving bevacizumab, which seems counterintuitive as this subtype is associated with lower VEGF expression, and the results could not be confirmed in the GLARIUS trial . In the progressive setting, NF1 mutation was associated with survival benefit from bevacizumab (OS approx. 17 vs. 8 months) in an exploratory biomarker analysis of EORTC‐26101, but these results need further validation . 6.2 Immunological targets The interaction of the immune system with GBM cells may provide further targets for precision oncology. To date, large clinical trials investigating immune checkpoint blockade in GBM have failed , and no predictive markers have been defined for matched therapy besides microsatellite instability (see Section ). Alone or in combination with PD‐1/programmed death‐ligand 1 (PD‐L1) inhibitors, some trials applied cell‐based (e.g. DCVax ) or – more promisingly – multi‐peptide‐based vaccination therapy after EGFRvIII‐directed mono‐peptide vaccination with rindopepimut failed as discussed above . The peptide vaccination approach may be personalized using an individual peptide mix informed by tumor tissue analysis. First results from the phase I GAPVAC‐101 trial, investigating highly individualized vaccinations against an individual selection of unmutated antigens and neoepitopes in 15 patients, document a sustained T‐cell immune response, while meaningful clinical efficacy (e.g. prolongation of PFS and OS) has still not been demonstrated . It remains a major challenge to identify immunomodulatory targets in the microenvironment that can overcome local immunosuppression and further enhance the immune reaction against tumor cells. Further, macrophages and microglia have been shown to interact with tumor cells and may even manipulate them to obtain a more aggressive phenotype . Despite promising preclinical data, first approaches with colony stimulating factor 1 receptor (CSF‐1R)‐targeted inhibition of macrophages failed . Macrophage exclusion from the tumor by inhibition of C‐X‐C motif chemokine receptor 4 (CXCR4) was shown to reduce post‐irradiation tumor revascularization in a small phase I/II trial. Additionally, inhibition of CXCR4 by plerixafor or inhibition of its ligand C‐X‐C motif chemokine ligand 12 (CXCL12; formerly known as stromal cell derived factor‐1, SDF‐1) by NOX‐A12 is being evaluated in several ongoing trials . Macrophages and microglia provide several other markers that may be targeted to enable a stronger and more precise immune reaction in GBM immunotherapy . Further, indirect targeting of immune cells with GBM‐specific stromal protein‐targeted immunostimulatory cytokines represents a novel approach. An antibody‐cytokine conjugate targeting a tumor‐associated fibronectin epitope to enable local distribution of tumor necrosis factor was associated with increasing tumor necrosis and local inflammation in a phase I study, with objective responses in 3/5 progressive GBM patients, and is currently being evaluated in further trials . 6.3 Targeting tumor‐tumor and neuron‐tumor networks The interactions of tumor cells with each other and with neuronal or glial cells offer further opportunities for precision oncology. The rising field of cancer neuroscience has provided a host of landmark publications, showing that GBM form tumor microtube‐based tumor cell networks that confer resistance to radiotherapy and chemotherapy and promote tumor cell invasion by recapitulating developmental neuronal programs . These observations may inform new targets for future therapeutic manipulation, e.g. or the disturbance of hub cells within the syncytium that dominate and organize the tumor cell network , or the disruption of tumor syncytia by gap junction inhibitors – the latter being explored in an ongoing phase I/II trial . Finally, several ways to modulate the neuronal input on tumor cell networks have been found and may be targeted. The synaptic protein neuroligin‐3 (NLGN3) was identified as the leading mitogen mediating neuronal activity‐induced glioma proliferation in patient‐derived xenograft models, and reduction of the release of its soluble form (sNLGN3) by ADAMS10 sheddase inhibition with INCB7839 is explored in a phase I trial (NCT04295759) . Similarly, neuronal activity was shown to mediate glioma invasion and growth via α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazoleproprionic acid receptor (AMPAR)‐mediated synaptic input from neurogliomal glutamatergic synapses in patient‐derived xenograft models. In line with this, targeting the modulation of AMPAR synaptic transmission using the antiepileptic drug perampanel is explored in a phase II trial . To further refine this as a precision therapy approach, predictive markers have yet to be defined. Antiangiogenic therapy The tumor‐vascularization axis has been targeted with the application of the VEGF‐A inhibitor bevacizumab, leading to prolonged PFS (potentially also due to antiedematous effects) but not OS, both in newly diagnosed and progressive GBM . Other anti‐angiogenic drugs, such as cediranib, a VEGFR inhibitor, had a similar PFS benefit in newly diagnosed GBM, but again, there was no OS benefit both in the newly diagnosed and progressive settings . Cilengitide, an integrin inhibitor targeting angiogenesis, did not find its way into clinical application following a PFS‐ and OS‐negative phase III trial . However, antiangiogenic therapy was applied in unselected cohorts without pretesting for the respective targeted angiogenic factors, and thus its potential might be higher in a precision oncology approach. Retrospective analyses aimed to identify molecular subgroups with an OS benefit from bevacizumab . In a biomarker analysis of AVAglio, the ‘proneural’ gene expression subtype was associated with a significant OS advantage (17.1 vs. 12.8 months) in newly diagnosed GBM receiving bevacizumab, which seems counterintuitive as this subtype is associated with lower VEGF expression, and the results could not be confirmed in the GLARIUS trial . In the progressive setting, NF1 mutation was associated with survival benefit from bevacizumab (OS approx. 17 vs. 8 months) in an exploratory biomarker analysis of EORTC‐26101, but these results need further validation . Immunological targets The interaction of the immune system with GBM cells may provide further targets for precision oncology. To date, large clinical trials investigating immune checkpoint blockade in GBM have failed , and no predictive markers have been defined for matched therapy besides microsatellite instability (see Section ). Alone or in combination with PD‐1/programmed death‐ligand 1 (PD‐L1) inhibitors, some trials applied cell‐based (e.g. DCVax ) or – more promisingly – multi‐peptide‐based vaccination therapy after EGFRvIII‐directed mono‐peptide vaccination with rindopepimut failed as discussed above . The peptide vaccination approach may be personalized using an individual peptide mix informed by tumor tissue analysis. First results from the phase I GAPVAC‐101 trial, investigating highly individualized vaccinations against an individual selection of unmutated antigens and neoepitopes in 15 patients, document a sustained T‐cell immune response, while meaningful clinical efficacy (e.g. prolongation of PFS and OS) has still not been demonstrated . It remains a major challenge to identify immunomodulatory targets in the microenvironment that can overcome local immunosuppression and further enhance the immune reaction against tumor cells. Further, macrophages and microglia have been shown to interact with tumor cells and may even manipulate them to obtain a more aggressive phenotype . Despite promising preclinical data, first approaches with colony stimulating factor 1 receptor (CSF‐1R)‐targeted inhibition of macrophages failed . Macrophage exclusion from the tumor by inhibition of C‐X‐C motif chemokine receptor 4 (CXCR4) was shown to reduce post‐irradiation tumor revascularization in a small phase I/II trial. Additionally, inhibition of CXCR4 by plerixafor or inhibition of its ligand C‐X‐C motif chemokine ligand 12 (CXCL12; formerly known as stromal cell derived factor‐1, SDF‐1) by NOX‐A12 is being evaluated in several ongoing trials . Macrophages and microglia provide several other markers that may be targeted to enable a stronger and more precise immune reaction in GBM immunotherapy . Further, indirect targeting of immune cells with GBM‐specific stromal protein‐targeted immunostimulatory cytokines represents a novel approach. An antibody‐cytokine conjugate targeting a tumor‐associated fibronectin epitope to enable local distribution of tumor necrosis factor was associated with increasing tumor necrosis and local inflammation in a phase I study, with objective responses in 3/5 progressive GBM patients, and is currently being evaluated in further trials . Targeting tumor‐tumor and neuron‐tumor networks The interactions of tumor cells with each other and with neuronal or glial cells offer further opportunities for precision oncology. The rising field of cancer neuroscience has provided a host of landmark publications, showing that GBM form tumor microtube‐based tumor cell networks that confer resistance to radiotherapy and chemotherapy and promote tumor cell invasion by recapitulating developmental neuronal programs . These observations may inform new targets for future therapeutic manipulation, e.g. or the disturbance of hub cells within the syncytium that dominate and organize the tumor cell network , or the disruption of tumor syncytia by gap junction inhibitors – the latter being explored in an ongoing phase I/II trial . Finally, several ways to modulate the neuronal input on tumor cell networks have been found and may be targeted. The synaptic protein neuroligin‐3 (NLGN3) was identified as the leading mitogen mediating neuronal activity‐induced glioma proliferation in patient‐derived xenograft models, and reduction of the release of its soluble form (sNLGN3) by ADAMS10 sheddase inhibition with INCB7839 is explored in a phase I trial (NCT04295759) . Similarly, neuronal activity was shown to mediate glioma invasion and growth via α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazoleproprionic acid receptor (AMPAR)‐mediated synaptic input from neurogliomal glutamatergic synapses in patient‐derived xenograft models. In line with this, targeting the modulation of AMPAR synaptic transmission using the antiepileptic drug perampanel is explored in a phase II trial . To further refine this as a precision therapy approach, predictive markers have yet to be defined. Summary and further perspectives Despite first successes with BRAF V600E‐directed dabrafenib/trametinib, and signs of some efficacy in a low percentage of GBM patients receiving molecular‐guided therapies, precision oncology has yet to find broad clinical application with proven efficacy in patients with GBM. Trials investigating targeted drugs in molecularly defined subgroups and treatment allocation based on broad NGS screening need further optimization, e.g. by taking into account the CNS penetration of drugs, more complex prediction models based on combinations of genetic vulnerabilities/interaction of pathways , and new targets beyond the tumor cell. New models of clinical trials are being conducted to allow efficient analysis of new substances and multi‐omics approaches. The results of N2M2, GBM AGILE, and INSIGhT exploring multiple targeted treatments in comparison to a common standard of care will significantly advance the field. Until more efficacy data are available, matched personalized therapy may, with the exceptions mentioned above, be reserved for the experimental treatment of relapsed GBM. UH reports advisory and speaker honoraria from Bayer and speaker honoraria from Medac. The other authors report no conflicts of interest. UH conceptualized and supervised the project. JW and UH researched the literature and wrote the first draft of the manuscript. A‐LP designed the figures. All authors contributed to the writing and editing of the manuscript. |
Gender Disparities in Adverse Events Resulting From Low-Value Practices in Family Practice in Spain: A Retrospective Cohort Study | 359b515d-e6e9-4acf-b81b-f5c1eca2e9de | 11286494 | Family Medicine[mh] | Despite the rising costs in developed Western societies, patient outcomes remain suboptimal , and adverse events continue to pose a significant challenge across all healthcare systems . Due to its role in orchestrating patient flow within the healthcare system, primary care is pivotal in achieving favorable patient outcomes . Although less studied, one of the causes of adverse events in primary care is directly related to recommending, administering, or prescribing healthcare services that are unlikely to benefit patients , which we consider as overuse . The volume of patients subjected to low-value practices (LVPs) in the United States, Canada, Australia, and Sweden reach up to 80%, depending on the type of practice . In Primary Care in Spain , nearly 6 out of 10 adult patients and 4 out of 10 pediatric patients annually receive at least one prescription classified as overuse. Examining only the overage in expenses resulting from unnecessary prescriptions of benzodiazepines, NSAIDs, lipid-lowering agents, paracetamol, and ibuprofen within a single year, reveals an annually total surpassing 290 million euros. This constitutes 2.8% of the entire Spanish pharmaceutical expenditure in 2018 , accounting solely for the cost of the prescribed medications. The continued occurrence of overuse in primary care is frequently linked to various factors, including limited time, constrained access to comprehensive patient data, defensive medical practices, and the approval of prescription decisions either made by healthcare colleagues or requested by patients . Recent studies also highlight differences in the frequency of LVPs between male and female patients . Moreover, the number of adverse events due to overuse has been suggested higher in women . Although it is known that women are negatively affected by a gender bias in the therapeutic effort, and they experience greater delays in diagnosis , the male and female difference has not yet been investigated in relation to overuse, which means that interventions aimed at reducing it do not consider the differential impact on female patients, who could be particularly and negatively affected by its consequences. Therefore, the overarching aim of this research is to assess if there are differences among male and female patients treated by male or female family physicians with regard the occurrence of preventable adverse events due to LVPs in the primary care setting. In this study, we reach out to test the following hypotheses developed based on the results of previous studies within primary care . H 1 . A higher number of LVPs are identified among female patients compared to male patients within similar age groups and reasons for consultation. H 2 . Male and female family physicians are responsible of a similar number of LVPs among their patients. H 3 . A higher number of preventable adverse events related to LVPs are identified among female patients compared to male patients within similar age groups and reasons for consultation. H 4 . Male and female family physicians are involved in a similar number of preventable adverse events related to LVPs among their patients. H 5 . Preventable adverse events stemming from overuse, either due to conditions or symptoms more commonly found in patients of a specific sex or those attributed to gender-related reasons, exhibit similarity. Design A retrospective cohort study in which a random selection of patients attending primary care consultations in Alicante province (Spain) was performed. The STROBE checklist was used as a guide for reporting the study . The study protocol was published first . Primary Care in Spain Spanish primary care is a cornerstone of the country’s healthcare system, offering accessible, and comprehensive healthcare services to individuals who require ongoing medical attention, often due to chronic illnesses. This level of care ensures universal access to quality healthcare for individuals of all ages. Preventive care, early intervention, and continuity of care is provided by multidisciplinary teams of family physicians, pediatricians, nurses, and allied health professionals. Definitions In this study, overuse was defined as continuing to do what should not be done (e.g., ignoring the “Do Not Do” recommendations). The LVPs considered in the study were derived from the Spanish Commitment to Quality initiative list of recommendations , formulated according to the Choosing Wisely campaign’s methodology to mitigate overuse . Adverse event was defined as injury resulting from medical management or a complication, rather than the underlying disease, leading to extended hospitalization and/or disability at discharge from medical care . Gender bias in health refers to unjustifiable differences in treatment between women and men based on scientific evidence. This bias arises from assuming gender differences where there are none or ignoring genuine differences that necessitate a distinct approach according to evidence . Ethics The Research Ethic Board of the Sant Joan Hospital approved the study protocol reference 21/061. It was registered on ClinicalTrials.gov https://clinicaltrials.gov/study/NCT05233852 (NCT05233852). Procedure A group of reviewers (n = 40) was formed and trained in the study LVPs identification and data collection procedures. Training was provided using anonymized records. During the training, all reviewers assessed the same cases, and concordance was measured using Cohen’s kappa coefficient. A score of 0.63 or higher was deemed acceptable, while a score of 0.84 or higher was considered excellent. Training concluded once an excellent level of concordance was reached. The list of reviewers involved in this study is provided in . Each reviewer independently assessed selected medical records and recorded study data. Upon identifying an LVP, the reviewer evaluated potential adverse events and, if present, assessed severity and harm extent using the Woods et al scale, where higher scores indicate greater severity and a stronger relationship between the practice and harm. Events with scores above 3 were classified as adverse events, while those above 4 were attributed to LVPs. A blinded recording system was employed. Data Collection Data were extracted from the primary care electronic medical records database, Abucasis, between 15 March 2023 and 31 August 2023. In Alicante, as well as in the rest of Spain, all the information about a patient is registered in a unique electronic medical record. Data from medical records were collected using an electronic data collection platform, which incorporated a trigger tool to facilitate the identification and recording of adverse events. This tool, previously used in the SOBRINA study , was based on recommendations by Rosenberg et al . The LVPs considered in this study were agreed in a previous study . An online consensus technique involving 33 health professionals from family medicine, cardiology, intensive care, and geriatrics was conducted to reach a consensus on LVPs considering three aspects: 1) if it was still a relatively frequent LVP in primary care; 2) its frequency of application was different between men and women, with a probable association with sex or gender; and 3) if the LVP could cause a severe adverse event in the patient. Panelists marked their level of agreement/disagreement on a scale of 0 (strongly disagree) to 10 (strongly agree). The resulting score was the sum of the three scales. The LVPs that yielded a score of 20 points, or more were retained (consensus criterion) and those scoring under 10 points discarded. Then, a select group of panelists were asked to review the final list of LVPs. Additionally, during a session with experts (clinicians and gender bias in health researchers), there was a debate and consensus reached on whether the differences between men and women that could be observed in these previously selected LVPs should be attributed to the presence of gender inequalities in healthcare. In cases where treatment (or test) is indicated for a condition or symptoms that are more prevalent in patients of a specific sex, it was assumed that the risk of overtreatment (or overuse) in patients of that particular sex is higher than in the other. However, when there is no evidence that the symptoms or prevalence of the condition for which the treatment is provided differ between sex, it was assumed that differences in practice application are due to gender-related reasons. We used a scale ranging from −5, entirely attributable to belief on that are more prevalent in patients of a specific sex, to +5, entirely attributable to gender bias. shows the outcome of this consensus among experts. Sample The proportion of medical records with at least one LVP was expected to be 50% . With an alpha risk of 0.05 and an accuracy of 2.5%, the minimum required sample size was determined to be 1,538 medical records (50% of which were from women). The study sample was stratified by age group and sex, considering the visit frequencies recorded in the National Health System’s primary care information system for 2018. Study participants were divided into three age groups: 18–59 years, 60–74 years, and >75 years, based on reference ages from prior studies . A simple random sampling method with k = 5 was used to select the medical records of patients attended in the past 3 years. Data Analysis Considering the higher frequency of female patients attending primary care consultations (In Spain, 9.6 vs. 5.7 visits per year in 2022 ) the adjusted LVPs and preventable adverse events rates have been calculated to correct for this effect in the interpretation of the data. The chi-square test with Yates correction were used to compare the frequency of LVPs in men and women, and the Cochran-Mantel-Haenszel test to analyze differences in the adjusted rate between the sexes. To analyze the relationship between the presence of an adverse event (dependent variable) and the corresponding independent variables such as age, the number of daily medications, patient’s gender, physician’s gender, and their interaction, a Generalized Linear Mixed Model (GLMM) was used. This model accounts for random effects to cover cases where the same patient is affected by more than one adverse event. Statistical significance for all tests was determined at p < 0.05 (two-tailed). The analyses were conducted using the SPSS statistical software and the RStudio V.1.1.463 programming language. A retrospective cohort study in which a random selection of patients attending primary care consultations in Alicante province (Spain) was performed. The STROBE checklist was used as a guide for reporting the study . The study protocol was published first . Spanish primary care is a cornerstone of the country’s healthcare system, offering accessible, and comprehensive healthcare services to individuals who require ongoing medical attention, often due to chronic illnesses. This level of care ensures universal access to quality healthcare for individuals of all ages. Preventive care, early intervention, and continuity of care is provided by multidisciplinary teams of family physicians, pediatricians, nurses, and allied health professionals. In this study, overuse was defined as continuing to do what should not be done (e.g., ignoring the “Do Not Do” recommendations). The LVPs considered in the study were derived from the Spanish Commitment to Quality initiative list of recommendations , formulated according to the Choosing Wisely campaign’s methodology to mitigate overuse . Adverse event was defined as injury resulting from medical management or a complication, rather than the underlying disease, leading to extended hospitalization and/or disability at discharge from medical care . Gender bias in health refers to unjustifiable differences in treatment between women and men based on scientific evidence. This bias arises from assuming gender differences where there are none or ignoring genuine differences that necessitate a distinct approach according to evidence . The Research Ethic Board of the Sant Joan Hospital approved the study protocol reference 21/061. It was registered on ClinicalTrials.gov https://clinicaltrials.gov/study/NCT05233852 (NCT05233852). A group of reviewers (n = 40) was formed and trained in the study LVPs identification and data collection procedures. Training was provided using anonymized records. During the training, all reviewers assessed the same cases, and concordance was measured using Cohen’s kappa coefficient. A score of 0.63 or higher was deemed acceptable, while a score of 0.84 or higher was considered excellent. Training concluded once an excellent level of concordance was reached. The list of reviewers involved in this study is provided in . Each reviewer independently assessed selected medical records and recorded study data. Upon identifying an LVP, the reviewer evaluated potential adverse events and, if present, assessed severity and harm extent using the Woods et al scale, where higher scores indicate greater severity and a stronger relationship between the practice and harm. Events with scores above 3 were classified as adverse events, while those above 4 were attributed to LVPs. A blinded recording system was employed. Data were extracted from the primary care electronic medical records database, Abucasis, between 15 March 2023 and 31 August 2023. In Alicante, as well as in the rest of Spain, all the information about a patient is registered in a unique electronic medical record. Data from medical records were collected using an electronic data collection platform, which incorporated a trigger tool to facilitate the identification and recording of adverse events. This tool, previously used in the SOBRINA study , was based on recommendations by Rosenberg et al . The LVPs considered in this study were agreed in a previous study . An online consensus technique involving 33 health professionals from family medicine, cardiology, intensive care, and geriatrics was conducted to reach a consensus on LVPs considering three aspects: 1) if it was still a relatively frequent LVP in primary care; 2) its frequency of application was different between men and women, with a probable association with sex or gender; and 3) if the LVP could cause a severe adverse event in the patient. Panelists marked their level of agreement/disagreement on a scale of 0 (strongly disagree) to 10 (strongly agree). The resulting score was the sum of the three scales. The LVPs that yielded a score of 20 points, or more were retained (consensus criterion) and those scoring under 10 points discarded. Then, a select group of panelists were asked to review the final list of LVPs. Additionally, during a session with experts (clinicians and gender bias in health researchers), there was a debate and consensus reached on whether the differences between men and women that could be observed in these previously selected LVPs should be attributed to the presence of gender inequalities in healthcare. In cases where treatment (or test) is indicated for a condition or symptoms that are more prevalent in patients of a specific sex, it was assumed that the risk of overtreatment (or overuse) in patients of that particular sex is higher than in the other. However, when there is no evidence that the symptoms or prevalence of the condition for which the treatment is provided differ between sex, it was assumed that differences in practice application are due to gender-related reasons. We used a scale ranging from −5, entirely attributable to belief on that are more prevalent in patients of a specific sex, to +5, entirely attributable to gender bias. shows the outcome of this consensus among experts. The proportion of medical records with at least one LVP was expected to be 50% . With an alpha risk of 0.05 and an accuracy of 2.5%, the minimum required sample size was determined to be 1,538 medical records (50% of which were from women). The study sample was stratified by age group and sex, considering the visit frequencies recorded in the National Health System’s primary care information system for 2018. Study participants were divided into three age groups: 18–59 years, 60–74 years, and >75 years, based on reference ages from prior studies . A simple random sampling method with k = 5 was used to select the medical records of patients attended in the past 3 years. Considering the higher frequency of female patients attending primary care consultations (In Spain, 9.6 vs. 5.7 visits per year in 2022 ) the adjusted LVPs and preventable adverse events rates have been calculated to correct for this effect in the interpretation of the data. The chi-square test with Yates correction were used to compare the frequency of LVPs in men and women, and the Cochran-Mantel-Haenszel test to analyze differences in the adjusted rate between the sexes. To analyze the relationship between the presence of an adverse event (dependent variable) and the corresponding independent variables such as age, the number of daily medications, patient’s gender, physician’s gender, and their interaction, a Generalized Linear Mixed Model (GLMM) was used. This model accounts for random effects to cover cases where the same patient is affected by more than one adverse event. Statistical significance for all tests was determined at p < 0.05 (two-tailed). The analyses were conducted using the SPSS statistical software and the RStudio V.1.1.463 programming language. In total, 1,538 electronic medical records were reviewed, but after exclusions (due to missing data), a total of 1,516 patients were included, being 911/1,516 (60.1%) female . The mean age of patients attended during the study period was for male 56.4 years (SD 19.4) and female patients 55.2 years old (DT 20.8). They were taking an average of 3.7 medications daily (range 1–25). A total of 245 (68.1%) patients treated by male family physicians were taking less than five drugs per day, while 115 (31.9%) were taking five or more drugs daily. In the case of patients treated by female family physicians, 769 (67.85%) were taking less than five drugs per day and 365 (32.1%) were taking five or more drugs per day. The most frequent main diagnoses in this sample were hypertension, and Type 2 Diabetes. H 1 A higher number of LVPs are identified among female patients. The prevalence of patients suffering LVPs was 465/1,516, 30.7%. A total of 221/605 (36.5%) LVPs occurred in male patients, meanwhile 417/911 (45.7%) LVPs occurred in female patients ( p -value = 0.022). As the patient’s age increased, the frequency of LVPs also increased ( p -value = 0.024). The number of patients who experienced at least one LVP was 465 (male patients 170/605, 28.1% and female patients 295/911, 32.4%). Among 286 patients, two or more LVPs were registered (103/605, 17.0% male patients; 183/911, 20.1% female patients). The data confirm H 1 , with the LVPs considered in this study being more frequent among women than among male patients. H 2 Male and female family physicians are responsible of a similar number of LVPs. A total of 156/360 (43.3%) LVPs were observed in patients treated by male physicians and 482/1,134 (42.5%) in patients treated by female physicians ( p -value = 0.950). Analyzing these LVPs considering both the patient’s sex and the professional’s sex , it was observed that only when the family doctor was female, female patients experienced more LVPs than male patients. The findings suggest rejecting, at least partially, H 2 , as there was a higher frequency of LVPs among female patients treated by female family physicians compared to male patients treated by the same female family physicians. H 3 A higher number of preventable adverse events related to LVPs are identified among female patients. During the review of electronic medical records, a total of 124 adverse events were identified among 105 patients subjected to one or multiple LVPs in the study (124/638, 19.4%), of which 35/221 (15.8%) were experienced by male patients and 89/417 (21.3%) by female patients. A total of 26 (26/105, 24.7%) patients experienced two or more preventable adverse events related to the included LVPs in the study. These occurrences of experiencing more than one adverse event related to LVPs were concentrated in individuals aged 60 or older. Among male patients, six (19.35%) of them experienced more than one adverse event, all of whom were treated by male physicians. Among female patients, 20 (27.03%) of them experienced more than one adverse event, of which 6 were treated by male physicians (30%) and 14 by female physicians (70%) ( p -value = 0.465). The severity tendency of the adverse events was slightly higher in the case of female patients, but the difference was not statistically significant ( p -value = 0.058). The data allow us to accept H 3 because the data trend suggests that female patients experience a higher volume of preventable adverse events related to LVPs than males treated for the same health issue. H 4 Male and female family physicians are involved in a similar number of preventable adverse events related to LVPs. When analyzing the interaction between patient sex and physician sex a higher proportion of patients attended by male physicians experienced an adverse event compared to those attended by female physicians ( p -value<0.000), and in cases where a female physicians attended, female patients experienced more adverse events than male patients ( p -value<0.002) . The severity of adverse events suffered by male and female patients were higher when the patients were attended by male family physicians ( p -value<0.000). Most adverse events were related to medication (99, 79.8%). No differences were identified in the nature of the adverse events suffered by patients when treated by male and female family physicians ( p -value = 0.286). As the patient’s age and the number of daily medications taken by the patient increase, the number of adverse events tends to rise. An interaction effect was observed between the patient’s sex and the family physician’s sex, such that when both the physician and the patient are female, there is a significant increase in the probability of adverse events. However, when the patient is male, being attended by a female physician reduces the probability of experiencing an adverse event. Based on suggestive data indicating that therapeutic decisions made by male or female family practice had a differentiated effect in terms of the occurrence of preventable adverse events related to LVPs among their patients of either sex, H 4 was rejected. H 5 Overuse-related adverse events attributed to sex/gender reasons exhibit similarities in specific conditions. Despite a similar frequency of unnecessary prescriptions or tests for both men and women, whether related to LVPs associated with conditions more prevalent in female patients or influenced by gender-based reasons, a higher number of adverse events occurred in cases linked to LVPs potentially driven by gender bias . Consequently, H 5 was rejected based on the data. The data from this study supports the notion that overutilization poses a risk to patient safety . Additionally, it suggests rejecting the assumption that the frequency of LVPs and the number of preventable adverse events involving male and female family physicians are similar; rather, it supports the idea that women experience a higher number of LVPs and related adverse events. The data suggests an interaction effect between the patient’s and physician’s gender regarding the frequency of both severe and mild adverse events, deserving further attention. This interaction may be specific to the type of LVPs studied in this research. Furthermore, LVPs influenced by gender-based conceptions are more likely to result in unsafe care. The extent and number of LVPs and their economic impact have been studied for years in various countries and healthcare levels . Some recent studies have emerged identifying the impact of LVPs in terms of patient safety, linking LVPs to the occurrence of preventable adverse events . In one of these initial studies on this topic, our group found that female patients experienced more adverse events related to LVPs than males . In this second study, we aimed to delve deeper into this issue that affects women’s health. To address this issue, first, a set of LVPs was identified where these differences between males and females could be more pronounced. Second, a review of a set of medical records of patients of both sexes was conducted to describe the frequency at which male and female patients experienced preventable adverse events related to these LVPs. In this study, women, whose medical records were analyzed, experienced a higher volume of these LVPs during the primary care they received. This data suggests that utilization play a significant role in overutilization. It also corroborates previous observations indicating gender differences that negatively impact the quality of care received by women . This study further delves into analyzing the discrepancy in LVPs frequency between men and women, specifying that when a female patient is treated by a female physician, there is a higher likelihood (up to 7% more) of experiencing one of the LVPs analyzed in this study. These results could be due to family physicians, as suggested in other studies , assuming differences between men and women when there are none. It is not new, the fact that some diseases are more often attributed to men and others to women generating a bias in diagnostic criteria and access to complementary tests or treatments . However, the higher number of adverse events in those cases suspected of gender bias is a novel finding. There is evidence that shows that gender, as a social construct, has a substantial impact on health behaviors, access to and use of health systems, and health system responses . Gender bias can be defined as a systematic error in the social construction of the disease’s history and symptoms, which produces inequitable responses to health problems from the health services, as well as discriminatory responses by professionals . The strategies designed to reduce overutilization could consider these findings and refine their approach, recognizing that female patients have a higher probability of receiving an LVP than male patients. One possible explanation is the higher utilization or healthcare-seeking behavior among women due to a persistent gender bias in our society, where they often take responsibility for family health. Another explanation lies in the recent feminization of the medical profession, which might result in a younger female workforce and therefore, less experience among these female physicians compared to their male counterparts. It could also be attributed to patients exerting more pressure on female physicians than on male physicians to undergo diagnostic tests or specific treatments. This could be influenced by the different status assigned to female professionals, owing to the enduring gender biases , as opposed to their male counterparts. Data collected reveals that nearly a quarter of LVPs ultimately result in a preventable adverse event . In other words, in 2 out of every 10 LVPs, harm is caused by an action on the patient through a treatment that should not have been initiated. Similar to other studies, we have also observed that among older patients, a higher number of preventable adverse events occurred . In this case, the data also suggests that in more severe adverse events, the involvement of male family physicians is higher than that of female physicians. Furthermore, female patients, when treated by female family physicians, exhibited a higher proportion of mild adverse events than male patients. We know that overutilization poses a threat to the survival of healthcare systems. Moreover, its risk to patients is becoming increasingly evident. In the majority of preventable adverse events identified, the severity of the damage was mild. However, nearly two out of ten resulted in severe permanent consequences for the patient. Both in hospitals and primary care, it has been emphasized that LVPs were not as innocuous as previously thought. Indicating, for example, a test when it’s unnecessary opens up possibilities of initiating equally unnecessary treatments, risking the patient and burdening the healthcare system with unnecessary costs, to the detriment of other patients in need of care. Considering the latest data indicating that around 7% of patients in primary care in Spain experience an adverse event in a year, the findings of this study clearly point to overutilization as a risk factor, given that the frequency of adverse events associated with LVPs is nearly 3 times higher than expected. Other studies conducted in various countries suggest rates of adverse events in primary care ranging between 1% and 24% , with the most common frequency being around 6%–7%, and 1.6% considered as severe events . LVPs pose a threat to the sustainability of healthcare systems due to the increased costs they entail . Initiatives implemented to reduce overuse have yielded diverse outcomes . The debate on overutilization and its impact on individuals and systems has expanded, verifying that multicomponent interventions are the most effective in reducing overuse. These interventions, combining various elements, should incorporate information regarding biases based on sex/gender related belief that contribute to women receiving more LVPs, especially when some culminate in adverse events. Implications These findings have implications for the content of programs aimed at raising awareness among professionals about the impact of overuse on health outcomes. Given these data, it is advisable to address these potential differences in outcomes between male and female patients when planning awareness campaigns. Some examples highlight that collaboration between patients, caregivers, and clinicians yields positive outcomes in primary care, and a similar approach could be pursued in this case to reduce overuse and concurrently enhance patient safety . Decision aids aimed at increasing patient safety could consider these results to prioritize situations where differences between men and women are more pronounced. Moreover, in clinical practice, particularly concerning these LVPs, clinicians should consider that an unnecessary indication may have an unexpected and negative impact leading to adverse events. Therefore, when making decisions, they should acknowledge that a low-value indication is not harmless and may negatively affect patient safety. They should assess whether the therapeutic approach is disproportionately affecting female patients compared to male patients, inadvertently impacting their health status. Finally, patient schools (e.g., patient experts) and informal caregiver education could serve as suitable platforms to educate both groups about the risks of LVPS in terms of patient safety. In essence, as patient safety remains a challenge for all primary care professionals , this data suggests initiating discussions about how overuse compromises patient safety. Despite practices that may seem inconsequential, they can result in a suboptimal level of care. These results raise new questions. For instance, to what extent do defensive medicine practices causing overuse differ between male and female professionals, and which patient profiles are more susceptible? Additionally, do decision aids integrated into digital systems reduce disparities in LVPs between male and female patients? Studies on overutilization have identified the frequency of various LVPs in different countries. However, transnational comparisons of these LVPs have not been conducted and could be valuable in determining which strategies are more effective in reducing overuse, considering diverse factors, among male and female patients. Limitations The sample size was calculated for a set of LVPs and not to determine the impact of gender on the outcome variables for each individual LVP. This study did not delve into differentiating whether the found differences were due to sex (biological) or gender (social) issues. Since the medical record system (Abucasis) does not include data on race, ethnic group, or socioeconomic status, these variables could not be considered. The clinical experience of the professionals who attended to the patients whose medical records were reviewed could not be determined since such information is not encoded and accessing it would have compromised the anonymization of the data. The data extraction for professionals was limited to gender. Professionals did not review their own histories, all coding and recording of information relied on the work of the reviewers. These data were collected from a limited number of cases of each LVP. More work is needed to understand the drivers of low-value care on males and females when attended by male and female family physicians. Conclusion The prescriptions and tests considered of low value for the patient, as studied in this research, correspond to common and frequent situations in primary care. They represent a small part of the myriad of conditions addressed at this healthcare level, showcasing only a fraction of the broader reality within primary care settings. Consequently, they serve as a mere sample, underscoring a much larger reality where overuse poses a severe challenge for professionals, patients and healthcare systems. This issue not only jeopardizes patients but also poses a risk, as although the majority of safety incidents are deemed minor and lack permanent consequences, our findings indicate that in some cases, they significantly impact patients’ health. Moreover, these results prompt a deeper reflection and exploration into the influence that gender differences—stemming from both biological and social reasons—might have on overuse and the frequency and nature of associated safety incidents. These findings have implications for the content of programs aimed at raising awareness among professionals about the impact of overuse on health outcomes. Given these data, it is advisable to address these potential differences in outcomes between male and female patients when planning awareness campaigns. Some examples highlight that collaboration between patients, caregivers, and clinicians yields positive outcomes in primary care, and a similar approach could be pursued in this case to reduce overuse and concurrently enhance patient safety . Decision aids aimed at increasing patient safety could consider these results to prioritize situations where differences between men and women are more pronounced. Moreover, in clinical practice, particularly concerning these LVPs, clinicians should consider that an unnecessary indication may have an unexpected and negative impact leading to adverse events. Therefore, when making decisions, they should acknowledge that a low-value indication is not harmless and may negatively affect patient safety. They should assess whether the therapeutic approach is disproportionately affecting female patients compared to male patients, inadvertently impacting their health status. Finally, patient schools (e.g., patient experts) and informal caregiver education could serve as suitable platforms to educate both groups about the risks of LVPS in terms of patient safety. In essence, as patient safety remains a challenge for all primary care professionals , this data suggests initiating discussions about how overuse compromises patient safety. Despite practices that may seem inconsequential, they can result in a suboptimal level of care. These results raise new questions. For instance, to what extent do defensive medicine practices causing overuse differ between male and female professionals, and which patient profiles are more susceptible? Additionally, do decision aids integrated into digital systems reduce disparities in LVPs between male and female patients? Studies on overutilization have identified the frequency of various LVPs in different countries. However, transnational comparisons of these LVPs have not been conducted and could be valuable in determining which strategies are more effective in reducing overuse, considering diverse factors, among male and female patients. The sample size was calculated for a set of LVPs and not to determine the impact of gender on the outcome variables for each individual LVP. This study did not delve into differentiating whether the found differences were due to sex (biological) or gender (social) issues. Since the medical record system (Abucasis) does not include data on race, ethnic group, or socioeconomic status, these variables could not be considered. The clinical experience of the professionals who attended to the patients whose medical records were reviewed could not be determined since such information is not encoded and accessing it would have compromised the anonymization of the data. The data extraction for professionals was limited to gender. Professionals did not review their own histories, all coding and recording of information relied on the work of the reviewers. These data were collected from a limited number of cases of each LVP. More work is needed to understand the drivers of low-value care on males and females when attended by male and female family physicians. The prescriptions and tests considered of low value for the patient, as studied in this research, correspond to common and frequent situations in primary care. They represent a small part of the myriad of conditions addressed at this healthcare level, showcasing only a fraction of the broader reality within primary care settings. Consequently, they serve as a mere sample, underscoring a much larger reality where overuse poses a severe challenge for professionals, patients and healthcare systems. This issue not only jeopardizes patients but also poses a risk, as although the majority of safety incidents are deemed minor and lack permanent consequences, our findings indicate that in some cases, they significantly impact patients’ health. Moreover, these results prompt a deeper reflection and exploration into the influence that gender differences—stemming from both biological and social reasons—might have on overuse and the frequency and nature of associated safety incidents. |
TRPS1, a sensitive marker for different histological and molecular types of breast cancer | 2ed8fbd7-6bc0-4610-948f-a78befde8d1f | 11378484 | Anatomy[mh] | Considering the high incidence and high rate of metastasis of breast cancer, metastatic breast carcinoma is commonly considered when metastatic disease is detected in lymph nodes or organs such as the lung, liver, bone, and brain in females. Immunohistochemical analysis is the most common way to identify breast cancer origin in addition to clinical history and histological features. GATA binding protein 3 (GATA3), gross cystic disease fluid protein 15 (GCDFP-15) and forkhead box transcription Factor C 1 (FOXC1) are relatively sensitive markers for breast cancer. GATA3 and GCDFP-15 are luminal markers, and the overall sensitivities of GATA3 and GCDFP-15 to breast cancer are approximately 70%-90% and 25–60%, respectively . However, the expression of GATA3 and GCDFP-15 is low in triple-negative breast cancer (TNBC) . Moreover, GATA3 and GCDFP-15 are expressed in urothelial carcinoma, salivary ductal carcinoma, skin adnexal tumors, T-cell lymphoma, prostate carcinoma and sweat gland carcinoma . FOXC1 is a reliable marker of TNBC and little expressed in luminal and human epidermal growth factor receptor 2 (HER2)-positive invasive breast carcinomas (IBCs) . Thus, there is a need to identify novel sensitive and specific markers to determine breast origin. Trichorhinophalangeal syndrome type 1 (TRPS1) is a newly discovered marker in the GATA family of zinc finger transcription factors. According to previous studies, TRPS1 is significantly upregulated in breast carcinomas compared with normal breast cells . TRPS1 functions as an essential regulator of the growth and differentiation of normal mammary epithelial cells and may be involved in the development of breast cancer. Recent studies have revealed that TRPS1 is a specific and sensitive marker for breast cancer . TRPS1 is highly expressed in 4 molecular types of breast cancer, especially TNBCs . Moreover, TRPS1 showed no or little expression in some tumor types, including lung adenocarcinoma, pancreatic adenocarcinoma, colon and gastric adenocarcinoma, renal cell carcinoma and melanoma . TRPS1 has been analyzed in some specific types of breast cancer, such as invasive lobular carcinoma, metaplastic breast carcinoma (MBC), neuroendocrine carcinomas, acinic cell carcinomas, cribriform adenoid cystic carcinomas (AdCCs) and secretory carcinoma (SC) . However, the sample sizes of these studies were relatively small. Some studies have revealed that TRPS1 is expressed in androgen receptor (AR)-positive apocrine carcinoma, but the results are conflicting. In addition, the correlation between TRPS1 and AR expression in TNBC has not been well studied. Some studies have reported TRPS1 expression in metastatic IBC. Wang et al. reported that TRPS1 had a high sensitivity of 88% for metastatic TNBC. However, research on TRPS1 expression in the most clinically challenging context of identifying a special metastatic type of breast cancer (BC) is limited. Previous studies have frequently compared TRPS1 with GATA3 expression in breast cancers ; however, few studies have compared the expression of TRPS1 and FOXC1, especially in TNBC. The main objective of the current study was to explore TRPS1 expression in special types of breast carcinoma (primary and metastatic). The secondary aims of the study included to assess the correlation between TRPS1 and AR expression in TNBC, and the expression of TRPS1, GATA3, GCDFP-15 and FOXC1 in the immunohistochemistry (IHC)—based molecular types of IBC. To the best of our knowledge, this is the largest reported series of TRPS1 expression in China. Study design and case selection Our study was divided into three parts: (1) evaluating TRPS1 in 801 patients with special types of BC; (2) analysing the correlation between TRPS1 and AR in 969 patients with TNBC (most of them were IBC of no special type); and (3) comparing the expression of TRPS1, GATA3, GCDFP-15 and FOXC1 in 1975 patients with different molecular types of BC (most of them were luminal breast cancers). All samples were collected from 2021–2023 in Fudan University Shanghai Cancer Center.. Tissue samples were obtained from surgical resection. The IHC was performed on full sections. All the pathological sections were reviewed, and the histological diagnosis was made according to the 5th edition of the World Health Organization (WHO) classification. All special types of breast carcinoma were composed of > 90% special carcinoma components according to WHO classification. Immunohistochemistry The details of the IHC antibodies used are shown in Table . All staining was performed with a Ventana BenchMark Ultra autostainer (Ventana Medical System Inc., Roche, Tucson, Arizona). According to the BenchMark ULTRA advanced staining system operator guide, the staining process was performed by applying the appropriate reagent, monitoring the incubation time, and rinsing slides between reagents. Only nuclear staining was considered to indicate positive TRPS1, GATA3, FOXC1 and AR expression, whereas cytoplasmic staining indicated positive GCDFP-15 expression. The expression of these proteins was divided by the percentage of positive cells (negative, < 1%; low positive, 1–10%; intermediate positive, 11–50%; high positive, 51–100%). Patients with intermediate to high positive expression were regarded as positive cases, and those with low positive expression were grouped as negative expression. The immunohistochemical surrogate type of all included IBCs was determined using the following definitions according to the 2013 St Gallen Consensus: luminal A subtype (estrogen receptor (ER) + , progesterone receptor (PR) + ≥ 20%), HER2-, and Ki-67 < 20%), luminal B subtype (ER + , HER2-, Ki-67 > 20% or PR-/ + ; ER + , HER2 + with any PR or Ki-67 index), HER2 + subtype (ER-,PR-,HER2 +) and triple-negative subtype (ER-, PR- and HER2-). According to the 2020 American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines , nuclear ER/PR staining in 1% or more of samples was defined as ER/PR-positive. HER2 status was assessed as proposed in the 2023 ASCO/CAP guidelines . HER2 was considered positive if the patient had an IHC score of 3 + or HER2 gene amplification. Ki-67 interpretation standards referred to the 2011 recommendations from the International Ki-67 in Breast Cancer Working Group . Ki-67 values less than 20% were considered low; otherwise, they were considered high. Statistical analysis SPSS software version 19 was used to perform all the statistical analyses. The data were compared between different subgroups, and the correlation between the expression of AR and that of TRPS1 was analyzed by using χ2 analysis or Fisher’s exact test. The level of significance was set at 0.05. A p value < 0.05 was considered to indicate statistical significance. Our study was divided into three parts: (1) evaluating TRPS1 in 801 patients with special types of BC; (2) analysing the correlation between TRPS1 and AR in 969 patients with TNBC (most of them were IBC of no special type); and (3) comparing the expression of TRPS1, GATA3, GCDFP-15 and FOXC1 in 1975 patients with different molecular types of BC (most of them were luminal breast cancers). All samples were collected from 2021–2023 in Fudan University Shanghai Cancer Center.. Tissue samples were obtained from surgical resection. The IHC was performed on full sections. All the pathological sections were reviewed, and the histological diagnosis was made according to the 5th edition of the World Health Organization (WHO) classification. All special types of breast carcinoma were composed of > 90% special carcinoma components according to WHO classification. The details of the IHC antibodies used are shown in Table . All staining was performed with a Ventana BenchMark Ultra autostainer (Ventana Medical System Inc., Roche, Tucson, Arizona). According to the BenchMark ULTRA advanced staining system operator guide, the staining process was performed by applying the appropriate reagent, monitoring the incubation time, and rinsing slides between reagents. Only nuclear staining was considered to indicate positive TRPS1, GATA3, FOXC1 and AR expression, whereas cytoplasmic staining indicated positive GCDFP-15 expression. The expression of these proteins was divided by the percentage of positive cells (negative, < 1%; low positive, 1–10%; intermediate positive, 11–50%; high positive, 51–100%). Patients with intermediate to high positive expression were regarded as positive cases, and those with low positive expression were grouped as negative expression. The immunohistochemical surrogate type of all included IBCs was determined using the following definitions according to the 2013 St Gallen Consensus: luminal A subtype (estrogen receptor (ER) + , progesterone receptor (PR) + ≥ 20%), HER2-, and Ki-67 < 20%), luminal B subtype (ER + , HER2-, Ki-67 > 20% or PR-/ + ; ER + , HER2 + with any PR or Ki-67 index), HER2 + subtype (ER-,PR-,HER2 +) and triple-negative subtype (ER-, PR- and HER2-). According to the 2020 American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines , nuclear ER/PR staining in 1% or more of samples was defined as ER/PR-positive. HER2 status was assessed as proposed in the 2023 ASCO/CAP guidelines . HER2 was considered positive if the patient had an IHC score of 3 + or HER2 gene amplification. Ki-67 interpretation standards referred to the 2011 recommendations from the International Ki-67 in Breast Cancer Working Group . Ki-67 values less than 20% were considered low; otherwise, they were considered high. SPSS software version 19 was used to perform all the statistical analyses. The data were compared between different subgroups, and the correlation between the expression of AR and that of TRPS1 was analyzed by using χ2 analysis or Fisher’s exact test. The level of significance was set at 0.05. A p value < 0.05 was considered to indicate statistical significance. Expression of TRPS1 in special types of breast carcinoma A total of 801 special types of breast cancers were stained with TRPS1 and GATA3. TRPS1 and GATA3 were positive in 100% (63/63) and 100% (63/63) of mucinous carcinoma, 100% (7/7) and 0% (0/3) adenoid cystic carcinomas (4 classic adenoid cystic carcinomas and 3 solid-basaloid adenoid cystic carcinomas), 100% (4/4) and 100% (4/4) tubular carcinomas, 100% (2/2) and 100% (2/2) secretory carcinomas, 99.59% (243/244) and 100% (244/244) invasive lobular carcinomas, 99.26% (267/269) and 98.51% (265/269) invasive micropapillary carcinomas, 97.44% (38/39) and 100% (34/34) ER-positive neuroendocrine tumors. TRPS1 and GATA3 were negative in all triple-negative neuroendocrine carcinomas (0/7). The above data are shown in Table and Fig. . Among all subtypes of MBC, 94.44% (34/36) were positive for TRPS1, and 41.18% (14/34) were positive for GATA3 (Table ). Among the 13 squamous cell carcinomas, 84.62% (11/13) were positive for TRPS1, and 46.15% (6/13) were positive for GATA3. In 1 patient with fibromatosis-like MBC, TRPS1 was weakly positive. Among the metaplastic carcinoma with mesenchymal differentiation (MC-MD), 100% (22/22) were positive for TRPS1, whereas 38.10% (8/21) were positive for GATA3. TRPS1 and GATA3 expression was also analyzed by IHC in 102 patients with apocrine carcinomas (AR positive and ER-negative), including 77 patients with a triple-negative phenotype and 25 patients with a HER2-positive phenotype (Table and Fig. ). Among the apocrine carcinomas, 63.73% (65/102) were positive for TRPS1, and 86.27% (88/102) were positive for GATA3. TRPS1 expression was positive in 58.44% (45/77) of the triple-negative phenotype and 80% (20/25) of the HER2-positive phenotype. GATA3 was positive in 84.42% (65/77) of the triple-negative phenotype and 92% (23/25) of the HER2-positive phenotype. A total of 28 patients had metastatic special types of BC, including 14 cases of invasive lobular carcinoma, 7 cases of neuroendocrine tumors of the breast (including 6 ER-positive neuroendocrine tumors and 1 triple-negative neuroendocrine carcinoma), 6 cases of MC-MD, and 1 case of invasive micropapillary carcinoma. These tumors generally metastasize to organs such as the lung, liver, bone, and brain (Table ). TRPS1 and GATA3 were positive in 100% (1/1) and 100% (1/1) of invasive micropapillary carcinomas, 100% (6/6) and 100% (6/6) ER-positive neuroendocrine tumors of the breast, 100% (6/6) and 20% (1/5) MC-MD, 92.86% (13/14) and 100% (11/11) invasive lobular carcinomas. TRPS1 and GATA3 were negative in 1 patient with triple-negative neuroendocrine carcinoma of the breast. The above data are shown in Table . Correlation of TRPS1 and AR expression in TNBC TRPS1 and AR expression was analyzed by IHC in 969 patients diagnosed with TNBC (Table ). We found that 90.40% were positive for TRPS1, and 42.41% were positive for AR. A significant inverse correlation between TRPS1 and AR expression was shown in TNBC ( p < 0.001). A total of 538 patients (55.52%) were TRPS1 positive and AR negative, and 73 patients (7.53%) were AR positive and TRPS1 negative. Expression of TRPS1, GATA3, GCDFP-15 and FOXC1 in 4 molecular types of breast cancer Four markers (TRPS1, GATA3, GCDFP-15, and FOXC1) were analyzed in a total of 1975 patients with breast cancer, including 607 patients with luminal A breast cancer, 708 patients with luminal B breast cancer, 296 patients with HER2 + breast cancer and 364 patients with triple-negative breast cancer. Our results revealed no difference in the percentage of luminal A breast cancers positive for TRPS1 or GATA3 (99.84% vs. 99.84%, respectively) and luminal B breast cancer (99.15% vs. 99.58%). However, differences in the expression of TRPS1 and GATA3 were most prominent in TNBC (93.13% vs. 53.57%), followed by HER2 + breast cancer (98.99% vs. 87.16%). We also tested GCDFP-15 and FOXC1 expression in 4 molecular types of breast cancer. The results demonstrated that TRPS1 had a significantly greater positivity rate than GCDFP-15 and FOXC1 in different molecular types of breast cancer ( p < 0.001) (Fig. and Table ). Notably, FOXC1 had a greater positivity rate (71.98%) than GATA3 (53.57%) and GCDFP-15 (32.14%) and a lower rate than TRPS1 (93.13%) in TNBC. A total of 801 special types of breast cancers were stained with TRPS1 and GATA3. TRPS1 and GATA3 were positive in 100% (63/63) and 100% (63/63) of mucinous carcinoma, 100% (7/7) and 0% (0/3) adenoid cystic carcinomas (4 classic adenoid cystic carcinomas and 3 solid-basaloid adenoid cystic carcinomas), 100% (4/4) and 100% (4/4) tubular carcinomas, 100% (2/2) and 100% (2/2) secretory carcinomas, 99.59% (243/244) and 100% (244/244) invasive lobular carcinomas, 99.26% (267/269) and 98.51% (265/269) invasive micropapillary carcinomas, 97.44% (38/39) and 100% (34/34) ER-positive neuroendocrine tumors. TRPS1 and GATA3 were negative in all triple-negative neuroendocrine carcinomas (0/7). The above data are shown in Table and Fig. . Among all subtypes of MBC, 94.44% (34/36) were positive for TRPS1, and 41.18% (14/34) were positive for GATA3 (Table ). Among the 13 squamous cell carcinomas, 84.62% (11/13) were positive for TRPS1, and 46.15% (6/13) were positive for GATA3. In 1 patient with fibromatosis-like MBC, TRPS1 was weakly positive. Among the metaplastic carcinoma with mesenchymal differentiation (MC-MD), 100% (22/22) were positive for TRPS1, whereas 38.10% (8/21) were positive for GATA3. TRPS1 and GATA3 expression was also analyzed by IHC in 102 patients with apocrine carcinomas (AR positive and ER-negative), including 77 patients with a triple-negative phenotype and 25 patients with a HER2-positive phenotype (Table and Fig. ). Among the apocrine carcinomas, 63.73% (65/102) were positive for TRPS1, and 86.27% (88/102) were positive for GATA3. TRPS1 expression was positive in 58.44% (45/77) of the triple-negative phenotype and 80% (20/25) of the HER2-positive phenotype. GATA3 was positive in 84.42% (65/77) of the triple-negative phenotype and 92% (23/25) of the HER2-positive phenotype. A total of 28 patients had metastatic special types of BC, including 14 cases of invasive lobular carcinoma, 7 cases of neuroendocrine tumors of the breast (including 6 ER-positive neuroendocrine tumors and 1 triple-negative neuroendocrine carcinoma), 6 cases of MC-MD, and 1 case of invasive micropapillary carcinoma. These tumors generally metastasize to organs such as the lung, liver, bone, and brain (Table ). TRPS1 and GATA3 were positive in 100% (1/1) and 100% (1/1) of invasive micropapillary carcinomas, 100% (6/6) and 100% (6/6) ER-positive neuroendocrine tumors of the breast, 100% (6/6) and 20% (1/5) MC-MD, 92.86% (13/14) and 100% (11/11) invasive lobular carcinomas. TRPS1 and GATA3 were negative in 1 patient with triple-negative neuroendocrine carcinoma of the breast. The above data are shown in Table . TRPS1 and AR expression was analyzed by IHC in 969 patients diagnosed with TNBC (Table ). We found that 90.40% were positive for TRPS1, and 42.41% were positive for AR. A significant inverse correlation between TRPS1 and AR expression was shown in TNBC ( p < 0.001). A total of 538 patients (55.52%) were TRPS1 positive and AR negative, and 73 patients (7.53%) were AR positive and TRPS1 negative. Four markers (TRPS1, GATA3, GCDFP-15, and FOXC1) were analyzed in a total of 1975 patients with breast cancer, including 607 patients with luminal A breast cancer, 708 patients with luminal B breast cancer, 296 patients with HER2 + breast cancer and 364 patients with triple-negative breast cancer. Our results revealed no difference in the percentage of luminal A breast cancers positive for TRPS1 or GATA3 (99.84% vs. 99.84%, respectively) and luminal B breast cancer (99.15% vs. 99.58%). However, differences in the expression of TRPS1 and GATA3 were most prominent in TNBC (93.13% vs. 53.57%), followed by HER2 + breast cancer (98.99% vs. 87.16%). We also tested GCDFP-15 and FOXC1 expression in 4 molecular types of breast cancer. The results demonstrated that TRPS1 had a significantly greater positivity rate than GCDFP-15 and FOXC1 in different molecular types of breast cancer ( p < 0.001) (Fig. and Table ). Notably, FOXC1 had a greater positivity rate (71.98%) than GATA3 (53.57%) and GCDFP-15 (32.14%) and a lower rate than TRPS1 (93.13%) in TNBC. Special types of BC, representing 25% of all breast cancers, encompass a collection of different diseases characterized by different biological and pathological features, clinical presentations, responses to treatments, clinical behaviors, and outcomes . We investigated TRPS1 expression in special types of BC, as mentioned in a few publications. TRPS1 is considered a “master controller” of both luminal and basal differentiation, while GATA3 is a regulator and indicator of luminal differentiation. As predicted, TRPS1 and GATA3 were highly expressed in some special types of BCs, including mucinous carcinomas, invasive micropapillary carcinomas, tubular carcinomas and invasive lobular carcinomas, ER-positive neuroendocrine tumors and secretory carcinomas. In addition, TRPS1 was more highly expressed than GATA3 in AdCCs, MBCs and special metastatic types of BC. In triple-negative neuroendocrine carcinomas, TRPS1 and GATA3 were not expressed. In apocrine carcinomas, TRPS1 expression was lower than GATA3 expression. According to the 5th WHO classification, breast neuroendocrine neoplasms are categorized as neuroendocrine tumors (mostly ER-positive), small cell neuroendocrine carcinomas or large cell neuroendocrine carcinomas. Our data demonstrated that TRPS1 and GATA3 expression was almost positive in ER-positive neuroendocrine tumors. However, triple-negative neuroendocrine carcinomas showed negative expression of TRPS1 and GATA3, similar to the findings of a previous study . Previous studies have demonstrated that fibromatosis-like MBC is not sensitive to TRPS1 . In our study, we found that TRPS1 was weakly positive in one patient with fibromatosis-like MBC. A larger sample of patients with fibromatosis-like MBC is needed to confirm the TRPS1 expression pattern. We also investigated TRPS1 expression in other MBCs, such as MC-MD and squamous cell carcinoma. All of these highly expressed TRPS1. Of the 7 patients with AdCC in our study, 3 had solid-basaloid adenoid cystic carcinoma, and the rest had classic adenoid cystic carcinoma. Our data revealed that TRPS1 expression was intermediate or highly positive in 100% of AdCCs. However, few studies have reported the opposite result of TRPS1 staining (2/5, 40%) in AdCC . Another study showed that TRPS1 was positive in only 50% (3/6) classic AdCCs . The main reason for this inconsistency is the small number of patients included in the previous study. TRPS1 was highly positive in SC, similar to that reported by others . A previous study showed that TNBC with apocrine differentiation was predominantly negative for TRPS1 expression, while 82% of HER2 + IBC with apocrine differentiation exhibited negative TRPS1 expression . Another study showed that in apocrine carcinomas, TRPS1 was negative in 4 patients (80%) and weakly positive in 1 patient (20%) . Schwartz reported that TRPS1 was weakly expressed in a minority of triple-negative apocrine carcinoma tumors (3/14, 21%) . These studies revealed low TRPS1 expression in AR-positive apocrine carcinoma. However, most of these studies included a relatively small number of apocrine carcinomas. In this study, we investigated TRPS1 expression in a relatively large sample of 102 patients with apocrine carcinoma. A total of 63.73% of all patients were positive for TRPS1. In 77 patients with triple-negative apocrine carcinomas, 45 patients (58.45%) were positive for TRPS1. Among the 25 patients with HER2-positive apocrine carcinomas, 20 patients (80%) were positive for TRPS1. Relatively greater TRPS1 expression in apocrine carcinomas (especially those with a HER2-positive phenotype) was observed in our study than in previous studies. This could relate to some reasons, such as using of different antibody , different diagnostic thresholds for rendering a diagnosis of apocrine carcinoma. Our study included a relative high number of pure apocrine carcinomas All apocrine carcinomas included in our study were composed of > 90% apocrine differentiation components. Although the expression of TRPS1 in apocrine carcinomas in our study was greater than that in previous studies, it was still lower than that in other histological types, and the percentage of TRPS1-positive cells was lower than that of GATA3-positive cells in apocrine carcinoma. We also explored the correlation between TRPS1 and AR in TNBC. Our study showed that TRPS1 and AR expression was inversely correlated in TNBC. TRPS1 has been found to be a highly sensitive marker for all breast cancer molecular types , consistent with our data. Our data showed an overall sensitivity of 93.13% (339/364) for TRPS1 expression in TNBCs. Parkinson et al. reported that TRPS1 was highly expressed in TNBCs (97.4%). Ai et al. reported that 86% of TNBCs were positive for TRPS1. These studies have provided further evidence suggesting that TRPS1 is highly important in TNBC. We evaluated TRPS1 expression in combination with other breast-specific markers (GATA3, GCDFP-15 and FOXC1) in a large cohort of breast cancer patients. FOXC1 is a reliable marker for TNBC. In addition, FOXC1 can serve as an additional diagnostic tool for triple-negative phenotypes and subclassifications in TNBC . Li reported that FOXC1 was positive in 77.84% of patients with TNBC . Our study revealed that TRPS1 was more strongly expressed than GATA3, GCDFP-15 and FOXC1 in TNBC, and FOXC1 was more strongly expressed than GATA3 and GCDFP-15 in TNBC. However, some studies have raised concerns about the specificity of TRPS1. It has been reported that some prostate cancer and muscle invasive bladder cancers show significant staining . Another report elucidated that TRPS1 was positive in non-breast tumors, such as soft tissue tumors, salivary gland tumors, squamous cell carcinomas, and gynecological cancers . Additional comprehensive studies are needed to elucidate the true specificity of TRPS1 IHC staining for many tumor types before it is widely adopted in clinical practice. The combination of TRPS1 and FOXC1 could be recognized as a reliable diagnostic panel for identifying TNBC in clinical practice. In conclusion, our study demonstrated that TRPS1 is a highly sensitive marker for most special types of breast carcinoma. TRPS1 was positive in 63.73% of apocrine carcinomas. TRPS1 and AR expression was inversely correlated in TNBC. TRPS1 was more highly expressed in 4 molecular types of breast cancer in Chinese patients. In TNBC, TRPS1 had a greater positivity rate than GATA3, GCDFP-15 and FOXC1. |
Stage-specific expression of Toll-like receptors in the seminiferous epithelium of mouse testis | 5b60d5e7-4440-498e-a22a-a286b307ee55 | 11364606 | Anatomy[mh] | Toll-like receptors (TLRs) that recognize pathogen-associated molecular patterns (PAMPs) are the best characterized pattern recognition receptors (PRRs) (Janeway and Medzhitov ; Roach et al. ; O’Neill and Bowie ; Hedger ). Binding of various ligands to specific TLRs triggers an innate immune response (Kawai and Akira ; Ward ), a prerequisite for the killing and clearance of various pathogens (Alexopoulou et al. ). Although 28 types of TLRs (TLR1 to TLR28) have been identified in vertebrates (Nie et al. ), mammals have only 13 TLRs (TLR1 − 13) (Akira et al. ; Roach et al. ; Takeda and Akira ). Of these, TLR-10 is a nonfunctional pseudogene in mice (Hasan et al. ; Lee et al. ). Although TLRs are mainly expressed by immune cells such as the dendritic cells, monocytes or macrophages (Underhill ), they are also expressed in various nonimmune tissues and organs (Cudicini et al. ; Cario and Podolsky ; Akhtar et al. ; Melmed et al. ; Smith et al. ; Schaefer et al. ; Zhang et al. ; Lauw et al. ; Girling and Hedger ; Palladino et al. , ; Nagaosa et al. ), including the testis tissue (Palladino et al. ; Bhushan et al. ; Wu et al. ; ; Shang et al. ; Wang et al. ; Chen et al. ; Saeidi et al. ; Nejsum et al. ; Sun et al. ; Özbek et al. ; Öztop et al. ). It has been suggested that TLRs might be involved in the regulation of various testicular functions (Hedger ) including spermatogenesis (Girling and Hedger ). Spermatogenesis, viz. development of functional spermatozoa, is a highly complex and developmentally regulated process initiated at the puberty, involving continuous and serial events of cellular proliferation and differentiation of germ cells within the seminiferous epithelium (Kimmins et al. ). This, what is referred to as the cycle of the seminiferous epithelium, occurs in twelve sequential stages (stages I–XII) in the mouse based on the periodic acid Schiff (PAS)–hematoxylin staining (Oakberg ). At stages I–VII, acrosomic granules occur and acrosome spreads over the periphery of the nucleus. At stage VIII, the acrosome moves away from the nucleus and approaches the surface of the cytoplasm. At this stage, step 16 elongated spermatids are delivered to the lumen through spermiation. While two generations of spermatids (round and elongated; steps 1–16) are present within the seminiferous epithelium at the first eight stages, only elongated spermatids (steps 9–12) are seen at stages IX–XII. At stages IX–XII, elongated spermatids are defined by their morphology and condensation of the chromatin (Oakberg ; Meistrich ; Hess and Renato de Franca ; Ahmed and de Rooij ; Meistrich and Hess ). There are several studies demonstrating that TLRs are expressed by germ cells both in the mouse (Wang et al. ; Chen et al. ; Nejsum et al. ; Sun et al. ) and the rat (Bhushan et al. ; Özbek et al. ; Öztop et al. ). Nevertheless, in the mouse, the expression pattern of TLRs by specific populations of germ cells has only been demonstrated for TLR-3, TLR-9, and TLR-11. Furthermore, spatiotemporal expression of TLRs in relation to the cycle of the seminiferous epithelium remains largely unknown. With these in mind, we examined in the present study the expression of all functional TLRs in the adult mouse testis in sequential sections stained with specific antibodies and PAS–hematoxylin in an attempt to reveal the expression of each TLR by germ cells throughout the cycle of the seminiferous epithelium. While confirming previous observations for the expression of TLR-11 and TLR-3, the present study further reveals a distinct and stage-specific pattern of expression for all functional TLR in the mouse testis. Animals Aguti F2 mice (C57BL/6× BALBc) were maintained on a 14 h light:10 h dark photoperiod (light on at 5 am) with free access to food and water. The experimental protocol was approved by the institutional Animal Ethics Committee of Adnan Menderes University, Aydin, Turkey (protocol no: 64583101/2019/121). Testis tissue samples of males were used to examine the expression of TLRs. Since all TLR antibodies used in the current study have been recently tested to show expression patterns in mouse lung tissues prepared in 4% paraformaldehyde/phosphate-buffered saline (PBS) fixation, testis tissues were fixed in 4% paraformaldehyde/PBS (pH 7.4) at 4 °C for 24 h instead of Bouin’s fixation. All tissue samples were dehydrated through a graded series of ethanol and were embedded in Paraplast X-TRA (Leica, Germany). Periodic acid Schiff (PAS) and immunohistochemical staining were performed on sequential sections taken at 100 µm intervals. PAS-hematoxylin staining and immunohistochemistry To determine the stage of the seminiferous epithelium, periodic acid Schiff (PAS) staining was used as described previously (Ahmed and de Rooij ). Thin (5 μm) tissue sections were deparaffinized and incubated in 1% periodic acid for 30 min at room temperature. Sections were washed in running water for 10 min and were incubated for 45 min in Schiff’s reagent. Following one more wash step in running water, sections were rinsed in distilled water and were counterstained with Mayer’s hematoxylin for 3 min. Images were captured using an Olympus BX51 microscope equipped with an Olympus DP70 camera and DP controller software (Olympus, Ver. 3.1.1.267). Immunohistochemistry was used to detect germ cells expressing TLR-1–13 (except for TLR-10) as described previously (Doğan et al. ; ). Histostain Plus Broad-Spectrum kit (Invitrogen) was used for the detection of TLRs. Working conditions of all TLR antibodies were previously optimized in our previous study in which specific protein bands of all TLRs were also demonstrated in the mouse testis tissue by western blotting method (Doğan et al. ). Anti-TLR-1 (B-23, Sc-130896, Santa Cruz Biotechnology, 1/50), anti-TLR 2 (NB100-56720, Novus, 1/50), anti-TLR-3 (NB100-56571, Novus, 1/50), anti-TLR-4 (NB100- 56,566, Novus, 1/50), anti-TLR-5 (H-127, Sc-10742, Santa Cruz Biotechnology, 1/50), anti-TLR-6 (NBP1-54,336, Novus, 1/50), anti-TLR-7 (NB100-56682, Novus, 1/50), anti-TLR-8 (NBP2- 24,917, Novus, 1/50), anti-TLR-9 (NBP2-24,729, Novus, 1/50), anti-TLR-11 (NBP1-77,204, Novus, 1/50), anti-TLR-12 (NBP2-24,833, Novus, 1/50), and anti-TLR-13 (NBP2-24,539, Novus, 1/50) were used as primary antibodies. TLRs were detected using 3,3′-diaminobenzidine tetrahydrochloride solution (DAB; 3 mg/mL in Tris–HCl, pH 7.6, with 3% H 2 O 2 ). Sections treated in an identical manner except for the use of TBS (pH 7.6) instead of a primary antibody were used as negative controls. Mayer’s hematoxylin was used for counter-staining. Images were captured using an Olympus BX51 microscope equipped with an Olympus DP70 camera and DP controller software (Olympus, Ver. 3.1.1.267). Aguti F2 mice (C57BL/6× BALBc) were maintained on a 14 h light:10 h dark photoperiod (light on at 5 am) with free access to food and water. The experimental protocol was approved by the institutional Animal Ethics Committee of Adnan Menderes University, Aydin, Turkey (protocol no: 64583101/2019/121). Testis tissue samples of males were used to examine the expression of TLRs. Since all TLR antibodies used in the current study have been recently tested to show expression patterns in mouse lung tissues prepared in 4% paraformaldehyde/phosphate-buffered saline (PBS) fixation, testis tissues were fixed in 4% paraformaldehyde/PBS (pH 7.4) at 4 °C for 24 h instead of Bouin’s fixation. All tissue samples were dehydrated through a graded series of ethanol and were embedded in Paraplast X-TRA (Leica, Germany). Periodic acid Schiff (PAS) and immunohistochemical staining were performed on sequential sections taken at 100 µm intervals. To determine the stage of the seminiferous epithelium, periodic acid Schiff (PAS) staining was used as described previously (Ahmed and de Rooij ). Thin (5 μm) tissue sections were deparaffinized and incubated in 1% periodic acid for 30 min at room temperature. Sections were washed in running water for 10 min and were incubated for 45 min in Schiff’s reagent. Following one more wash step in running water, sections were rinsed in distilled water and were counterstained with Mayer’s hematoxylin for 3 min. Images were captured using an Olympus BX51 microscope equipped with an Olympus DP70 camera and DP controller software (Olympus, Ver. 3.1.1.267). Immunohistochemistry was used to detect germ cells expressing TLR-1–13 (except for TLR-10) as described previously (Doğan et al. ; ). Histostain Plus Broad-Spectrum kit (Invitrogen) was used for the detection of TLRs. Working conditions of all TLR antibodies were previously optimized in our previous study in which specific protein bands of all TLRs were also demonstrated in the mouse testis tissue by western blotting method (Doğan et al. ). Anti-TLR-1 (B-23, Sc-130896, Santa Cruz Biotechnology, 1/50), anti-TLR 2 (NB100-56720, Novus, 1/50), anti-TLR-3 (NB100-56571, Novus, 1/50), anti-TLR-4 (NB100- 56,566, Novus, 1/50), anti-TLR-5 (H-127, Sc-10742, Santa Cruz Biotechnology, 1/50), anti-TLR-6 (NBP1-54,336, Novus, 1/50), anti-TLR-7 (NB100-56682, Novus, 1/50), anti-TLR-8 (NBP2- 24,917, Novus, 1/50), anti-TLR-9 (NBP2-24,729, Novus, 1/50), anti-TLR-11 (NBP1-77,204, Novus, 1/50), anti-TLR-12 (NBP2-24,833, Novus, 1/50), and anti-TLR-13 (NBP2-24,539, Novus, 1/50) were used as primary antibodies. TLRs were detected using 3,3′-diaminobenzidine tetrahydrochloride solution (DAB; 3 mg/mL in Tris–HCl, pH 7.6, with 3% H 2 O 2 ). Sections treated in an identical manner except for the use of TBS (pH 7.6) instead of a primary antibody were used as negative controls. Mayer’s hematoxylin was used for counter-staining. Images were captured using an Olympus BX51 microscope equipped with an Olympus DP70 camera and DP controller software (Olympus, Ver. 3.1.1.267). In an attempt to demonstrate the expression pattern of all functional TLRs throughout the cycle of the seminiferous epithelium, we performed PAS staining for staging of the seminiferous epithelium (Supplementary Fig. ) along with immunohistochemistry on sequential sections (Supplementary Figs. – ). Microscopic evaluation of the sections at low magnification (40×) reveals the expression of TLR-1,-2, -3 and -4 (Supplementary Fig. ); TLR-5 and -7 (Supplementary Fig. ); andTLR-11, -12, and -13 (Supplementary Fig. ) by germ cells at specific stages in the cycle of the seminiferous epithelium. No immune positivity was detected in any of the negative control sections used for each antibody (Supplementary Fig. ). Microscopic evaluation of the sections at a higher magnification (100×) further revealed that TLR-1, -2, -3, -4, -5, -7, -11, -12, and -13 were expressed by distinct populations of germ cells. Figures , , , , , , , , and show the expression of TLR-1, -2, -3, -4, -5, -7, -11, -12, and -13, respectively. TLR-1 was expressed by spermatocytes, round, and elongated spermatids. TLR-2, -4, -7, and -13 were only expressed by elongated spermatids. TLR-3, TLR-5, TLR-11, and TLR-12 were expressed by spermatocytes, round, and elongated spermatids, while spermatogonia expressed only TLR-11. Of these, expression of TLR-1, -3, -5, -11, and TLR-12 appeared at endosomal compartments of spermatocytes. TLR-1, -2, -3, -5, -11, and TLR-12 were expressed at the acrosomal complex at round and elongated spermatids. TLR-4, -5, -11, and -13 were specifically expressed in residual bodies either at the luminal surface of the seminiferous epithelium and/or near to the nuclei of Sertoli cells. A summary of the expression of TLRs by the type of germ cells is provided in Table . Results further revealed that the expression of TLR-1, -2, -3, -4, -5, -7, -11, -12, and -13 was not arbitrary and follows a distinct spatiotemporal pattern throughout the cycle of seminiferous epithelium (Figs. , , , , , , , , ). PAS staining on sequential section showing the expression of TLR-2, -3, -4, -5, -7, -11, -12, and -13 corresponding to a specific cycle of the seminiferous epithelium is provided in Supplementary Figs. – . TLR-1, -3, -5, -11, and -12 were expressed in all (the early, middle, and late) stages of the spermatogenic cycle. While the expression of TLR-4 was observed at the early and middle stages of spermatogenic cycle, TLR-2, -7, and -13 were expressed only at the early stage in the cycle of the seminiferous epithelium. A summary of the expression of TLRs by germ cells coinciding with a specific cycle of the seminiferous epithelium is provided in Fig. . TLRs are evolutionarily conserved proteins that play an indispensable role in innate immune system by recognizing various PAMPs derived from bacteria, fungi ,and protozoa (Takeda and Akira ). In addition to the specific PAMPs, TLRs can also detect endogenous ligands referred to as damage-associated molecular patterns released from damaged or dying cells (DAMPs, Janeway ; Yu et al. ; Behzadi et al. ). Binding of any of these ligands to a specific TLR triggers an innate immune response (Kawai and Akira ; Ward ), a prerequisite for the killing and clearance of various pathogens (Alexopoulou et al. ). In the testis tissue, involvement of TLRs in mediating testicular innate immune response is relatively well established for various somatic cells including Sertoli (Riccioli et al. ; Starace et al. ; Wu et al. ; Sun et al. ; Winnall et al. ) and Leydig cells (Shang et al. ). Nevertheless, to what extent germ cells are involved in this process remains unknown, except for TLR-3 and TLR-11. It was demonstrated that activation of TLR3 through a synthetic double-strained RNA analog leads to increased production of various proinflammatory cytokines and antiviral proteins in germ cells (Wang et al. ). Similarly, Chen et al. demonstrated in the mouse that activation of TLR-11 by Toxoplasma gondii -derived profilin and uropathogenic Escherichia coli (UPEC) can induce an innate immune response in germ cells through the production of inflammatory cytokines. In light of evidence provided in the present study demonstrating the expression of all functional TLRs in various populations of germ cells, further investigations are warranted to obtain a better and more comprehensive understanding of the role of each TLRs in modulating testicular innate immune response. There is ample evidence demonstrating that all TLRs, except for TLR-13, are expressed by male germ cells (Bhushan et al. ; Wang et al. ; Chen et al. ; Saeidi et al. ; Nejsum et al. ; Sun et al. ; Özbek et al. ; Öztop et al. ). However, in the mouse, with the model used in the present study, the expression pattern of TLRs by specific populations of germ cells has only been demonstrated for TLR-3 (Wang et al. ; Nejsum et al. ), TLR-9 (Mihara et al. ) and TLR-11 (Chen et al. ). Accordingly, spermatogonia and spermatocytes express both TLR-3 and TLR-11 (Wang et al. ; Chen et al. ; Nejsum et al. ). While TLR-11 is also expressed by spermatids (Chen et al. ), TLR-9 is expressed only by spermatozoa (Mihara et al. ). While confirming the expression of TLR-11 by spermatogonia, spermatocytes, and spermatids as well as the expression of TLR-3 by spermatocytes, results of the present study further reveal the expression pattern of the remaining functional TLRs by specific populations of germ cells. It appears that, on top of TLR-3 and -11, spermatocytes also express TLR-1, -5, and -12. It is also evident that the expression of TLRs by spermatids is not limited to TLR-11. While TLR-1, -3, -5, and -12 are expressed by round, and elongated spermatids, elongated spermatids express only TLR-2, -4, -7, and -13. To the best of our knowledge, the present study is the first revealing the expression pattern of all functional TLRs simultaneously by germ cells in the mouse testis throughout the cycle of the seminiferous epithelium from stage I to XII. It is evident from these observations that throughout spermatogenesis TLRs are differentially expressed by various populations of germ cells. Expression of TLR-1, -3, -5, -11, and -12 at endosomal compartments of elongated spermatids (Fig. ) and spermatocytes (Figs. , , , , ) and confinement of the expression of TLR-1, -2, -3, -5, -11 and -12 (Figs. , , , , , ) to the acrosomes of round and/or elongated spermatids are the two most interesting and novel findings of the present study. TLRs are synthesized in the endoplasmic reticulum (ER), transported to the Golgi bodies, and finally travel either to the cell surface (TLR-1–6 and -10) or stay in endosomes and/or lysosomes (TLR-3, -7, -8, -9, -11, -12, and -13) (Kawai and Akira ; Celhar et al. ; Kawasaki and Kawai ; Lee and Barton ). In light of evidence suggesting that endosomes might also give rise to the acrosome (Martínez-Menárguez et al. ; Sun-Wada et al. ; Moreno and Alvarado, ), results of the present study revealing to the best of our knowledge for the first time, the expression of specific TLRs at endosomal compartments and/or acrosome appear to be coherent with the synthesis and trafficking of TLRs. However, whether or not TLRs are differentially expressed by specific subsets of endosomal compartments such as early endosomes (Lakadamyali et al. ), recycling endosomes (Rink et al. ), late endosomes/multivesicular bodies (Russell et al. ), and lysosomes (Stein et al. ) warrant further investigations. Another interesting finding of the present study was the expression of TLR -4, -5, -11, and -13 at residual bodies, composed of various organelles, such as Golgi complex and ER, that the sperm cell no longer needs (de Kretser and Kerr ). Residual bodies are the cytoplasmic fragments of late spermatids which are removed at the time of sperm release (Syed et al. ). When spermatids are released into the lumen of the seminiferous epithelium, residual bodies are phagocytosed by Sertoli cells (O’Donnell et al. ), transported to the basal compartment, and catabolized (Johnson ). Lysosomes of Sertoli cells then fuse with the residual bodies to form phagolysosomes (de Kretser and Kerr ) resulting in phagocytosis of the residual bodies of germ cells (Wu et al. ; Chojnacka et al. ; Chen et al. ). In light of the fact that phagocytosis of residual bodies by Sertoli cells is an essential process for spermatogenesis (Wu et al. ; Li et al. ; Chojnacka et al. ; Chen et al. ), details surrounding the role of TLR -4, -5, -11, and -13 in this process warrant more in-depth studies. It is well established that activation of TLRs facilitates phagosome maturation (Blander and Medzhitov ) and is involved in the activation of autophagy (Xu et al. ). It is also interesting to note in this context that the removal of damaged and/or dysfunctional mitochondria in the residual bodies by mitophagy, viz. selective degradation of mitochondria by autophagy, is a critical process for the generation of individual spermatozoa and proper rearrangement of mitochondria (Sakai and Yamashina ; Ho and Wey ; Huang et al. ). Whether or not TLRs expressed at residual bodies of spermatids are involved in any of these processes remains to be determined. In any case, if and to what extent expression of specific TLRs at endosomal compartments, acrosomes, and/or residual bodies play in the regulation of the cycle of spermatogenesis remains an open question. Evidence gathered in the present study demonstrating differential expression of TLR-1, -2, -3, -4, -5, -7, -11, -12, and -13 by germ cells in accordance with the cycle of the seminiferous epithelium is arguably the most important finding of the present study. While TLR-1, -2, -3, -4, -5, -7, -11, -12, and -13 were expressed at the early (I–V) stages, TLR-1, -3, -4, -5, -11, and -12 were expressed at the middle (VI–VIII) stages in the spermatogenic cycle of the seminiferous epithelium. On the other hand, TLR-1, -3, -5, -11, and -12 were expressed at the late (IX–XII) stages. To the best of our knowledge, this is the first study demonstrating the spatiotemporal expression of all functional TLRs throughout the cycle of the seminiferous epithelium in a stage-specific manner. How stage-specific expression of TLRs is regulated remains elusive. In light of the fact that developing spermatogenic cells produce various autoantigens after gaining the ability to generate an immune response (Yule et al. ; Zhao et al. ), it is tempting to speculate that stage-specific expression of TLRs might be involved in preventing immune response to germ cell-specific as well as paternal major histocompatibility complex (MHC) antigens (Zhao et al. ). To what extent this process is associated with the immune-privileged status of the testis tissue also warrants further investigations (Head and Billingham ; Fijak et al. ). Considering that apoptosis is a physiological process of spermatogenesis (Nakanishi and Shiratsuchi ; Zhao et al. ; Zakariah et al. ) and that some members of the TLR family are capable of inducing apoptosis (Aliprantis et al. ; Salaun et al. ), TLRs might also be involved in the regulation of germ cell apoptosis. Furthermore, there is a substantial body of evidence indicating that, apart from their immune functions, TLRs play a role in various developmental processes including the regulation of neurogenesis (Rolls et al. ) and aging (Okun et al. ). In any case, whether or not the expression of TLRs by specific populations of germ cells in a cycle-dependent manner has any non-immune and/or developmental function(s) in the regulation of spermatogenesis remains to be determined. Taken together, results of the present study strengthen the hypothesis that the expression of TLRs by male germ cells is a developmentally regulated process and point to their possible involvement in the regulation of testicular functions (Hedger ) and spermatogenesis (Girling and Hedger ). Nevertheless, specific function of each TLR in sequential stages of proliferation, growth, maturation, and differentiation of germ cells throughout the cycle of the seminiferous epithelium remains elusive and warrants further investigations. Below is the link to the electronic supplementary material. Supplementary file1 (PDF 2628 KB) |
Accurate diagnosis and prognosis prediction of gastric cancer using deep learning on digital pathological images: A retrospective multicentre study | 4227c20c-2960-4e04-9133-b58f76842d57 | 8529077 | Pathology[mh] | We first searched PubMed and learned about relevant researches, and then carried out our project (Dec. 5, 2019). We searched for the keywords “gastric cancer” AND “deep learning” OR "artificial intelligence", with no restrictions on language or publication date. We learned that gastric cancer (GC) was the fifth most common type of malignant disease, and it ranks as the third leading cause of cancer-related deaths worldwide. Pathological evaluation remains the gold standard for the diagnosis of GC. When deciding on the necessity for further expensive and painful adjuvant treatments after surgery, clinicians tend to make decisions according to evidence-based information about the risk of death. Convolutional neural network (CNN) is a high-efficient deep learning method for image recognition and has excelled in quite a few image interpretation tasks and might be utilized to abstract additional characteristics from pathological images of GC patients. There had been a number of AI studies focusing on GC, most of which relied on endoscopy and medical radiologic technology. A few articles had focused on the application of deep learning algorithm to pathology of GC. There was still much room for improvement in developing AI models to diagnose GC and predict survival outcome through pathological pictures, especially to improve the performance of AI models. In this study, we designed a CNN-based model, GastroMIL, for the accurate diagnosis of GC directly from digital H&E-stained pictures. Encouragingly, this diagnosis model achieved excellent performance in differentiating GC from normal tissues in the training and internal validation sets (AUC nearly 1.000). Moreover, we have successfully developed a deep learning-based model, MIL-GC, to automatically predict OS in patients with GC with C-index of 0.728 and 0.671 in the training and internal validation sets. And we also used an independent external validation set, and the two models showed good diagnostic (AUC = 0.978) and prognostic (C-index = 0.657) prediction performance, indicating good robustness of the two designed models. And the risk score computed by MIL-GC was proved to be of independent prognostic value of GC by univariate and multivariable Cox analyze. In the comparison with human pathologists, the diagnostic model (GastroMIL) achieved accuracy better than that of the junior pathologist and comparable to that of expert pathologists. More importantly, we further constructed the first webpage (https://baigao.github.io/Pathologic-Prognostic-Analysis/) for the automatic diagnosis of GC and survival prediction. Our models can be adopted to make diagnosis with high accuracy and help clinicians select the appropriate adjuvant therapy following surgery, by identifying patients at high risk who would benefit from intensive regimens as well as patients at low risk who might be cured through surgery alone. It will help improve the survival status of GC patients and reduce the high mortality. Gastric cancer (GC) is the fifth most common type of malignant disease, and it ranks as the third leading cause of cancer-related deaths worldwide . For patients with early GC, the 5-year survival rate can exceed 90% . However, approximately half of patients with GC already proceed the advanced stage at the time of diagnosis, with the 5-year survival rate dropping below 30% . To reduce the mortality of GC, early detection and appropriate treatment are crucial, and precise and efficient pathology services are indispensable to realize this goal. Pathological evaluation remains the gold standard for the diagnosis of GC. Conventionally carried out by pathologists, this method is labor-intensive, tedious, and time-consuming. A severe shortage of pathologists and a heavy workload of diagnosis are widespread problems globally, which negatively affect the diagnostic accuracy . Accordingly, it is necessary to design a new method to conveniently and accurately diagnose GC using pathological pictures. Surgery is the main treatment for GC, followed by adjuvant treatments including chemoradiotherapy and molecular targeted therapy , , . When deciding on the necessity for further expensive and painful adjuvant treatments, clinicians tend to make decisions according to evidence-based information about the risk of death. Clinical practice has confirmed that prognoses of almost all human cancers, including GC, are closely related to pathological criteria , especially the tumour-node-metastasis (TNM) staging system specified and revised by the American Joint Committee on Cancer (AJCC) . However, manual histological analysis of tumour tissues is still not accurate enough to stratify and identify those who may benefit from adjuvant treatment. Hence, there is an urgent need to develop succinct and reliable methods to predict overall survival (OS) of patients with GC, which could assist in developing individualized therapeutic strategies and maximizing the benefits. In recent years, deep learning has gradually attracted the attention of oncologists. Deep learning belongs to the class of machine learning that can successively identify more abstract information from the input data , , . Deep learning has progressed remarkably in the field of oncology, and has been demonstrated to be superior to conventional machine learning techniques . Convolutional neural network (CNN) is a high-efficient deep learning method for image recognition and has excelled in quite a few images interpretation tasks . Many studies have reported that artificial intelligence (AI) trained with endoscopic images could detect GC precisely , , , . When it comes to the field of tumour detection and prediction of prognosis of GC using AI through pathological images, some progress has been made. Song et al. reported a histopathological diagnosis system for GC detection using deep learning with the sensitivity near 100% and average specificity of 80.6%. Another research developed recalibrated multi-instance deep learning for whole slide gastric image classification with 86% accuracy . Wang et al. successfully predicted GC outcome from resected lymph node histopathology images using deep learning. Before proposing our models, we had established a number of challenges that needed to be overcome in order to make the developed AI models better applicable to clinical practice. First of all, a large sample size from multiple centres should be available for training and validating the proposed model to ensure the robustness. While ensuring the effectiveness of model training, it is preferable not to rely on extensive manual pixel-level annotation, which would be laborious and time consuming and might hinder the development of AI in the field of pathology. The developed model should be able to be applied in clinical practice, and it should be simple, affordable and accessible enough to be easily used by people without an AI background or in places where the economy is not particularly developed. We hoped to accomplish these challenges better than previous studies. In this study, we developed deep learning-based models, named GastroMIL and MIL-GC, for precisely and conveniently detecting tumour and predicting outcome of GC by analyzing pathological pictures, respectively. GastroMIL and MIL-GC were proved to be novel and strong predictors for diagnosis and outcome of GC patients on both internal and independent external validation sets. In the comparison with human pathologists, our GastroMIL model outperformed the junior pathologist and achieved a great agreement with expert pathologists. Furthermore, we designed an online website ( https://baigao.github.io/Pathologic-Prognostic-Analysis/ ) based on our analysis to make this prediction more available to users who have no knowledge of AI. Patient Population Three different cohorts were retrospectively collected to achieve a broad patient representation and thereby improve the ability to generalize results to other cohorts. In the Renmin Hospital of Wuhan University (RHWU; Wuhan, Hubei, China), we continuously collected 871 candidate patients with GC from 2012 to 2017, together with corresponding 588 tumour tissue blocks (made from surgically removed tumour tissue, which was formalin-fixed and paraffin-embedded) and 1276 pathological images. 1057 digital H&E-stained pictures of 449 GC patients from The Cancer Genome Atlas (TCGA) public dataset were collected, 934 of which were malignant and 123 were normal. In addition, 91 GC patients with 175 digital pictures were acquired from National Human Genetic Resources Sharing Service Platform (NHGRP; Shanghai, China) and served as the independent validation set to evaluate the robustness of our models. We adopted the following inclusion criteria for developing the diagnostic model: (a) patients unequivocal diagnosed with GC by preoperative biopsy or postoperative pathological examination; (b) patients older than 18 years old and assentient to participate in this study; and (c) pathological images available and clear rather than loss, destruction or mildew. Pictures used in the diagnostic model were excluded from the prognostic model when they met the following conditions: (a) identified as normal; (b) lack of follow-up information; and (c) no critical clinicopathologic information available. Sample collection Digital images of H&E-stained pathological images were used to construct the computer frameworks. For each GC patient, we selected two representative images in principle, which included tissues from not only GC tumour but also surrounding normal gastric tissues. For candidate patients from RHWU, the corresponding formalin-fixed, paraffin-embedded tumour tissue blocks, made from surgically removed tumour tissue, and their H&E-stained slides were obtained. Next, two expert pathologists A and B selected preferred blocks and marked areas that were cancerous or normal independently. When it came to a disagreement between them, the diagnostic opinion of another expert pathologist C was final adopted. Expert pathologists A and B were associate chief pathologists, while expert pathologists C was chief pathologist. The marked areas were utilized to construct tissue microarrays (TMAs), which were then photographed to obtain 1276 digital H&E-stained images, of which 640 were malignant and 636 were normal. For the TCGA cohort, 1057 pathological pictures (malignant 936, normal 123) were downloaded from the website ( https://www.cbioportal.org/study/summary?id=lihc_tcga ). In view of the uneven number of cancerous and normal pictures in the TCGA cohort, the data augmentation technique was used to equalize the distribution of images. In the external validation cohort from NHGRP, there were 91 malignant digital pathological images and 84 normal. Clinical and pathological information was additionally needed for survival analysis, including survival state, OS time, age, sex, tumour size, neoplasm histologic grade, and pathologic T,N, M and TNM stages (according to the American Joint Committee on Cancer (AJCC) Cancer Staging Manual, Eighth Edition, 2017) . Clinicopathological data of patients from the RHWU cohort were collected through electronic medical records, and those of the TCGA and NHGRP cohorts were downloaded directly from the official website. This retrospective study was checked and approved by the clinical ethics committees of RHWU (No. WDRY2021-K002). And informed consents were gained from patients. Diagnostic Model Firstly, we designed a diagnostic model, named GastroMIL, to distinguish GC images from normal gastric tissue images. In order to avoid complex manual annotation, we applied weak supervised learning to our algorithm framework, specifically multiple instance learning (MIL) , , , . Based on the assumption of MIL, each input image was a bag, and the tiles it contained were the example instances. To develop the model, we only needed coarse-grained labels of bags, that is, pathological diagnosis of each image. When the target picture was marked positive, at least one tile was positive; if the target picture was marked negative, all tiles should be negative. Given that the images used to train the model had different magnifications, specifical the original magnification of images from TCGA was 20 × (without fixed size, could larger than 30000*30000 pixels), whereas that of RHWU was 30 × (3200*2400 pixels), we uniformly processed these images into 5 ×, 10 ×, and 20 × magnification and use them to develop the algorithm separately. Our GastroMIL model comprised two-step algorithms ( a-b). First of all, each input image was split into fixed-size tiles with 224 × 224 pixels, the labels of which were the same as the pathological diagnoses of the image itself. These tiles were used as training data for the first step algorithm, the MIL classifier. There were 10548460 tiles used for training and 4523755 tiles used for internal validation. Considering accuracy and efficiency, we chose RegNet developed by Facebook to constitute the backbone of MIL. The output of RegNet was the probability of these tiles being malignant. To obtain the inference results of the complete pictures, we introduced a recurrent neural network (RNN) as the second step classifier. Feature vectors with dimension 608 of the 32 most suspicious tiles gained from each picture by the first step were sequentially passed on to the RNN classifier to predict the probability of malignancy of the entire picture. The GastroMIL model could thus not only distinguish malignant images from normal, but also identify regions of interest (ROIs) by the segmentation and analysis of tiles. ROI indicated the area recognized as malignant by GastroMIL, which could be visualized in the form of heat map ( b) and provide additional guiding information for clinicians. Prognostic Model After identifying the malignant images, we developed another model, MIL-GC, to predict the prognosis of GC patients. As shown in a-c, the first step of MIL-GC was similar to that of the GastroMIL model, and the 128 most suspicious tiles were selected and output as feature vectors with dimension 608. In the second step, each feature vector would finally yield a probability value between 0 and 1, through a multilayer perceptron (MLP) algorithm. Probability values of the 128 most suspicious tiles of the input picture were merged to generate an average value as the output risk score. Statistical analysis Receiver operating characteristic (ROC) curves and areas under the curve (AUCs), analysed with scikit-learn, a Python software package for machine learning, were used to quantify the performance of the diagnostic classifier as well as accuracy, sensitivity, and specificity. The cut-off value of ROC curves was set as 0.5. Cohen's kappa coefficient was used to assess the inter-observer agreement of the diagnostic model (GastroMIL) and human pathologists. To assess the predictive performance of the prognostic classifier, we adopted Harrell's concordance index (C-index) as a metric. Kaplan-Meier survival curve plotted with GraphPad Prism_9 was used to evaluate the correlation between risk score generated by prognostic models and OS of GC patients and Log-Rank test was performed. Prognostic factors were identified using univariate and multivariate Cox proportional hazards models implemented in SPSS26.0. The statistical significance level was set at 0.05 (two-tailed). Statistical significance threshold was adjusted for multiple comparisons using the Bonferroni correction. Python and Pytorch were employed to build the algorithm. Role of the funding source Not applicable. Three different cohorts were retrospectively collected to achieve a broad patient representation and thereby improve the ability to generalize results to other cohorts. In the Renmin Hospital of Wuhan University (RHWU; Wuhan, Hubei, China), we continuously collected 871 candidate patients with GC from 2012 to 2017, together with corresponding 588 tumour tissue blocks (made from surgically removed tumour tissue, which was formalin-fixed and paraffin-embedded) and 1276 pathological images. 1057 digital H&E-stained pictures of 449 GC patients from The Cancer Genome Atlas (TCGA) public dataset were collected, 934 of which were malignant and 123 were normal. In addition, 91 GC patients with 175 digital pictures were acquired from National Human Genetic Resources Sharing Service Platform (NHGRP; Shanghai, China) and served as the independent validation set to evaluate the robustness of our models. We adopted the following inclusion criteria for developing the diagnostic model: (a) patients unequivocal diagnosed with GC by preoperative biopsy or postoperative pathological examination; (b) patients older than 18 years old and assentient to participate in this study; and (c) pathological images available and clear rather than loss, destruction or mildew. Pictures used in the diagnostic model were excluded from the prognostic model when they met the following conditions: (a) identified as normal; (b) lack of follow-up information; and (c) no critical clinicopathologic information available. Digital images of H&E-stained pathological images were used to construct the computer frameworks. For each GC patient, we selected two representative images in principle, which included tissues from not only GC tumour but also surrounding normal gastric tissues. For candidate patients from RHWU, the corresponding formalin-fixed, paraffin-embedded tumour tissue blocks, made from surgically removed tumour tissue, and their H&E-stained slides were obtained. Next, two expert pathologists A and B selected preferred blocks and marked areas that were cancerous or normal independently. When it came to a disagreement between them, the diagnostic opinion of another expert pathologist C was final adopted. Expert pathologists A and B were associate chief pathologists, while expert pathologists C was chief pathologist. The marked areas were utilized to construct tissue microarrays (TMAs), which were then photographed to obtain 1276 digital H&E-stained images, of which 640 were malignant and 636 were normal. For the TCGA cohort, 1057 pathological pictures (malignant 936, normal 123) were downloaded from the website ( https://www.cbioportal.org/study/summary?id=lihc_tcga ). In view of the uneven number of cancerous and normal pictures in the TCGA cohort, the data augmentation technique was used to equalize the distribution of images. In the external validation cohort from NHGRP, there were 91 malignant digital pathological images and 84 normal. Clinical and pathological information was additionally needed for survival analysis, including survival state, OS time, age, sex, tumour size, neoplasm histologic grade, and pathologic T,N, M and TNM stages (according to the American Joint Committee on Cancer (AJCC) Cancer Staging Manual, Eighth Edition, 2017) . Clinicopathological data of patients from the RHWU cohort were collected through electronic medical records, and those of the TCGA and NHGRP cohorts were downloaded directly from the official website. This retrospective study was checked and approved by the clinical ethics committees of RHWU (No. WDRY2021-K002). And informed consents were gained from patients. Firstly, we designed a diagnostic model, named GastroMIL, to distinguish GC images from normal gastric tissue images. In order to avoid complex manual annotation, we applied weak supervised learning to our algorithm framework, specifically multiple instance learning (MIL) , , , . Based on the assumption of MIL, each input image was a bag, and the tiles it contained were the example instances. To develop the model, we only needed coarse-grained labels of bags, that is, pathological diagnosis of each image. When the target picture was marked positive, at least one tile was positive; if the target picture was marked negative, all tiles should be negative. Given that the images used to train the model had different magnifications, specifical the original magnification of images from TCGA was 20 × (without fixed size, could larger than 30000*30000 pixels), whereas that of RHWU was 30 × (3200*2400 pixels), we uniformly processed these images into 5 ×, 10 ×, and 20 × magnification and use them to develop the algorithm separately. Our GastroMIL model comprised two-step algorithms ( a-b). First of all, each input image was split into fixed-size tiles with 224 × 224 pixels, the labels of which were the same as the pathological diagnoses of the image itself. These tiles were used as training data for the first step algorithm, the MIL classifier. There were 10548460 tiles used for training and 4523755 tiles used for internal validation. Considering accuracy and efficiency, we chose RegNet developed by Facebook to constitute the backbone of MIL. The output of RegNet was the probability of these tiles being malignant. To obtain the inference results of the complete pictures, we introduced a recurrent neural network (RNN) as the second step classifier. Feature vectors with dimension 608 of the 32 most suspicious tiles gained from each picture by the first step were sequentially passed on to the RNN classifier to predict the probability of malignancy of the entire picture. The GastroMIL model could thus not only distinguish malignant images from normal, but also identify regions of interest (ROIs) by the segmentation and analysis of tiles. ROI indicated the area recognized as malignant by GastroMIL, which could be visualized in the form of heat map ( b) and provide additional guiding information for clinicians. After identifying the malignant images, we developed another model, MIL-GC, to predict the prognosis of GC patients. As shown in a-c, the first step of MIL-GC was similar to that of the GastroMIL model, and the 128 most suspicious tiles were selected and output as feature vectors with dimension 608. In the second step, each feature vector would finally yield a probability value between 0 and 1, through a multilayer perceptron (MLP) algorithm. Probability values of the 128 most suspicious tiles of the input picture were merged to generate an average value as the output risk score. Receiver operating characteristic (ROC) curves and areas under the curve (AUCs), analysed with scikit-learn, a Python software package for machine learning, were used to quantify the performance of the diagnostic classifier as well as accuracy, sensitivity, and specificity. The cut-off value of ROC curves was set as 0.5. Cohen's kappa coefficient was used to assess the inter-observer agreement of the diagnostic model (GastroMIL) and human pathologists. To assess the predictive performance of the prognostic classifier, we adopted Harrell's concordance index (C-index) as a metric. Kaplan-Meier survival curve plotted with GraphPad Prism_9 was used to evaluate the correlation between risk score generated by prognostic models and OS of GC patients and Log-Rank test was performed. Prognostic factors were identified using univariate and multivariate Cox proportional hazards models implemented in SPSS26.0. The statistical significance level was set at 0.05 (two-tailed). Statistical significance threshold was adjusted for multiple comparisons using the Bonferroni correction. Python and Pytorch were employed to build the algorithm. Not applicable. Patient characteristics A total of 871 GC patients were initially screened from the RHWU cohort, and 588 with tumour tissue blocks were eligible for the study. There were 449 GC patients with digital H&E-stained pathological images were eligible for this study in the TCGA cohort and 91 in the NHGRP cohort. A total of 1276 images from the RHWU cohort and 1057 images from the TCGA cohort were obtained for the development of the GastroMIL model. Through data augmentation, 3221 pictures (malignant: normal = 1574: 1647) were finally enrolled in the GastroMIL model and 70% ( N = 2261) were randomly assigned to the training set while the remaining 30% ( N = 960) were included in the internal validation set. 175 pictures from the independent NHGRP cohort were used as the external validation set. The detailed data distribution was shown in Supplementary Table 1. A total of 199 malignant pathological pictures with intact follow-up and clinicopathological information from the RHWU cohort and 440 from the TCGA cohort participated in the construction of the prognostic model and then randomly spilt into training set ( N = 443) and internal validation set ( N = 196) at a ratio of 70: 30. 91 GC digital pathological pictures with the required information from NHGRP were included in the external validation set. exhibits the baseline characteristics of the pictures used in MIL-GC. It is worth noting that the OS time of the TCGA cohort was significantly lower than that of the RHWU cohort (median of 13.8 months vs. 43 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) and NHGRP cohort (median of 13.8 months vs. 44 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) (Supplementary Fig. 1a). Differences between the RHWU cohort (median 43 months) and NHGRP cohort (median 44 months) were not statistically significant ( P = 0.075, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017). Since pictures in the TCGA cohort were collected from different medical centres, the distribution of their OS time was much more heterogeneous. To better adapt our models to the heterogeneity caused by patients from different sources and to improve the generalizability of the developed models, we used a mixture of images from the RHWU and TCGA cohort together as the training and internal validation sets. The independent NHGRP cohort of images was used as the external validation set. The OS time of the external validation set (median 44 months) was significantly higher compared with the training set (median 20.2 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) and the internal validation set (median 22.4 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) (Supplementary Fig. 1b). Performance of the Diagnostic Model In the diagnostic model, GastroMIL, all 3221 pictures after data augmentation were mixed and then randomly split into the training set and the internal validation set at a ratio of 7: 3. As shown in , ROC curves and AUCs represented the ability to discriminate malignant pathological images of the GastroMIL model when pictures were at 5 ×, 10 × and 20 × magnification. The accuracy (Acc), sensitivity (Sen), and specificity (Spe) of each magnification in the training ( d) and internal validation ( h) sets are shown in a. In the training set, the AUCs achieved 1.000 at three different magnifications. The differences of Acc between the three groups was statistically significant ( P = 0.003, Chi-square test) (significance threshold P < 0.05). The group of 10 × magnification (Acc = 1.000) outperformed that of 20 × (Acc = 0.996) ( P = 0.004, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017). There was no statistically significant difference in Acc between the groups of 5 × and 10 × (0.999 vs. 1.000, P = 0.250, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017) and between the groups of 5 × and 20 × (0.999 vs. 0.996, P = 0.109, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017). In the internal validation set, the AUC achieved 0.995 when pictures were magnified 10 times. The AUCs were also very close to it when images were magnified 5 times (AUC = 0.995) and 20 times (AUC = 0.994). The differences in Acc among the three groups of different magnifications (0.976, 0.976, and 0.979) were not statistically significant ( P = 0.870, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017). It can be seen that our GastroMIL model achieved excellent diagnostic ability for the differentiation of malignant and normal gastric pathological pictures, and the generalization performance was excellent for images at these three magnifications. The diagnosis prediction by the GastroMIL model was based on the classification probability output of all tiles, which can be used to visualize the localization of highly suspected lesions on positive sections. That is, tiles predicted positive could show chromatic aberration, and through appropriate strides, suspected areas would overlap many times, colour of which thus became warmer and darker than other areas. The warmer the colour, the higher the probability that GastroMIL predicted malignant in this area. b showed how heat maps generated by GastroMIL. Heat maps of the RHWU and TCGA (Supplementary Fig. 2) cohorts could almost accurately outline the area where the tumour was located, regardless of different cohorts or pathological TNM stage, indicating excellent generalization performance of GastroMIL. Performance of the prognostic models In the process of outcome prediction, 639 malignant images from RHWU and TCGA cohorts were mixed and then randomly divided into the training set and the internal validation set at a 7:3 ratio. Considering that the discriminatory power of the GastroMIL model for different magnification images was basically identical to each other, we chose 10 × images to be applied in the prognosis model. The MIL-GC model performed well in both the training set and the internal validation set, with C-index of 0.728 and 0.671, respectively. Prognostic model assigned risk score to each picture, and we divided the GC patients into high-risk and low-risk score groups. We used the median value of risk score in the training set as the threshold for stratifying patients. Then we adopted Kaplan-Meier plots and univariate and multivariable Cox models to assess the association between risk score and prognosis among patients with GC. In the training set, the MIL-GC classifier was a strong predictor of survival in the univariate analysis (HR = 4.209, P < 0.0001, Cox analyse; Supplementary Table 2 and Supplementary Fig. 3). The classifier remained strong in multivariable analysis (HR = 3.549, P < 0.0001, Cox analyse; Supplementary Table 2) after adjusting for significant prognostic indexes in univariable analyses: age, pT stage, pN stage, pM stage and pTNM stage. In the internal validation set, our model stratified the population accurately based on univariate analysis (HR = 3.249, P < 0.0001, Cox analyse; a and c). The MIL-GC classifier predicted survival even after stratification for other features (such as age, histologic grade, pT grade, pN grade and pTNM grade; ). The risk score computed by MIL-GC was of independent prognostic value (HR = 2.976, P < 0.0001, Cox analyse; b). The results showed that the prognostic model based on CNN was equipped to predict OS of GC and might provide a basis for the choice of treatment. Independent external validation of developed models Figures from the independent NHGRP cohort were employed as the external validation set for the diagnosis prediction model, GastroMIL and outcome prediction model, MIL-GC. The magnification of the original images is 20 × (3900*3900 pixels), and we pre-processed them into 10 × magnification. The GastroMIL model showed good performance in identifying malignant pathological images on the external validation set (AUC = 0.978, a). Heat maps of the independent external validation set were displayed in Supplementary Fig. 4. The C-index of MIL-GC in the external validation set was 0.657. The MIL-GC classifier was a strong predictor of OS in the univariate analysis (HR = 2.414, P < 0.0001, Cox analyse; b and c). The MIL-GC classifier predicted survival among the various subgroups (such as age > 60, tumour size ≤ 5, histologic grade 3, pT stage 3, pN stage 0 and 3, pM stage 0, pTNM stage II and pTNM stage III; ). After adjusting for significant prognostic indexes in univariable analyses: histologic grade, pT stage, pN stage, pM stage and pTNM stage, the risk score output by MIL-GC remained strong in multivariable analysis (HR = 1.803, P = 0.043, Cox analyse; d). Good diagnostic and prognostic prediction performance demonstrated on the external validation set, indicating good robustness of the two designed models. Comparing diagnostic performance with human pathologists To explore how the diagnostic performance of our model compared to that of human pathologists, we employed three expert pathologists D, E, and F who were chief or associate chief pathologists and one junior pathologist G who was under training to diagnose images in the external validation set. Human pathologists were blind for patients’ information before examination. The accuracy, sensitivity, and specificity of manual diagnosis were exhibited in Table 2c. The performance of our GastroMIL model (Accuracy = 0.920) was significantly better than the junior pathologist G (Accuracy = 0.874) ( P < 0.0001, paired chi‐square test). There was no significantly difference when our model compared to expert pathologist D (Accuracy = 0.971) ( P > 0.05, paired chi‐square test), expert pathologist E (Accuracy = 0.983) ( P > 0.05, paired chi‐square test), and expert pathologist F (Accuracy = 0.983) ( P > 0.05, paired chi‐square test), respectively. And the diagnostic model achieved substantial interobserver agreement with the expert pathologists ( kappa = 0.805, 0.806, and 0.806, respectively). Moreover, we designed an online website ( https://baigao.github.io/Pathologic-Prognostic-Analysis/ ) to make the process of prediction more available and much easier for users without AI knowledge. The detail process of prediction is seen in Supplementary Fig. 5. Analysis of representative predictive tiles Our models predicted diagnosis and outcome of GC patients from the 32 and 128 most suspicious tiles automatically gained from the input HE-stained pathological pictures, respectively. We extracted these suspicious tiles and had them reviewed by expert pathologists A and B. Here we displayed some representative predictive tiles with interpretation by pathological experts . These tiles were of obvious tumour heterogeneity, including necrosis, nerve invasion, signet ring cell, intravasated cancer cells, muscularis propria invasion, and mucous secretion, hiding significant diagnostic and prognostic information. These suspicious tiles provide a preliminary indication that our model could automatically identify regions of pathological significance and classify GC pathology images based on these regions. Further details on the analysis of deep learning, which has been traditionally treated as a black box, deserve our further study in the future. A total of 871 GC patients were initially screened from the RHWU cohort, and 588 with tumour tissue blocks were eligible for the study. There were 449 GC patients with digital H&E-stained pathological images were eligible for this study in the TCGA cohort and 91 in the NHGRP cohort. A total of 1276 images from the RHWU cohort and 1057 images from the TCGA cohort were obtained for the development of the GastroMIL model. Through data augmentation, 3221 pictures (malignant: normal = 1574: 1647) were finally enrolled in the GastroMIL model and 70% ( N = 2261) were randomly assigned to the training set while the remaining 30% ( N = 960) were included in the internal validation set. 175 pictures from the independent NHGRP cohort were used as the external validation set. The detailed data distribution was shown in Supplementary Table 1. A total of 199 malignant pathological pictures with intact follow-up and clinicopathological information from the RHWU cohort and 440 from the TCGA cohort participated in the construction of the prognostic model and then randomly spilt into training set ( N = 443) and internal validation set ( N = 196) at a ratio of 70: 30. 91 GC digital pathological pictures with the required information from NHGRP were included in the external validation set. exhibits the baseline characteristics of the pictures used in MIL-GC. It is worth noting that the OS time of the TCGA cohort was significantly lower than that of the RHWU cohort (median of 13.8 months vs. 43 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) and NHGRP cohort (median of 13.8 months vs. 44 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) (Supplementary Fig. 1a). Differences between the RHWU cohort (median 43 months) and NHGRP cohort (median 44 months) were not statistically significant ( P = 0.075, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017). Since pictures in the TCGA cohort were collected from different medical centres, the distribution of their OS time was much more heterogeneous. To better adapt our models to the heterogeneity caused by patients from different sources and to improve the generalizability of the developed models, we used a mixture of images from the RHWU and TCGA cohort together as the training and internal validation sets. The independent NHGRP cohort of images was used as the external validation set. The OS time of the external validation set (median 44 months) was significantly higher compared with the training set (median 20.2 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) and the internal validation set (median 22.4 months, P < 0.0001, Kruskal-Wallis nonparametric test) (Bonferroni-adjusted significance threshold P’ < 0.017) (Supplementary Fig. 1b). In the diagnostic model, GastroMIL, all 3221 pictures after data augmentation were mixed and then randomly split into the training set and the internal validation set at a ratio of 7: 3. As shown in , ROC curves and AUCs represented the ability to discriminate malignant pathological images of the GastroMIL model when pictures were at 5 ×, 10 × and 20 × magnification. The accuracy (Acc), sensitivity (Sen), and specificity (Spe) of each magnification in the training ( d) and internal validation ( h) sets are shown in a. In the training set, the AUCs achieved 1.000 at three different magnifications. The differences of Acc between the three groups was statistically significant ( P = 0.003, Chi-square test) (significance threshold P < 0.05). The group of 10 × magnification (Acc = 1.000) outperformed that of 20 × (Acc = 0.996) ( P = 0.004, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017). There was no statistically significant difference in Acc between the groups of 5 × and 10 × (0.999 vs. 1.000, P = 0.250, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017) and between the groups of 5 × and 20 × (0.999 vs. 0.996, P = 0.109, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017). In the internal validation set, the AUC achieved 0.995 when pictures were magnified 10 times. The AUCs were also very close to it when images were magnified 5 times (AUC = 0.995) and 20 times (AUC = 0.994). The differences in Acc among the three groups of different magnifications (0.976, 0.976, and 0.979) were not statistically significant ( P = 0.870, Chi-square test) (Bonferroni-adjusted significance threshold P’ < 0.017). It can be seen that our GastroMIL model achieved excellent diagnostic ability for the differentiation of malignant and normal gastric pathological pictures, and the generalization performance was excellent for images at these three magnifications. The diagnosis prediction by the GastroMIL model was based on the classification probability output of all tiles, which can be used to visualize the localization of highly suspected lesions on positive sections. That is, tiles predicted positive could show chromatic aberration, and through appropriate strides, suspected areas would overlap many times, colour of which thus became warmer and darker than other areas. The warmer the colour, the higher the probability that GastroMIL predicted malignant in this area. b showed how heat maps generated by GastroMIL. Heat maps of the RHWU and TCGA (Supplementary Fig. 2) cohorts could almost accurately outline the area where the tumour was located, regardless of different cohorts or pathological TNM stage, indicating excellent generalization performance of GastroMIL. In the process of outcome prediction, 639 malignant images from RHWU and TCGA cohorts were mixed and then randomly divided into the training set and the internal validation set at a 7:3 ratio. Considering that the discriminatory power of the GastroMIL model for different magnification images was basically identical to each other, we chose 10 × images to be applied in the prognosis model. The MIL-GC model performed well in both the training set and the internal validation set, with C-index of 0.728 and 0.671, respectively. Prognostic model assigned risk score to each picture, and we divided the GC patients into high-risk and low-risk score groups. We used the median value of risk score in the training set as the threshold for stratifying patients. Then we adopted Kaplan-Meier plots and univariate and multivariable Cox models to assess the association between risk score and prognosis among patients with GC. In the training set, the MIL-GC classifier was a strong predictor of survival in the univariate analysis (HR = 4.209, P < 0.0001, Cox analyse; Supplementary Table 2 and Supplementary Fig. 3). The classifier remained strong in multivariable analysis (HR = 3.549, P < 0.0001, Cox analyse; Supplementary Table 2) after adjusting for significant prognostic indexes in univariable analyses: age, pT stage, pN stage, pM stage and pTNM stage. In the internal validation set, our model stratified the population accurately based on univariate analysis (HR = 3.249, P < 0.0001, Cox analyse; a and c). The MIL-GC classifier predicted survival even after stratification for other features (such as age, histologic grade, pT grade, pN grade and pTNM grade; ). The risk score computed by MIL-GC was of independent prognostic value (HR = 2.976, P < 0.0001, Cox analyse; b). The results showed that the prognostic model based on CNN was equipped to predict OS of GC and might provide a basis for the choice of treatment. Figures from the independent NHGRP cohort were employed as the external validation set for the diagnosis prediction model, GastroMIL and outcome prediction model, MIL-GC. The magnification of the original images is 20 × (3900*3900 pixels), and we pre-processed them into 10 × magnification. The GastroMIL model showed good performance in identifying malignant pathological images on the external validation set (AUC = 0.978, a). Heat maps of the independent external validation set were displayed in Supplementary Fig. 4. The C-index of MIL-GC in the external validation set was 0.657. The MIL-GC classifier was a strong predictor of OS in the univariate analysis (HR = 2.414, P < 0.0001, Cox analyse; b and c). The MIL-GC classifier predicted survival among the various subgroups (such as age > 60, tumour size ≤ 5, histologic grade 3, pT stage 3, pN stage 0 and 3, pM stage 0, pTNM stage II and pTNM stage III; ). After adjusting for significant prognostic indexes in univariable analyses: histologic grade, pT stage, pN stage, pM stage and pTNM stage, the risk score output by MIL-GC remained strong in multivariable analysis (HR = 1.803, P = 0.043, Cox analyse; d). Good diagnostic and prognostic prediction performance demonstrated on the external validation set, indicating good robustness of the two designed models. To explore how the diagnostic performance of our model compared to that of human pathologists, we employed three expert pathologists D, E, and F who were chief or associate chief pathologists and one junior pathologist G who was under training to diagnose images in the external validation set. Human pathologists were blind for patients’ information before examination. The accuracy, sensitivity, and specificity of manual diagnosis were exhibited in Table 2c. The performance of our GastroMIL model (Accuracy = 0.920) was significantly better than the junior pathologist G (Accuracy = 0.874) ( P < 0.0001, paired chi‐square test). There was no significantly difference when our model compared to expert pathologist D (Accuracy = 0.971) ( P > 0.05, paired chi‐square test), expert pathologist E (Accuracy = 0.983) ( P > 0.05, paired chi‐square test), and expert pathologist F (Accuracy = 0.983) ( P > 0.05, paired chi‐square test), respectively. And the diagnostic model achieved substantial interobserver agreement with the expert pathologists ( kappa = 0.805, 0.806, and 0.806, respectively). Moreover, we designed an online website ( https://baigao.github.io/Pathologic-Prognostic-Analysis/ ) to make the process of prediction more available and much easier for users without AI knowledge. The detail process of prediction is seen in Supplementary Fig. 5. Our models predicted diagnosis and outcome of GC patients from the 32 and 128 most suspicious tiles automatically gained from the input HE-stained pathological pictures, respectively. We extracted these suspicious tiles and had them reviewed by expert pathologists A and B. Here we displayed some representative predictive tiles with interpretation by pathological experts . These tiles were of obvious tumour heterogeneity, including necrosis, nerve invasion, signet ring cell, intravasated cancer cells, muscularis propria invasion, and mucous secretion, hiding significant diagnostic and prognostic information. These suspicious tiles provide a preliminary indication that our model could automatically identify regions of pathological significance and classify GC pathology images based on these regions. Further details on the analysis of deep learning, which has been traditionally treated as a black box, deserve our further study in the future. In this study, we designed a CNN-based model, Gastro-MIL, for the accurate diagnosis of GC directly from digital H&E-stained pictures. Encouragingly, this diagnosis model achieved excellent performance in differentiating GC from normal tissues in the training and internal validation sets (AUC nearly 1.000). Moreover, we have successfully employed deep learning to automatically predict OS in patients with GC with C-index of 0.728 and 0.671 in the training and internal validation sets. The predictive models can be adopted to help clinicians select the appropriate adjuvant therapy following surgery, by identifying patients at high risk who would benefit from intensive regimens as well as patients at low risk who might be cured through surgery alone. And we also used an independent external validation set, and the two models showed good diagnostic (AUC = 0.978) and prognostic (C-index = 0.657) prediction performance, indicating good robustness of the two designed models. Moreover, the risk score computed by MIL-GC was proved to be of independent prognostic value of GC by univariate and multivariable Cox analyse. In the comparison with human pathologists, our diagnostic model GastroMIL outperformed the junior pathologists and demonstrated a high degree of consistence with expert pathologists ( kappa > 0.8). More importantly, we further constructed the first webpage ( https://baigao.github.io/Pathologic-Prognostic-Analysis/ ) for the automatic diagnosis of GC and survival prediction. In recent years, deep learning, such as CNN, has attracted much attention and has achieved particular success in computer vision tasks. In our previous study, we developed an AI model to distinguish abnormal images from normal images in small bowel capsule endoscopy, and it was validated to exceed human performance . In this study, we adopted CNN to analyse digital H&E-stained GC pathological images. Song et al. reported deep learning model for GC detection by analysing histopathological images with validation in multicentre sample. They reached good performance and developed the system for pathologists to use the proposed model. However, they applied a large number of pixel-level manual annotations to train the model, which consumed a lot of time and effort of pathologists. The need for extensive manual annotation was also seen in the studies using AI for GC and bladder pathological diagnosis, different from the models proposed in our study. Due to the adoption of a weakly supervised model (specifical MIL) in our study, the only label we needed for training was the reported diagnoses made by pathologists in the course of their daily work, eliminating large manually annotated tasks that used to hinder the development and clinical practice of AI in pathology. And the system Song et al. have developed may be too expensive (Small hospital: $84, 000-$87, 000; Large hospital: $161, 000-$164, 000) for economically underdeveloped areas with a shortage of pathologists, limiting its promotion in primary hospitals to a certain extent. We designed a website based on our analysis, simple and easy to use, and all the users need is a computer with an Internet connection or even an Internet-connected phone or tablet. When a histological image is uploaded without any professional annotation, the webpage will show a brief result of the primary type and survival prediction. The website identifies suspicious areas, thus improving diagnostic accuracy in a limited amount time, which will prove particularly useful in areas with a shortage of pathologists and in improving the diagnosis performance of junior pathologists who are under training. There have been a number of AI studies focusing on GC . Most of these previous studies relied on endoscopy, and a small amount of selected medical radiologic technology, such as computed tomography. Moreover, previous studies generally employed a small sample size from a single centre, lacking effective proof to validate the robustness of models. We constructed models on much larger datasets from two different cohorts, and validated on another independent external validation set, greatly enhancing the universality of the diagnostic and survival models. In the meantime, images in the external validation were diagnosed by human pathologists. By comparing the diagnosis performance of our model with that of junior and expert pathologists, the accuracy and reliability of our model was further confirmed. For survival prediction, we designed MIL-GC algorithm in this study. The risk score generated by MIL-GC exhibited a distinguished performance in predicting OS among patients with GC, as reflected by the C-index of 0.728, 0.671, and 0.657 in the training, internal validation and external validation sets, respectively. Furthermore, we applied Cox regression analysis to determine whether the risk score generated by MIL-GC was an independent biomarker for the prediction of OS in GC patients. Fortunately, the risk score generated by MIL-GC remained strong in multivariate regression (HR = 2.976, P < 0.0001 in the internal validation set, and HR = 1.803, P = 0.043 in the external validation set) after the adjustment for established prognostic features, indicating that the risk score generated by MIL-GC will be a promising supplement to the established markers and help refine risk stratification among GC patients. A recent study predicting the outcome of GC from resected lymph node histopathology images also yielded meaningful results. Our study focused on the pathological histological features from the stomach tissue, while their study concentrated on the lymph node metastasis of GC. Both of these two studies able to make predictions about the prognosis of patients with GC. If the key points of the two researches could be combined in the future, we may achieve a more satisfactory performance in predicting GC patients’ prognosis. Furthermore, our webpage could conveniently provide predictions of patient prognosis, serving as an important reference for selection of adjuvant therapy after surgery in patients with GC. There are some limitations to our study. First, the survival time of GC individuals from the TGCA cohort was different from that in the RHWU cohort due to the progress of treatment. Therefore, it is not appropriate for us to use the TCGA or RHWU cohort as the training set and the other as the validation set. Hence, we mixed them together and randomly split them into the training or internal validation set. Furthermore, datasets we collected in the RHWU, TCGA and NHGRP cohorts were retrospective and thus suffered from inherent biases. In the future, we plan to conduct a prospective, randomized, multicentre clinical trial to validate the performance of precisely diagnosing GC and stratifying patients into high-risk and low-risk score groups to assist in selecting the suitable individualized adjuvant treatment regiments. In conclusion, we developed deep learning models to diagnose GC and predict the survival outcomes of GC patients by analyzing H&E-stained pathological images. To make our models more intuitive and easier to use, an online website ( https://baigao.github.io/Pathologic-Prognostic-Analysis/ ) based on developed algorithms was designed. Our models assist oncologists in the identification of GC and selection of appropriate treatment, thus reducing the physical and economic burdens of patients. W.D. was associated with Conceptualization, Funding acquisition, Resources, and Supervision of the study. B.H., S.T., N.Z., J.M., Y.L., P.H., B.D., and J.H. conducted Data curation. Z.H., C.Z., H.Z., and F.M. took charge of Formal analysis, Methodology and Software. B.H., S.T., N.Z., and J.M. performed the Investigation and Project administration. S.T., B.H., Z.H., C.Z., H.Z., and F.M. completed the Validation and Visualization. B.H. and S.T. completed Writing – original draft. W.D., B.H., S.T., F.L., M.J., and J.Z. conducted Writing – review & editing. All authors verified the underlying data and reviewed and approved the final manuscript. All authors have no conflicts of interest to disclose. |
Mental health literacy of reproductive age women: a qualitative study | 8931dd49-bd28-4dba-a91b-dadb49a62995 | 11740350 | Health Literacy[mh] | Currently, mental illness is one of the five main causes of disability, which accounts for more than 30% of all disabilities during life and creates a significant economic and social burden for the individuals, families and societies . The Global Burden of Mental Disease highlights around 45 million increase in Disability-Adjusted Life Years (DALYs) attributable to mental disorders worldwide (form 80 million to over 125 million). With this surge, mental disorders have moved into the top 10 significant causes of DALYs over the last three decades . For this reason, the management of mental disorders is one of the main priorities of health systems in different countries, and it is an issue that attracted the attention of many researchers and policy makers . Previous studies have reported that 418 million disability-adjusted life years (DALYs) could be attributed to mental disorders in 2019 (16% of global DALYs). Also the economic value associated with this burden is estimated at 5 trillion dollars . A growing body of literature also suggests that mental disorders are costly, both in direct medical costs of care, outpatient visits, and hospitalizations, and in indirect costs, such as lost income and productivity due to Disability, which may cause absenteeism . In one study, it was reported that DALYs due to mental disorders increase sharply with the onset of reproductive age. Also, DALYs differ significantly between genders and more disability-adjusted life years are reported in women than men . Around one in five women have a common mental health problem, such as depression and anxiety. While there can be many reasons why these develop, some risk factors affect many women . Women are more likely than men: to be cares, which can lead to stress, anxiety and isolation, to live in poverty which, along with concerns about personal safety and working mainly in the home, can lead to social isolation, to experience physical and sexual abuse, which can have a long-term impact on their mental health and to experience sexual violence, which can cause PTSD . When women find it hard to talk about difficult feelings, they tend to internalize them. This can lead to depression, eating disorders and self-harm . On the other hand, Life events and hormonal changes can affect women’s mental health. Having a baby is a life-changing event. For some women, it can trigger postnatal depression (after birth) and/or antenatal depression (during pregnancy). The term ‘perinatal depression’ covers both . While every woman’s experience of the menopause is different, many women find they have symptoms in addition to their periods stopping. These can include changes to mental health, such as mood swings, anxiety and feeling low . WHO has released the Comprehensive Mental Health Action Plan 2013–2020 and its next update is due by 2030. The goal of this action plan is to ensure that 80% of countries have at least three national and multispectral mental health promotion and prevention programs by 2030. These initiatives should target vulnerable groups using programs such as mental health awareness/anti-stigma or school-based mental health prevention and promotion . In this framework lies the concept of Mental Health Literacy (MHL), described by Jorm as “knowledge and beliefs about mental health problems that help to recognize, manage and prevent them” . This definition includes the capacity to recognize and inform about mental disorders and their risk factors and how to seek treatment and professional support . Women of reproductive age, in addition to their key role in the family, also play an essential role in society and the workforce . They have different duties related to their different roles that may endanger their mental health . Women bear a greater burden in many societies, as they are often expected to work and earn to support the family economically, while also taking care of household duties and chores . In addition, passing through various physiological processes such as puberty, menstruation, pregnancy, childbirth and menopause, as well as the possibility of greater risk of poverty, hunger and malnutrition, heavy workload and discrimination also affect their mental health . Some of these problems can be linked to limited mental health literacy . In a research (2023), 15.3% of young women reported that they experienced periodic poverty. Higher odds of poor mental health were estimated for women experiencing period poverty. a considerable number of young women living in an urban setting in a high-income country cannot afford menstrual products, and this may have an impact on their mental wellbeing . Mental health literacy includes several components, including (a) the ability to distinguish specific disorders from types of mental distress; b) Knowledge and belief about risk factors and causes. (c) Knowledge and beliefs about self-help interventions. (d) knowledge and beliefs about available professional help; (e) Attitudes that facilitate identification and appropriate help-seeking. and (f) knowledge about how to search for mental health information . It can be said that in addition to all aspects of general mental health literacy, due to the prevalence of specific mental health disorders and the transitional periods of women of reproductive age, special aspects of this special group are also needed. Based on this, considering the direct effect of mental health literacy on improving the mental health of women of reproductive age and evaluating the state of mental health literacy among women of reproductive age as key and effective people in the family and society, it seems necessary. Then, if necessary, focus on improving it can be considered. A systematic review of mental health literacy and women recognizes the importance of mental health literacy in women and at the same time emphasizes the very limited number of studies in this field. It has also been reported, women often report more mental health problems than men, and mental health problems are often more persistent among women, further highlighting the need for future studies on MHL in women. This systematic review also reported low consistency in measuring women’s mental health literacy and a lack of specific tools for measuring women’s mental health literacy . Mental disorders can affect men and women differently. Some disorders such as depression and anxiety are more common in women. There are also certain types of depression that are specific to women. Some women may experience symptoms of mental disorders during times of hormonal changes, such as perinatal depression, premenstrual dysphoric disorder, and menopause-related depression . Mental health concerns for women who experience domestic violence are also well documented. For instance, women who have experienced trauma such as sexual abuse and physical violence in childhood are three to four times more likely to experience depression as adults . As such determining the main dimensions of mental health literacy in reproductive age women is essential and can help in planning to protect and prevent psychological disorders among this population. Paying attention to the dimensions of mental health literacy in women of reproductive age, in designing mental health programs, can have an important effect on improving their quality of life. Thus, this study aimed to elucidate the concept and provide a better understanding of the topic through a qualitative approach. It was hoped that the findings could help to design and develop a specific measure that could assess mental health literacy in reproductive age women. Design and sampling This was a qualitative study and the data obtained was based on 14 in-depth semi-structured interviews with women of reproductive age (15–49 years) and six semi- structured interviews with formal service providers, in three provinces of Tehran, Alborz and Qazvin in Iran, in 2022–2023. For sampling, a purposive sampling method was used with maximum diversity according to age, education level, marital status and employment status. The inclusion criteria for the study were women of reproductive age (according to the World Health Organization standard in the age group of 15–49 years) , being literate in reading and writing, able to express words and willing to participate in the study. The formal service providers were included if they were engaged with given advice to women of reproductive age. Sampling continued until data saturation reached. That is, when no new codes and data were identified to add to the previous ones. Data collection Semi-structured face-to-face interviews were conducted to collect the data. After asking the main and opening questions of each interview, participants’ experiences were further explored during the interview using probing questions. All interviews were conducted by the first author (AS, a Ph.D. candidate in reproductive health) at a time and place preferred by the participants. As mentioned earlier in addition to women of reproductive age, the formal service providers including psychiatrists, psychologists and counselors were also interviewed, which helped to improve the depth of data collection. We tried to convey information we received from participants to formal service providers and started with general questions and ended with more specific contents. The answer to each of the questions lead to the next probing questions and then to deepen the data, questions such as: “Could you explain more about this?“, “If possible, explain about your feelings at that time”, “After “What happened to this experience you had” was asked. Also, during the interview, all the non-verbal behaviors of the participants, including body movements, facial expressions, tone of voice, emotional reactions, and expressions of emotions, were recorded by the researcher in order to better explain the interviews. The length of the interviews was ranged from 22 to 48 minutes, with mean value of 35 minutes. The data collection in this stage was from August 14 2022 to March 24, 2023. All interviews were recorded and field notes were taken with the consent of the participants. Since there was a little ambiguity in a number of interviews thus for clarity 4 additional interviews were conducted with participants 1, 2, 3, and 6. Examples of interview opening questions are given in Table . Data analysis Conventional (inductive) content analysis was used to analyze the data . Data analysis was done by the main investigator (AS) under the supervision of the research team. Inductive content analysis is an inductive process that involves iterative coding. First, the main investigator (AS) transcribed the interviews word for word and codes used to label the data was created during the coding process based on the actual content of the data set. In this research, the codes were identified by the researcher within the data itself, or as it is often said, “emerged” and the codes are or “emerged” from the data. The coding process was not done just once for each transcript, but was refined and then repeated based on comparisons between transcripts. Each transcript was coded several times with more detailed repetitions. Since the coding was inductive and by analyzing more transcripts, new aspects of the data were identified, this recoding was repeated and new codes were identified. It is very common for ideas identified in later transcripts to be present in earlier transcripts, perhaps in more indirect ways, but not recognized when those transcripts were first coded. Therefore, the researcher added to the list of codes and adapted it during the analysis process. The coding process, following the comparison, grouping and division of groups of codes, led to categories and subcategories. Participants’ quotes (initial codes) were grouped and summarized based on similarities, and dense codes were obtained. Then the dense codes were categorized based on the area and similarity and formed sub-categories. And sub-categories were formed based on the subject area of categories and then categories particle for themes. It is necessary to explain that the data obtained from the interviews of formal service providers were analyzed simultaneously with women of reproductive age. Finally, the dimensions and areas of the themes of mental health literacy in reproductive age women were determined. Trustworthiness According to the definition of Lincoln and Goba (1986), the well-known criteria for evaluating the validity and reliability of qualitative data include credibility, transferability, dependability and confirm ability were used . In the present study, to ensure the credibility of the findings, the coded text in the interview was given to the participants to confirm the extracted codes (Member Check) and if there was a contradiction, the necessary investigation was done and clarification was done with the help of the participants’ opinions. In addition, key informants including psychologists, psychiatrists and counselors were used in order to use their experiences. The interviews were analyzed by the research team. In this way, the problems in the interview process and the coding process were determined. The researcher tried to increase the credibility of the research by spending enough and long time to review the information and data, constantly rethinking the codes and classes and holding numerous meetings with the professors of the research team until reaching a consensus. Also, the integration of different methods of data collection, such as in-depth interviews, note-taking, daily notes, and the selection of participants with the maximum variety of other actions taken in this direction. In order to improve the dependability of the research, he used the precise introduction of the participants and the complete explanation of the data analysis steps to enable the audit of the research. Also, an attempt was made to guide the interview and coding process in line with the research question so that the research findings have a good logic and consistency. In order to improve the confirm ability, the researcher tried to provide a complete and step-by-step explanation of the different stages of the research, from data collection to the formation of sub-classes, classes and topics, so that the review of the research is possible for the readers. Also, the accuracy of the research process step by step, including the process of interviews, coding, extraction of classes and themes, was confirmed by the research team. Finally, in order to improve the transferability, it was tried to provide clear and complete explanations about the research field, experiences and events during the research so that other researchers are also able to judge the research process. Also, participants with maximum diversity in terms of age, level of education, occupation and marital status entered the research. This was a qualitative study and the data obtained was based on 14 in-depth semi-structured interviews with women of reproductive age (15–49 years) and six semi- structured interviews with formal service providers, in three provinces of Tehran, Alborz and Qazvin in Iran, in 2022–2023. For sampling, a purposive sampling method was used with maximum diversity according to age, education level, marital status and employment status. The inclusion criteria for the study were women of reproductive age (according to the World Health Organization standard in the age group of 15–49 years) , being literate in reading and writing, able to express words and willing to participate in the study. The formal service providers were included if they were engaged with given advice to women of reproductive age. Sampling continued until data saturation reached. That is, when no new codes and data were identified to add to the previous ones. Semi-structured face-to-face interviews were conducted to collect the data. After asking the main and opening questions of each interview, participants’ experiences were further explored during the interview using probing questions. All interviews were conducted by the first author (AS, a Ph.D. candidate in reproductive health) at a time and place preferred by the participants. As mentioned earlier in addition to women of reproductive age, the formal service providers including psychiatrists, psychologists and counselors were also interviewed, which helped to improve the depth of data collection. We tried to convey information we received from participants to formal service providers and started with general questions and ended with more specific contents. The answer to each of the questions lead to the next probing questions and then to deepen the data, questions such as: “Could you explain more about this?“, “If possible, explain about your feelings at that time”, “After “What happened to this experience you had” was asked. Also, during the interview, all the non-verbal behaviors of the participants, including body movements, facial expressions, tone of voice, emotional reactions, and expressions of emotions, were recorded by the researcher in order to better explain the interviews. The length of the interviews was ranged from 22 to 48 minutes, with mean value of 35 minutes. The data collection in this stage was from August 14 2022 to March 24, 2023. All interviews were recorded and field notes were taken with the consent of the participants. Since there was a little ambiguity in a number of interviews thus for clarity 4 additional interviews were conducted with participants 1, 2, 3, and 6. Examples of interview opening questions are given in Table . Conventional (inductive) content analysis was used to analyze the data . Data analysis was done by the main investigator (AS) under the supervision of the research team. Inductive content analysis is an inductive process that involves iterative coding. First, the main investigator (AS) transcribed the interviews word for word and codes used to label the data was created during the coding process based on the actual content of the data set. In this research, the codes were identified by the researcher within the data itself, or as it is often said, “emerged” and the codes are or “emerged” from the data. The coding process was not done just once for each transcript, but was refined and then repeated based on comparisons between transcripts. Each transcript was coded several times with more detailed repetitions. Since the coding was inductive and by analyzing more transcripts, new aspects of the data were identified, this recoding was repeated and new codes were identified. It is very common for ideas identified in later transcripts to be present in earlier transcripts, perhaps in more indirect ways, but not recognized when those transcripts were first coded. Therefore, the researcher added to the list of codes and adapted it during the analysis process. The coding process, following the comparison, grouping and division of groups of codes, led to categories and subcategories. Participants’ quotes (initial codes) were grouped and summarized based on similarities, and dense codes were obtained. Then the dense codes were categorized based on the area and similarity and formed sub-categories. And sub-categories were formed based on the subject area of categories and then categories particle for themes. It is necessary to explain that the data obtained from the interviews of formal service providers were analyzed simultaneously with women of reproductive age. Finally, the dimensions and areas of the themes of mental health literacy in reproductive age women were determined. According to the definition of Lincoln and Goba (1986), the well-known criteria for evaluating the validity and reliability of qualitative data include credibility, transferability, dependability and confirm ability were used . In the present study, to ensure the credibility of the findings, the coded text in the interview was given to the participants to confirm the extracted codes (Member Check) and if there was a contradiction, the necessary investigation was done and clarification was done with the help of the participants’ opinions. In addition, key informants including psychologists, psychiatrists and counselors were used in order to use their experiences. The interviews were analyzed by the research team. In this way, the problems in the interview process and the coding process were determined. The researcher tried to increase the credibility of the research by spending enough and long time to review the information and data, constantly rethinking the codes and classes and holding numerous meetings with the professors of the research team until reaching a consensus. Also, the integration of different methods of data collection, such as in-depth interviews, note-taking, daily notes, and the selection of participants with the maximum variety of other actions taken in this direction. In order to improve the dependability of the research, he used the precise introduction of the participants and the complete explanation of the data analysis steps to enable the audit of the research. Also, an attempt was made to guide the interview and coding process in line with the research question so that the research findings have a good logic and consistency. In order to improve the confirm ability, the researcher tried to provide a complete and step-by-step explanation of the different stages of the research, from data collection to the formation of sub-classes, classes and topics, so that the review of the research is possible for the readers. Also, the accuracy of the research process step by step, including the process of interviews, coding, extraction of classes and themes, was confirmed by the research team. Finally, in order to improve the transferability, it was tried to provide clear and complete explanations about the research field, experiences and events during the research so that other researchers are also able to judge the research process. Also, participants with maximum diversity in terms of age, level of education, occupation and marital status entered the research. Participants The characteristics of participants are shown in Table . The mean age of women was 31 ± 9.7 years ranging from 15 to 49. The educational level varied from secondary to higher education. Table also presents the characteristics of formal service providers. Overall findings After summarizing the semantic units, 965 quotations and 293 condensed codes were obtained. The main codes were obtained by examining the similarities and differences. These codes were placed in 34 different subcategories based on their similarities and differences. By comparing the subcategories, 9 main categories and finally four themes emerged (Fig. ). The detailed findings are presented as . Two main categories of “awareness of mental health information resources” and “understanding of mental health information” led to the explanation of the theme of knowledge of information sources and the ability to understand mental health. The main categories of “effective and positive communication with others, “management of external emotions” and “psychological self-management” led to the formation of the theme of the ability to use mental health information in women’s lives. The main categories of “awareness and beliefs about mental health” and “mental health autonomy” led to the formation of the theme of adapting to mood changes specific to reproductive age. The main categories of “self-help to promote mental health” and “asking for help and helping” led to the formation of the theme of action for mental health promotion. Knowledge of information sources and the ability to understand mental health Knowledge of mental health information sources Women of reproductive age considered it essential to be able to search for mental health information in order to improve their mental health. Most participants acknowledged that they have obtained the information needed to maintain their mental health from available printed and written sources, internet sources and virtual social networks. They emphasized the importance of being aware of the barriers and facilitators of seeking mental health information. “I read psychology books. Reading these books has made me stop doing a series of behaviors that were wrong. For example , getting stressed easily.” (Woman No. 2). Another participant said: “If the drug I am using is new and I don’t know anything about it , I will definitely read the brochure to see if it has any side effects.” (Woman No. 13). Women indicated that they usually search internet browsers such as Google, dedicated mental health sites, social networks such as Telegram, Instagram, Twitter, YouTube, as well as sites related to pharmaceutical information. “When I have a problem related to my mental health , I search on Google , if I see there is need for training , I search on YouTube and Twitter and watch some videos.” (Woman No. 5). A woman pointed out: “Because Instagram is visual , I get most of the information from there.” (Woman No. 8). The participants expressed that they obtained useful information about mental health from applications that even can suggest mental health disorders. “I have the ‘Calm’ program , it works on four skills: meditation , breathing , sleep , and relaxation , which has breathing programs to reduce stress , soothing music , and nature sounds for relaxation.” (Woman No. 14). Women said there are several sources for seeking information they need. “If I feel bad or suspect a disease , I talk to my friends , they guide me and tell me about their experiences.” (Woman No. 9). Service providers emphasized on obtaining mental health information from therapists. “Women of reproductive age can get advice from their doctor” (Service providers No. 3). Introduction of information sources by friends and family and the therapist, the existence of brief and simple and understandable materials for the public, the existence of educational pamphlets in offices and electronic versions of psychology books, were among the strengthening factors that played a significant role in obtaining information. “I have called the national counseling helpline several times when I had a problem with my partner and I got help , which was very good.” (Woman No. 13). Women also indicated that their doctors were a trusted source for information: “The most reliable way to seek information is to talk to your doctor who usually patiently explain everything to you.” (Woman No. 3). In order to obtain mental health information, the participants highlighted that they faced various obstacles and inhibiting factors such as limited access to the internet and slow internet speed, small font and non-Persian drug brochures, the extent of information, the lack of electronic versions of some psychology books and the lack of an official website. Mental health information provider mentioned. “In addition to the variety of information , the problem of the Internet is a big obstacle to get information.” (Woman No. 1). Understanding mental health information Early diagnosis of mental health disorders, diagnosis of mood disorders specific to the reproductive age, information validity and ability to use information were the topic for understanding mental health information. A participant said: “I know the symptoms of depression and I was able to understand that I have depression now with these symptoms and I have to do something for myself.” (Woman No. 6). In the current study, the participants stated about the experience of mental health disorders specific to women of reproductive age and how to diagnose them: “After I gave birth , I was in a very bad mood and at the same time the responsibilities of my child fell on me , I felt I’m not normal and there might be a problem , I asked my family for help.” (Woman No. 1). Another participant said: “During my period , I get anxiety and I’m all confused. On the other hand , the heartache and headache bother me a lot , but I know that it’s related to my hormones and I try to control it.” (Woman No. 11). In this regard, one of service providers stated: “If women recognize the disorders that are related to hormonal changes and their fertility , to know how far these symptoms are normal and when they become abnormal and they need help , it will be effective in treating them.” ( Service provider No. 20) . Women of reproductive age trusted the information obtained from doctors and specialists the most, and also considered the information obtained from books, scientific sites, and virtual networks related to therapists to be valid. “First of all , I look at sites that are safe. For example , if it is a psychologist or psychiatrist , or about drugs , I look at sites that are only about drug information.” (Woman No. 9). Women evaluated information by performing actions such as comparing information from different sources, comparing information with the person’s previous experience and previous information, and asking questions from experts. A woman said: “Regarding the drug , I checked the pharmacy website , which was written according to the drug brochure. I also checked several sites.” (Woman No. 3). Women stated that they used the mental health information obtained to improve family relationships, control mental health disorders and followed up treatment, self-care and positive thinking and problem solving. “I use the information in different situations that occur to me. For example , I might be arguing with someone and feel very angry , then I remember to take a few deep breaths and count to ten , then react.” (Woman No. 2). Ability to use mental health information in women’s lives Women of reproductive age need strategies to using the information gained in women’s life. Apparently women indicated three strategies included: positive communication with others, Management of external emotions, and psychological self-management. Effective and positive communication with others Women used effective and positive communication with others to prevent mental health disorders in the two areas effective communication with others and maintaining favorable family relationships, feeling responsible for family members. “I tried to learn more about how to interact with different people with different lifestyles and minds , I tried to improve my knowledge over time.” (Woman No. 8). A woman highlighted: “ I used the information I gained to improve my family relationships.” (Woman No. 3). Management of external emotions Women considered the ability to control anger, control anxiety and stress, adapt to changes and cope with life’s adversities, control emotions related to the reproductive period, and the ability to be influential in society as important mental health facets. “ A therapist taught me some anger management skills. I use them , and I have a bit more control over myself.” (Woman No. 10) . The ability to adapt to changes and cope with adversities in life, such as adapting to job problems, the ability to cope with the bereavement of loved ones, the ability to cope with responsibilities related to children, adapting to economic problems, correct coping styles and having knowledge of impulse control in health is impressive. “After my father’s death , it was very difficult at first , but gradually we used to the situation.” (Woman No. 11). Women were concerned with the impact of gender roles and responsibilities, especially the motherhood role, unwanted pregnancy, infertility, concern for the health of the fetus during pregnancy, mood changes after childbirth and around menstruation, and the impact of abortion and loss. They also mentioned about mental health and communication, as well as the need to adapt to the emotions caused by these and stated that: “ After giving birth , I couldn’t cope with the baby’s chores at all. I thought. I felt very bad because of this. I was bored , I was crying all the time , I threw out all my pre-pregnancy clothes.” (Woman No. 3). Another woman indicated that: “After abortion , in addition to the physical complications I experienced a very bad post-abortion period.” (Woman No. 13). Psychological self-management Women of reproductive age reported that they have skills such as self-knowledge, awareness, self-acceptance, and ultimately maintaining dynamism and vitality in life in the direction of psychological self-management. “I know what kind of people I like to hang out with , what kind of sports I like , I know who I want to marry , how to control my anger” (Woman No. 4). A participant pointed out: “For my mental health , I mostly watch movies , listen to music , and go out with my friends.” (Woman No. 5). Also a woman said that: “To control myself , I try to use relaxation techniques , and to control the environment , it depends on the situation. For example , in my work environment , I try to maintain a good relationship with my colleagues. In general , I try to manage the situation somehow. take my hand.” (Woman No. 2). Adapting to mood changes specific to reproductive age This topic is formed from the following categories: “awareness and beliefs about mental health” and “mental health self-esteem”. According to the experience of the participants, a number of factors such as loneliness, family problems, poverty and economic problems, problems in emotional relationships or relationships outside the family framework, gender discrimination and experience of violence, etc. were considered as risk factors for mental health disorders. Awareness and beliefs about mental health “It is a heavy burden to understand that your child has a disease that cannot be cured. It is very difficult. I suffered from depression and was very nervous. It was very tense , both my husband and I were very depressed and sad.” (Woman No. 10). Another participant said that: “My husband and I studied together in the same university , after graduation , my husband went to work very easily and I couldn’t find a job. Wherever I went looking for an architect , I was very disappointed from life.” (Woman No. 9). Mental health autonomy Participants acknowledged things such as fear of drugs for mental health disorders, uncertainty about the effectiveness of non-drug treatments, the stigma of mental illness, lack of recognition of the symptoms of mental health disorders by the person and those around him, and the high cost of services as a deterrent to seeking help. Also women of reproductive age emphasized the recognition of symptoms of mental health disorders by those around them as a facilitating factor for seeking help. “I don’t take the medicine for my sleep disorder that the psychiatrist gave me , I think it’s addictive.” (Woman No. 2). Or a women said that: “After Corona , there were people around me who didn’t take my panic symptoms seriously , I said hey , I’m sorry , I have anxiety and I’m stressed , they said it’s nothing , don’t think about it , don’t be afraid , you’re fine now , or they even laughed to see that once Corona No matter how scared you are , not everyone will die.” (Woman No. 9). Women emphasized on reducing the stigma if mental health information is increased. “The biggest effect of reading psychology books is that changed my view of people who have these problems.” (Woman No. 1). “When you understand that someone has a mental problem , it affects one’s perspective. I myself usually deal with them differently than others.” (Woman No. 3). Most participants admitted that they take self-medication measures to protect their mental health during the hormonal disturbances of the reproductive period. They also stated that if they see abnormal symptoms and symptoms worsen, they will refer to therapists for more treatment and care. “During my period and a few days before , I get anxious and agitated. For my anxiety , the doctor prescribed Inderal to reduce my heart rate and anxiety.” (Woman No. 7). Action for mental health promotion Participants stated that they take steps to improve their mental health. This topic consists of self-help to improve mental health and help seeking and giving help. Women admitted that in case of abuse, they will take the necessary legal and psychological measures. They also emphasized on following up the necessary treatments in case of mental health disorders to improve their mental health. Self-help to promote mental health “I try to have a healthy lifestyle , for example , I exercise and use vitamins before my period to reduce my symptoms. Now I pay attention to my sleep , and I feel relax.” (Woman No. 2). Another participant said: “When I was dealing with infertility , I was trying to calm myself down with anxiety and stress control practices.” (Woman No. 6). Women stated that they use the help of professionals including psychiatrists, psychologists, counselors, doctors, comprehensive health service centers, social workers and psychoanalysts in case of mental health disorders. “When I was a student , one of my friends had a seizure in the dormitory. After that , every sound I heard , I got anxious. I went to the doctor , and he gave me Inderal and Alprazolam.” (Woman No. 2). One-woman state that: “After giving birth , I had developed mood problems and depression , all my time was spent on the baby. I only got help from my family and friends and they supported me” (Woman No. 1). Asking for help and helping Also participants emphasized on the ability to recognize the symptoms of mental health disorders in the people around them and help them to find a solution or lead them to start the treatment process and follow it. “I knew the symptoms of my mother’s panic. When this happens , we would keep the house quiet and bring her medicine and we wouldn’t get on her nerves. If there is a movie about dying or someone we knew had died , we don’t tell her. Because she was feeling bad.” (Woman No. 11). The characteristics of participants are shown in Table . The mean age of women was 31 ± 9.7 years ranging from 15 to 49. The educational level varied from secondary to higher education. Table also presents the characteristics of formal service providers. After summarizing the semantic units, 965 quotations and 293 condensed codes were obtained. The main codes were obtained by examining the similarities and differences. These codes were placed in 34 different subcategories based on their similarities and differences. By comparing the subcategories, 9 main categories and finally four themes emerged (Fig. ). The detailed findings are presented as . Two main categories of “awareness of mental health information resources” and “understanding of mental health information” led to the explanation of the theme of knowledge of information sources and the ability to understand mental health. The main categories of “effective and positive communication with others, “management of external emotions” and “psychological self-management” led to the formation of the theme of the ability to use mental health information in women’s lives. The main categories of “awareness and beliefs about mental health” and “mental health autonomy” led to the formation of the theme of adapting to mood changes specific to reproductive age. The main categories of “self-help to promote mental health” and “asking for help and helping” led to the formation of the theme of action for mental health promotion. Knowledge of mental health information sources Women of reproductive age considered it essential to be able to search for mental health information in order to improve their mental health. Most participants acknowledged that they have obtained the information needed to maintain their mental health from available printed and written sources, internet sources and virtual social networks. They emphasized the importance of being aware of the barriers and facilitators of seeking mental health information. “I read psychology books. Reading these books has made me stop doing a series of behaviors that were wrong. For example , getting stressed easily.” (Woman No. 2). Another participant said: “If the drug I am using is new and I don’t know anything about it , I will definitely read the brochure to see if it has any side effects.” (Woman No. 13). Women indicated that they usually search internet browsers such as Google, dedicated mental health sites, social networks such as Telegram, Instagram, Twitter, YouTube, as well as sites related to pharmaceutical information. “When I have a problem related to my mental health , I search on Google , if I see there is need for training , I search on YouTube and Twitter and watch some videos.” (Woman No. 5). A woman pointed out: “Because Instagram is visual , I get most of the information from there.” (Woman No. 8). The participants expressed that they obtained useful information about mental health from applications that even can suggest mental health disorders. “I have the ‘Calm’ program , it works on four skills: meditation , breathing , sleep , and relaxation , which has breathing programs to reduce stress , soothing music , and nature sounds for relaxation.” (Woman No. 14). Women said there are several sources for seeking information they need. “If I feel bad or suspect a disease , I talk to my friends , they guide me and tell me about their experiences.” (Woman No. 9). Service providers emphasized on obtaining mental health information from therapists. “Women of reproductive age can get advice from their doctor” (Service providers No. 3). Introduction of information sources by friends and family and the therapist, the existence of brief and simple and understandable materials for the public, the existence of educational pamphlets in offices and electronic versions of psychology books, were among the strengthening factors that played a significant role in obtaining information. “I have called the national counseling helpline several times when I had a problem with my partner and I got help , which was very good.” (Woman No. 13). Women also indicated that their doctors were a trusted source for information: “The most reliable way to seek information is to talk to your doctor who usually patiently explain everything to you.” (Woman No. 3). In order to obtain mental health information, the participants highlighted that they faced various obstacles and inhibiting factors such as limited access to the internet and slow internet speed, small font and non-Persian drug brochures, the extent of information, the lack of electronic versions of some psychology books and the lack of an official website. Mental health information provider mentioned. “In addition to the variety of information , the problem of the Internet is a big obstacle to get information.” (Woman No. 1). Understanding mental health information Early diagnosis of mental health disorders, diagnosis of mood disorders specific to the reproductive age, information validity and ability to use information were the topic for understanding mental health information. A participant said: “I know the symptoms of depression and I was able to understand that I have depression now with these symptoms and I have to do something for myself.” (Woman No. 6). In the current study, the participants stated about the experience of mental health disorders specific to women of reproductive age and how to diagnose them: “After I gave birth , I was in a very bad mood and at the same time the responsibilities of my child fell on me , I felt I’m not normal and there might be a problem , I asked my family for help.” (Woman No. 1). Another participant said: “During my period , I get anxiety and I’m all confused. On the other hand , the heartache and headache bother me a lot , but I know that it’s related to my hormones and I try to control it.” (Woman No. 11). In this regard, one of service providers stated: “If women recognize the disorders that are related to hormonal changes and their fertility , to know how far these symptoms are normal and when they become abnormal and they need help , it will be effective in treating them.” ( Service provider No. 20) . Women of reproductive age trusted the information obtained from doctors and specialists the most, and also considered the information obtained from books, scientific sites, and virtual networks related to therapists to be valid. “First of all , I look at sites that are safe. For example , if it is a psychologist or psychiatrist , or about drugs , I look at sites that are only about drug information.” (Woman No. 9). Women evaluated information by performing actions such as comparing information from different sources, comparing information with the person’s previous experience and previous information, and asking questions from experts. A woman said: “Regarding the drug , I checked the pharmacy website , which was written according to the drug brochure. I also checked several sites.” (Woman No. 3). Women stated that they used the mental health information obtained to improve family relationships, control mental health disorders and followed up treatment, self-care and positive thinking and problem solving. “I use the information in different situations that occur to me. For example , I might be arguing with someone and feel very angry , then I remember to take a few deep breaths and count to ten , then react.” (Woman No. 2). Women of reproductive age considered it essential to be able to search for mental health information in order to improve their mental health. Most participants acknowledged that they have obtained the information needed to maintain their mental health from available printed and written sources, internet sources and virtual social networks. They emphasized the importance of being aware of the barriers and facilitators of seeking mental health information. “I read psychology books. Reading these books has made me stop doing a series of behaviors that were wrong. For example , getting stressed easily.” (Woman No. 2). Another participant said: “If the drug I am using is new and I don’t know anything about it , I will definitely read the brochure to see if it has any side effects.” (Woman No. 13). Women indicated that they usually search internet browsers such as Google, dedicated mental health sites, social networks such as Telegram, Instagram, Twitter, YouTube, as well as sites related to pharmaceutical information. “When I have a problem related to my mental health , I search on Google , if I see there is need for training , I search on YouTube and Twitter and watch some videos.” (Woman No. 5). A woman pointed out: “Because Instagram is visual , I get most of the information from there.” (Woman No. 8). The participants expressed that they obtained useful information about mental health from applications that even can suggest mental health disorders. “I have the ‘Calm’ program , it works on four skills: meditation , breathing , sleep , and relaxation , which has breathing programs to reduce stress , soothing music , and nature sounds for relaxation.” (Woman No. 14). Women said there are several sources for seeking information they need. “If I feel bad or suspect a disease , I talk to my friends , they guide me and tell me about their experiences.” (Woman No. 9). Service providers emphasized on obtaining mental health information from therapists. “Women of reproductive age can get advice from their doctor” (Service providers No. 3). Introduction of information sources by friends and family and the therapist, the existence of brief and simple and understandable materials for the public, the existence of educational pamphlets in offices and electronic versions of psychology books, were among the strengthening factors that played a significant role in obtaining information. “I have called the national counseling helpline several times when I had a problem with my partner and I got help , which was very good.” (Woman No. 13). Women also indicated that their doctors were a trusted source for information: “The most reliable way to seek information is to talk to your doctor who usually patiently explain everything to you.” (Woman No. 3). In order to obtain mental health information, the participants highlighted that they faced various obstacles and inhibiting factors such as limited access to the internet and slow internet speed, small font and non-Persian drug brochures, the extent of information, the lack of electronic versions of some psychology books and the lack of an official website. Mental health information provider mentioned. “In addition to the variety of information , the problem of the Internet is a big obstacle to get information.” (Woman No. 1). Early diagnosis of mental health disorders, diagnosis of mood disorders specific to the reproductive age, information validity and ability to use information were the topic for understanding mental health information. A participant said: “I know the symptoms of depression and I was able to understand that I have depression now with these symptoms and I have to do something for myself.” (Woman No. 6). In the current study, the participants stated about the experience of mental health disorders specific to women of reproductive age and how to diagnose them: “After I gave birth , I was in a very bad mood and at the same time the responsibilities of my child fell on me , I felt I’m not normal and there might be a problem , I asked my family for help.” (Woman No. 1). Another participant said: “During my period , I get anxiety and I’m all confused. On the other hand , the heartache and headache bother me a lot , but I know that it’s related to my hormones and I try to control it.” (Woman No. 11). In this regard, one of service providers stated: “If women recognize the disorders that are related to hormonal changes and their fertility , to know how far these symptoms are normal and when they become abnormal and they need help , it will be effective in treating them.” ( Service provider No. 20) . Women of reproductive age trusted the information obtained from doctors and specialists the most, and also considered the information obtained from books, scientific sites, and virtual networks related to therapists to be valid. “First of all , I look at sites that are safe. For example , if it is a psychologist or psychiatrist , or about drugs , I look at sites that are only about drug information.” (Woman No. 9). Women evaluated information by performing actions such as comparing information from different sources, comparing information with the person’s previous experience and previous information, and asking questions from experts. A woman said: “Regarding the drug , I checked the pharmacy website , which was written according to the drug brochure. I also checked several sites.” (Woman No. 3). Women stated that they used the mental health information obtained to improve family relationships, control mental health disorders and followed up treatment, self-care and positive thinking and problem solving. “I use the information in different situations that occur to me. For example , I might be arguing with someone and feel very angry , then I remember to take a few deep breaths and count to ten , then react.” (Woman No. 2). Women of reproductive age need strategies to using the information gained in women’s life. Apparently women indicated three strategies included: positive communication with others, Management of external emotions, and psychological self-management. Effective and positive communication with others Women used effective and positive communication with others to prevent mental health disorders in the two areas effective communication with others and maintaining favorable family relationships, feeling responsible for family members. “I tried to learn more about how to interact with different people with different lifestyles and minds , I tried to improve my knowledge over time.” (Woman No. 8). A woman highlighted: “ I used the information I gained to improve my family relationships.” (Woman No. 3). Management of external emotions Women considered the ability to control anger, control anxiety and stress, adapt to changes and cope with life’s adversities, control emotions related to the reproductive period, and the ability to be influential in society as important mental health facets. “ A therapist taught me some anger management skills. I use them , and I have a bit more control over myself.” (Woman No. 10) . The ability to adapt to changes and cope with adversities in life, such as adapting to job problems, the ability to cope with the bereavement of loved ones, the ability to cope with responsibilities related to children, adapting to economic problems, correct coping styles and having knowledge of impulse control in health is impressive. “After my father’s death , it was very difficult at first , but gradually we used to the situation.” (Woman No. 11). Women were concerned with the impact of gender roles and responsibilities, especially the motherhood role, unwanted pregnancy, infertility, concern for the health of the fetus during pregnancy, mood changes after childbirth and around menstruation, and the impact of abortion and loss. They also mentioned about mental health and communication, as well as the need to adapt to the emotions caused by these and stated that: “ After giving birth , I couldn’t cope with the baby’s chores at all. I thought. I felt very bad because of this. I was bored , I was crying all the time , I threw out all my pre-pregnancy clothes.” (Woman No. 3). Another woman indicated that: “After abortion , in addition to the physical complications I experienced a very bad post-abortion period.” (Woman No. 13). Psychological self-management Women of reproductive age reported that they have skills such as self-knowledge, awareness, self-acceptance, and ultimately maintaining dynamism and vitality in life in the direction of psychological self-management. “I know what kind of people I like to hang out with , what kind of sports I like , I know who I want to marry , how to control my anger” (Woman No. 4). A participant pointed out: “For my mental health , I mostly watch movies , listen to music , and go out with my friends.” (Woman No. 5). Also a woman said that: “To control myself , I try to use relaxation techniques , and to control the environment , it depends on the situation. For example , in my work environment , I try to maintain a good relationship with my colleagues. In general , I try to manage the situation somehow. take my hand.” (Woman No. 2). Women used effective and positive communication with others to prevent mental health disorders in the two areas effective communication with others and maintaining favorable family relationships, feeling responsible for family members. “I tried to learn more about how to interact with different people with different lifestyles and minds , I tried to improve my knowledge over time.” (Woman No. 8). A woman highlighted: “ I used the information I gained to improve my family relationships.” (Woman No. 3). Women considered the ability to control anger, control anxiety and stress, adapt to changes and cope with life’s adversities, control emotions related to the reproductive period, and the ability to be influential in society as important mental health facets. “ A therapist taught me some anger management skills. I use them , and I have a bit more control over myself.” (Woman No. 10) . The ability to adapt to changes and cope with adversities in life, such as adapting to job problems, the ability to cope with the bereavement of loved ones, the ability to cope with responsibilities related to children, adapting to economic problems, correct coping styles and having knowledge of impulse control in health is impressive. “After my father’s death , it was very difficult at first , but gradually we used to the situation.” (Woman No. 11). Women were concerned with the impact of gender roles and responsibilities, especially the motherhood role, unwanted pregnancy, infertility, concern for the health of the fetus during pregnancy, mood changes after childbirth and around menstruation, and the impact of abortion and loss. They also mentioned about mental health and communication, as well as the need to adapt to the emotions caused by these and stated that: “ After giving birth , I couldn’t cope with the baby’s chores at all. I thought. I felt very bad because of this. I was bored , I was crying all the time , I threw out all my pre-pregnancy clothes.” (Woman No. 3). Another woman indicated that: “After abortion , in addition to the physical complications I experienced a very bad post-abortion period.” (Woman No. 13). Women of reproductive age reported that they have skills such as self-knowledge, awareness, self-acceptance, and ultimately maintaining dynamism and vitality in life in the direction of psychological self-management. “I know what kind of people I like to hang out with , what kind of sports I like , I know who I want to marry , how to control my anger” (Woman No. 4). A participant pointed out: “For my mental health , I mostly watch movies , listen to music , and go out with my friends.” (Woman No. 5). Also a woman said that: “To control myself , I try to use relaxation techniques , and to control the environment , it depends on the situation. For example , in my work environment , I try to maintain a good relationship with my colleagues. In general , I try to manage the situation somehow. take my hand.” (Woman No. 2). This topic is formed from the following categories: “awareness and beliefs about mental health” and “mental health self-esteem”. According to the experience of the participants, a number of factors such as loneliness, family problems, poverty and economic problems, problems in emotional relationships or relationships outside the family framework, gender discrimination and experience of violence, etc. were considered as risk factors for mental health disorders. Awareness and beliefs about mental health “It is a heavy burden to understand that your child has a disease that cannot be cured. It is very difficult. I suffered from depression and was very nervous. It was very tense , both my husband and I were very depressed and sad.” (Woman No. 10). Another participant said that: “My husband and I studied together in the same university , after graduation , my husband went to work very easily and I couldn’t find a job. Wherever I went looking for an architect , I was very disappointed from life.” (Woman No. 9). Mental health autonomy Participants acknowledged things such as fear of drugs for mental health disorders, uncertainty about the effectiveness of non-drug treatments, the stigma of mental illness, lack of recognition of the symptoms of mental health disorders by the person and those around him, and the high cost of services as a deterrent to seeking help. Also women of reproductive age emphasized the recognition of symptoms of mental health disorders by those around them as a facilitating factor for seeking help. “I don’t take the medicine for my sleep disorder that the psychiatrist gave me , I think it’s addictive.” (Woman No. 2). Or a women said that: “After Corona , there were people around me who didn’t take my panic symptoms seriously , I said hey , I’m sorry , I have anxiety and I’m stressed , they said it’s nothing , don’t think about it , don’t be afraid , you’re fine now , or they even laughed to see that once Corona No matter how scared you are , not everyone will die.” (Woman No. 9). Women emphasized on reducing the stigma if mental health information is increased. “The biggest effect of reading psychology books is that changed my view of people who have these problems.” (Woman No. 1). “When you understand that someone has a mental problem , it affects one’s perspective. I myself usually deal with them differently than others.” (Woman No. 3). Most participants admitted that they take self-medication measures to protect their mental health during the hormonal disturbances of the reproductive period. They also stated that if they see abnormal symptoms and symptoms worsen, they will refer to therapists for more treatment and care. “During my period and a few days before , I get anxious and agitated. For my anxiety , the doctor prescribed Inderal to reduce my heart rate and anxiety.” (Woman No. 7). “It is a heavy burden to understand that your child has a disease that cannot be cured. It is very difficult. I suffered from depression and was very nervous. It was very tense , both my husband and I were very depressed and sad.” (Woman No. 10). Another participant said that: “My husband and I studied together in the same university , after graduation , my husband went to work very easily and I couldn’t find a job. Wherever I went looking for an architect , I was very disappointed from life.” (Woman No. 9). Participants acknowledged things such as fear of drugs for mental health disorders, uncertainty about the effectiveness of non-drug treatments, the stigma of mental illness, lack of recognition of the symptoms of mental health disorders by the person and those around him, and the high cost of services as a deterrent to seeking help. Also women of reproductive age emphasized the recognition of symptoms of mental health disorders by those around them as a facilitating factor for seeking help. “I don’t take the medicine for my sleep disorder that the psychiatrist gave me , I think it’s addictive.” (Woman No. 2). Or a women said that: “After Corona , there were people around me who didn’t take my panic symptoms seriously , I said hey , I’m sorry , I have anxiety and I’m stressed , they said it’s nothing , don’t think about it , don’t be afraid , you’re fine now , or they even laughed to see that once Corona No matter how scared you are , not everyone will die.” (Woman No. 9). Women emphasized on reducing the stigma if mental health information is increased. “The biggest effect of reading psychology books is that changed my view of people who have these problems.” (Woman No. 1). “When you understand that someone has a mental problem , it affects one’s perspective. I myself usually deal with them differently than others.” (Woman No. 3). Most participants admitted that they take self-medication measures to protect their mental health during the hormonal disturbances of the reproductive period. They also stated that if they see abnormal symptoms and symptoms worsen, they will refer to therapists for more treatment and care. “During my period and a few days before , I get anxious and agitated. For my anxiety , the doctor prescribed Inderal to reduce my heart rate and anxiety.” (Woman No. 7). Participants stated that they take steps to improve their mental health. This topic consists of self-help to improve mental health and help seeking and giving help. Women admitted that in case of abuse, they will take the necessary legal and psychological measures. They also emphasized on following up the necessary treatments in case of mental health disorders to improve their mental health. Self-help to promote mental health “I try to have a healthy lifestyle , for example , I exercise and use vitamins before my period to reduce my symptoms. Now I pay attention to my sleep , and I feel relax.” (Woman No. 2). Another participant said: “When I was dealing with infertility , I was trying to calm myself down with anxiety and stress control practices.” (Woman No. 6). Women stated that they use the help of professionals including psychiatrists, psychologists, counselors, doctors, comprehensive health service centers, social workers and psychoanalysts in case of mental health disorders. “When I was a student , one of my friends had a seizure in the dormitory. After that , every sound I heard , I got anxious. I went to the doctor , and he gave me Inderal and Alprazolam.” (Woman No. 2). One-woman state that: “After giving birth , I had developed mood problems and depression , all my time was spent on the baby. I only got help from my family and friends and they supported me” (Woman No. 1). Asking for help and helping Also participants emphasized on the ability to recognize the symptoms of mental health disorders in the people around them and help them to find a solution or lead them to start the treatment process and follow it. “I knew the symptoms of my mother’s panic. When this happens , we would keep the house quiet and bring her medicine and we wouldn’t get on her nerves. If there is a movie about dying or someone we knew had died , we don’t tell her. Because she was feeling bad.” (Woman No. 11). “I try to have a healthy lifestyle , for example , I exercise and use vitamins before my period to reduce my symptoms. Now I pay attention to my sleep , and I feel relax.” (Woman No. 2). Another participant said: “When I was dealing with infertility , I was trying to calm myself down with anxiety and stress control practices.” (Woman No. 6). Women stated that they use the help of professionals including psychiatrists, psychologists, counselors, doctors, comprehensive health service centers, social workers and psychoanalysts in case of mental health disorders. “When I was a student , one of my friends had a seizure in the dormitory. After that , every sound I heard , I got anxious. I went to the doctor , and he gave me Inderal and Alprazolam.” (Woman No. 2). One-woman state that: “After giving birth , I had developed mood problems and depression , all my time was spent on the baby. I only got help from my family and friends and they supported me” (Woman No. 1). Also participants emphasized on the ability to recognize the symptoms of mental health disorders in the people around them and help them to find a solution or lead them to start the treatment process and follow it. “I knew the symptoms of my mother’s panic. When this happens , we would keep the house quiet and bring her medicine and we wouldn’t get on her nerves. If there is a movie about dying or someone we knew had died , we don’t tell her. Because she was feeling bad.” (Woman No. 11). The current study revealed four general themes including (a) knowledge of information sources and the ability to understand mental health; (b) ability to use mental health information in women’s lives; (c) adapting to mood changes specific to reproductive age, and (d) action for mental health promotion. These themes are presented individually with enlightening categories. Quadrilateral finding, they present an interesting quadruple interaction that provides insight into practical translations of reproductive-age women’s lives. Each topic will be discussed separately. The findings suggested that mental health literacy in reproductive age women includes women’s ability to overcome and solve mental health information needs by mastering the search for mental health information and ability to understand and apply information on mental health. They believed that such abilities can help to achieve prevention of mental disorders especially mood disorders specific to the reproductive age, as well as maintaining and promoting mental health and correct management of physiological mood changes related to hormonal instabilities specific to the reproductive age. The results of the present study identified new aspects of mental health literacy that are specific to women of reproductive age. Emphasizing the lack of similar research in the field of the concept of mental health literacy from the empirical perspective of reproductive ages, for the purpose of comparison, we will discuss the limited studies that are somehow related to the present research. Based on the results of the present study, in women of childbearing age, knowledge of information sources and the ability to understand mental health includes mastery of searching for mental health information and the ability to understand mental health information. Health information is used by people for various reasons. The high variety of information, the rapid development of science, health-related concerns, receiving disease prevention information has made people follow this information in multiple communication channels. Knowledge of resources and the ability to search for mental health information in reproductive age women, in the field of knowledge of printed and written sources of mental health information, knowledge of internet resources and social networks of mental health, knowledge of mental health software, ability to access mental health information from human sources, knowledge of the facilitating and strengthening factors as well as the obstacles and barriers to searching for mental health information and subsequently in the field of applying the information obtained, including early diagnosis of mental health disorders, diagnosis of mood disorders specific to the reproductive period, distinguishing the validity of information and the ability to use it. In this regard, woman number 6 stated that “in order to see what depression is , I used to search on Google and read articles. I also joined a telegram group , which is very good , all the women who have this problem are members there and talk about their experiences , and anyone who has any information writes about it , or if there is a question or if I am stressed about something , you are there. I ask that group”. A study reported that people searched the most on mental health topics in search engines such as Google, and the most searched item was mental health services available near the person and based on geographical conditions. The same study indicated that search for services based on the proximity of their location was more in women than in men . Similarly, a qualitative study showed that in a mental health website, factors that are considered trustworthy include relevant design, high quality, and the credibility of website creators, while unreliable factors included privacy policy and cookie settings . Rajabi et al.‘s research (2023) showed the positive effect of electronic health literacy and the use of electronic resources on people’s mental health . In the present study, the ability to use mental health information in women’s lives included the use of external emotion management skills and psychological self-management, which will ultimately lead to creating and maintaining effective and positive relationships with others. The ability to use mental health information in the life of women of reproductive age, including mastering the control of relationships and effective communication with others, maintaining favorable family relationships and feeling responsible for the family, managing emotions, including controlling anger, anxiety, stress, and coping with new developments. And the adversities in life and emotions are related to hormonal changes and changes in the roles of the reproductive period and ultimately influencing the society, and also with self-knowledge, awareness of self-relief methods and self-acceptance, maintaining dynamism and vitality in life and psychological self-management and self-control and environment. Defined under self-control. Investigating the effect of stress management intervention including cultural components on the levels of stress biomarkers and mental health indicators among women, the results showed the positive effect of the interventions and the reduction of cortisol levels and perceived stress . In this regard, woman number 9 stated: “The counselor I go to has taught me some anger management skills , and I use them. My control over myself has increased a bit. Before , I used to yell and scream , but now I am a bit better. For example , when I get angry , I close my eyes and count to ten , or I go to drink a glass of water , or , for example , I sit on the balcony for a few minutes to get some fresh air”. In the review research, the concepts of self-management, self-care and self-help were described and strategies or techniques related to these concepts were depicted for young people with emotional problems. In this research, the results showed the positive effect of self-management, self-care and self-help on the participants’ mental health . Woman number 12 stated: “Since I got help and practiced , I can control myself better than before when I am angry or stressed , I can say that I know myself more than before and have control over myself”. In a research, maternal experiences for risk management in pregnancy were formed in 9 subcategories and three categories: emotional excitement including: “feelings of worry and despair”, “joy in the shadow of hope and optimism”, “momentary impulses and emotions” and “immobility and helplessness”; Self-thinking includes: “Active analysis of ways to adjust risk”, “Cognitive denial” and “Indifference to risk”. In addition, there was evidence of actions such as: “problem-oriented and rational actions” and “avoidance and ineffective participation”. The results of this research showed that risk management experiences of pregnant women with high-risk pregnancies include a wide range of positive and negative emotions, effective and ineffective thoughts and behaviors. Cooperation between mothers and midwives/obstetricians by providing high quality risk management advice can lead to choosing effective risk management strategies and thus improving mental health . Woman number 3 stated: “I found out that my friend’s son has autism during pregnancy , and I was stressed until the end of my pregnancy so that my child would not have problems , especially since it was unplanned and I had not followed many things. After giving birth , I felt relieved about the baby , but once I came to my senses , I saw that I had given birth and the baby was in my arms , but my body was so fat that everywhere I went , they thought I was pregnant. I was very sick because of this. I was bored , I kept crying. On the one hand , I had a guilty conscience that I could not reach my baby well”. In the present study, adaptation to mood changes specific to reproductive age included improving awareness and beliefs about mental health and self-reliance in order to maintain mental health. In fact, adapting to the mood changes specific to the reproductive age, including recognizing the real causes and factors of mental health disorders, attitudes that hinder the receipt of mental health services, reducing the stigma of mental illness, taking action to treat mood disorders during the reproductive period and protecting oneself from abuse and abuse. Perinatal mental illness is often not diagnosed and treated, and early identification of risk factors can help women receive timely intervention to reduce maternal and child co-morbidities . Also, evidence suggests a significant psychosocial burden associated with infertility, which has been described as a stigmatized condition that makes it difficult for individuals to disclose information . In a research aimed at investigating the psychological impact of infertility, it was shown that patients may feel embarrassed and vulnerable. Effective psychosocial care is critical to help reduce distress and worry about fertility interventions and outcomes and improve patient well-being and quality of life. While fertility treatment may not enable all patients to conceive, available psychological care is essential for patients to have a healthy experience of their fertility journey . Woman number 6 stated: “We had financial problems , we were in debt for the whole house , at the same time , there were also costs related to treatment and infertility drugs , and the worst of all was the stress that I suffered every time and it didn’t go away , and the feeling of depression , discomfort and failure that followed for several months. Medicines also made me feel very bad. When I was taking medicine , I was always nauseous and I felt bad. Now that she is pregnant , I am stressed in a different way. I am all stressed about my child’s health. Any pain , any change in my body makes me very stressed. Especially since the first one was twins , the heart of one of them stopped in the first weeks. It was very difficult for me”. Findings shown that effective psychosocial care is important to help reduce distress and worry about fertility interventions and outcomes and improve the patient’s well-being and quality of life. Available psychological care is essential for patients to have a healthy experience of their fertility journey . In the present study, actions to promote mental health in women of reproductive age include activities for self-help, seeking help, and assisting people with mental health disorders. Women of childbearing age perform various activities to improve their mental health and in case of injury, they check their mental health and follow up treatment of their mental health disorders if needed. They also know professional and non-professional mental health services and help, and in case of encountering a person with a mental health disorder, they have the ability to help him, including support and guidance to receive services. Previous studies showed that there is a correlation between mental health literacy and attitudes about seeking help for mental health . Also several things such as stigma, fear and lack of trust have an effect on the amount of people seeking help . Woman No. 11 stated: “If I have any problem , I tell my mom or my friend first. But for example , my mother , as I said , had a panic attack , first she went to the health center , then they sent her to a psychiatrist. He went and gave him medicine , he used it for a while and he got better”. The current investigation has some limitations. This was a qualitative study and thus inherently cannot provide representative evidence of the distribution of different perspectives in a population. However, the use of purposive sampling methods allowed us to capture a diverse range of relevant perspectives and to uncover marginalized or unanticipated perspectives. Our use of qualitative thematic analysis brought with it the inherent reliability issue of the method, which we attempted to mitigate through careful examination of the processes of recruitment, data collection, and analysis. Also, the participants were only from three provinces of Iran and did not reflect the geographical diversity of the population, thus limiting the external validity of the study. This study sought to better understand the factors affecting mental health literacy in women of reproductive age and provided a detailed descriptive analysis of their knowledge and understanding of various mental health issues. The main dimensions of mental health literacy in women of reproductive age include “knowledge of information sources and ability to understand mental health”, “ability to use mental health information in women’s lives”, “adaptation to mood changes specific to reproductive age” and “action to promote mental health”. in fact The obtained themes form a quadrilateral, in the heart of which there are special issues and challenges of women’s mental health, including hormonal mood changes, changing roles and social interactions, transitional periods and specific mood disorders. Women of reproductive age. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Kinetic
and Thermodynamic Interplay of Polymer-Mediated
Liquid–Liquid Phase Separation for Poorly Water-Soluble Drugs | 1492e774-d65d-47c3-af0e-cd1df758aeaa | 11151203 | Pharmacology[mh] | Introduction Amorphous solid dispersion (ASD) has been well recognized as an enabling formulation strategy for the bioavailability enhancement of poorly water-soluble drugs. The high free energy state and disordered nature of these ASDs can lead to remarkably high water solubility and enhanced bioavailability. However, mechanistically understanding the ASD’s phase separation process during dissolution and storage is still challenging. In the dissolution of most amorphous drug formulations, various levels of supersaturation are anticipated, where the amount of dissolved drug is above the crystalline drug solubility in the respective aqueous media. The relationships between the maximum drug solubility in its crystalline form and the dynamics of the noncrystalline drug–water miscibility are critical to the phase behaviors of the amorphous formulation during dissolution. When the supersaturation of the drug solution is moderate, the system undertakes a classic nucleation pathway to reduce the free energy and form solid crystalline precipitates. Remarkably, if the level of supersaturation in the drug solution exceeds the spinodal boundary, drug-rich liquid or solid transient phases are often observed in the solution before the crystalline precipitates. The appearance of these transient phases suggests the mechanism of a different nucleation pathway for these amorphous formulations during dissolution, perhaps through liquid–liquid phase separation (LLPS). LLPS is a common phenomenon in which a fluid separates into solute-rich and solute-lean phases. LLPS occurs in cells, plays a vital role in infections, and is critical to the self-assembling of amphiphilic molecules. Thanks to high-resolution analytical techniques, the importance of LLPS has been widely investigated. Its impacts in the scientific fields of chemistry, biology, and pharmaceutics for the nucleation and crystallization of polymers, proteins, minerals, and small organic molecules are also discussed. − Indeed, numerous studies have impressively illustrated the applications of LLPS in understanding and designing amorphous solids. , In a typical binary phase diagram, binodal and spinodal curves govern the mixture’s phase behaviors. When the temperature and composition of the mixture sit within the binodal regions, the thermodynamic favored LLPS occurs. Interestingly, a recent study demonstrates that the liquid phases are likely the dynamic aggregates of clusters of the solute. Thus, at a given temperature, changing the composition of the mixture can result in the occurrence of LLPS in the solution, perhaps simply shifting the location of the mixture from outside the binodal boundary into the binodal region. Therefore, discussing LLPS is often accompanied by the nonclassical nucleation theory, suggesting the complex dynamic behaviors of these supersaturation drug solutions. − The free energy barrier to form these metastable transient phases is smaller than that to form solid crystal nuclei. The condensed liquid or cluster is primarily formed from the original phase during LLPS and is suggested to be the precursor for further nucleation. Although many efforts have been made to elucidate the nonclassical nucleation pathway, this process has not been fully understood. The two-step mechanism is one possible hypothesis to describe the nonclassical nucleation pathway. The concentration and structure fluctuations are the two parameters accompanying the nuclei-forming process. With the formation of LLPS, the concentration and structural fluctuation can further decrease the energy barrier to form stable nuclei within the clusters of droplets. Additives and impurities are revealed to significantly affect the appearance and dynamics of the LLPS and the resulting different polymorphic forms. − Additives can influence the nucleation of the solute from the solution by changing the Gibbs free energy landscape of the drug–polymer–water system, the position and width of the metastable zoom (miscibility gap), crystallization introduction time, and structure evolution of the clusters. In the case of ASD dissolution, polymers can be treated as the most essential additive in the supersaturated drug solution. The formation of nanosized liquid or solid transient phases is frequently observed in polymer-mediated drug–water–polymer ternary systems, suggesting its critical impact in maintaining the supersaturation of the drug solutions. , Two main scenarios have been presented from the currently known cases of polymer-mediated LLPS in supersaturated solutions. In most cases, polymeric excipients have been reported not to change or to have a limited influence on the LLPS onset point (binodal curve). However, the concentration of polymers is in orders of magnitude higher than the drug in the aqueous media. , Some other examples suggested that polymers can alter the LLPS onset point of the drug-water system. − It is worth noting that the polymer types and concentrations were often randomly selected or fixed in these studies without a clear understanding of the boundaries of the drug–polymer–water ternary system. Thus, any systematic method to guide a polymer-mediated LLPS may help us improve our knowledge of this screening approach for polymer selections. In this work, the drug–polymer–water ternary phase diagrams were constructed for celecoxib (CXB)–water solutions with polymers of poly(vinylpyrrolidone) (PVP), poly(vinylpyrrolidone/vinyl acetate) (PVPVA), hydroxypropyl methylcellulose acetate succinate-M grade (HPMCAS-MF), and hydroxypropyl methylcellulose phthalate (HPMCP). The kinetics of LLPS were detected using the UV/vis spectroscopic method. The interplay between the kinetics of the mixing and thermodynamics of the ternary systems was discussed. The impacts of several critical parameters, such as the drug–polymer–water interaction parameters and the types of drug–polymer interactions, were discussed in relation to the positions of the resulting LLPS. Materials and Methods 2.1 Materials Celecoxib (CXB) was purchased from Kemprotec (Carnforth, U.K.). Poly(vinylpyrrolidone) (PVP) and poly(vinylpyrrolidone/vinyl acetate) E-635 (PVPVA) were obtained from Ashland (Kidderminster, U.K.). Hydroxypropyl methylcellulose acetate succinate-M grade (HPMCAS-MF) and hydroxypropyl methylcellulose phthalate (HPMCP) were donated by Shin-Etsu Chemical Company Ltd. (Tokyo, Japan). Methanol (MeOH) and phosphate-buffered saline (PBS) solution (including chemicals of sodium chloride, potassium chloride, sodium phosphate dibasic, and potassium phosphate monobasic) were purchased from Sigma-Aldrich Company Ltd. (Gillingham, U.K.). The purified water was obtained using a PKPD Millipore water purification system 7. Water resistivity was 18.2 MΩ·cm (Merck, U.K.). NMR deuterated solvent dimethyl sulfoxide-d 6 (DMSO- d 6 ) was purchased from Sigma-Aldrich Company Ltd. (Gillingham, U.K.). 2.2 Methods 2.2.1 Solution 1 H Nuclear Magnetic Resonance (NMR) Spectroscopy The solution 1 H NMR spectra investigated interactions between the CXB and polymers. One dimension 1 H NMR spectra were collected using a Bruker Magnet System Ascend 400 MHz spectrometer (Bruker GmbH, Mannheim, Germany) with an acquisition time of 4 seconds, 2-second relaxation delay, 64 scans per sample at 25 °C. Pure CXB, PVP, PVPVA, HPMCAS-MF, HPMCP, and drug–polymer mixtures were dissolved in DMSO- d 6 . The drug concentration was fixed at 3 mg/mL, and the weight ratio of the drug and polymer was 1:5. 2.2.2 Construction of Drug–Polymer–Water Ternary Phase Diagram Drug–polymer–water ternary phase diagrams were constructed using a previously published method. Two approaches were utilized to obtain the binary Flory–Huggins interaction parameters. For water–polymer–drug systems, sorption isotherm experiments of water with the ingredients were collected using a DVS advantage system (Surface Measurement Systems, London, U.K.) at a temperature of 25 °C. Approximately 50–100 mg of ingredients were placed in a sample holder (mesh) within the DVS chamber. The sample environment humidity was then gradually increased from 0 to 90% RH at 10% RH intervals, using 120 min per step. The amount of water (in weight) absorbed into the sample at each water partial pressure was used to calculate the water–ingredient interaction parameter. Strong localized water–polymer bonding may occur for some partially frozen water–polymer systems such as HPMCAS. Hence, in this study, only a completely dried sample was used. Based on the DVS approach, the F–H interaction parameter may be derived using the activity of water in the mixture: 1 where the ϕ is the volume fraction of water (ϕ w ) or solute (ϕ s ), and m is the molar volume ratio of the solute over water. The solvent in this study is the water; therefore, the change of water vapor partial pressure in the DVS tests can be used to define water activity ( a w ) in . For drug–polymer F–H interaction parameters at 25 °C, the Hildebrand solubility parameter approach was utilized for the calculations ( Supporting Information ). 2.2.3 Ultraviolet/Visible (UV) Extinction Study PVP, PVPVA, HPMCAS, and HPMCP polymers with a 1 mg/mL concentration were dissolved in the pH 7.4 PBS buffer at 37 °C. UV spectra of polymer solutions were scanned using the GENESYS 180 UV/Vis spectrophotometer (Thermo Fisher Scientific, Madison, USA) connecting with a fiber optic probe coupler in the range from 200 to 800 nm, with a scanning speed of 1 nm/s. In a typical procedure, 50–250 μL of CXB-MeOH stock solution (4 mg/mL) was gradually added into a 20 mL PBS solution (pH 7.4) using a syringe pump at various flow rates. The final CXB concentrations were 10, 20, 30, 40, and 50 μg/mL, in which 10 mg of polymer was predissolved. 2.2.4 Determination of the CXB–Polymer–Water Ternary System LLPS Onset Concentration The LLPS onset concentration point was determined using the ultraviolet (UV) extinction method. Polymers of PVP, PVPVA, HPMCAS-MF, and HPMCP with concentrations of 100, 500, or 1000 μg/mL were predissolved in PBS or codissolved with the CXB in the MeOH stock solution. Four mg/mL of the CXB stock solution in the 10 mL syringe was gradually added into 20 mL PBS solution using the Aladdin SyringeONE programmable syringe pump (AL-1000, Hitchin, U.K.) with various mixing rates of the stock solution and PBS. The mixing rates were controlled by altering the pumping rates at 3, 1, or 0.5 mL/h, generating the 50 μg/mL CXB solution in 30, 15, and 5 min, respectively. The PBS solution was stirred by a magnetic stirrer at 200 rpm, and a water bath was controlled at a constant temperature of 37 °C. Values of UV extinction were measured at the interval of 0.167 μg/mL CXB concentration until a clear extinction slope difference can be observed from the initial drug concentration, which indicates the formation of the drug-rich phase in the CXB–polymer–PBS ternary system. The UV extinction was determined at the wavelength confirmed ( Supporting Information Figure S1 ), where UV absorption of CXB and polymer molecules can be insignificant. 2.2.5 Verification of the LLPS of CXB–Polymer–Water via Cryo-TEM and Total Internal Reflection Fluorescence Microscopy Cryogenic transmission electron microscopy (Cryo-EM, FEI, Thermo Fisher Scientific, Eindhoven, The Netherlands) was used to characterize the appearance of intransit nanoparticles/nanodroplets after the LLPS onset points. 3 μL of the liquid was pipetted onto a previously glow discharged, lacey carbon film EM grid, blotted for 1.2 s, and plunged frozen into liquid ethane using a Leica GP plunge freezer (Leica Microsystems, Wetzlar, Germany). The sample was kept at liquid nitrogen temperature while transferred to a Gatan 626 Cryotransfer holder (Gatan, Pleasanton, CA) and imaged using Cryo-EM. Images were acquired on a CETA camera (FEI, Thermo Fisher Scientific, Eindhoven, The Netherlands) using low-dose acquisition software. A total internal reflection fluorescence microscope (TIRFM, Lecia, Wetzlar, Germany) coupled with a 40X/0.85 NA HC PL APO objective lens was used to assess the appearance of the CXB–polymer LLPS phase. Pyrene was used as the hydrophobic fluorescence probe and dissolved with the CXB in methanol solution before pumping into the PBS using the 3 mL/h rate described above. For TIRFM fast acquisition, videos of the CXB–polymer suspensions were recorded using an Andor Zyla sCMOS camera with 4.2 megapixels (Oxford Instruments, Oxford, U.K.) at 100 ms per frame for 20 s. The excitation at 336 ± 40 nm and emission at 384 ± 40 nm were used. Samples were also recorded using the TIRFM with a polarizer at a crossed position for comparison. All videos were analyzed using ImageJ software (version 1.54f, National Institutes of Health, USA). 2.2.6 Statistical Analysis Data were analyzed using GraphPad Prism (version 9.0.0) and presented as mean ± standard deviation of three replicates. The statistical analysis was carried out using ordinary one- and two-way ANOVA. A significant difference was considered when p < 0.05. Materials Celecoxib (CXB) was purchased from Kemprotec (Carnforth, U.K.). Poly(vinylpyrrolidone) (PVP) and poly(vinylpyrrolidone/vinyl acetate) E-635 (PVPVA) were obtained from Ashland (Kidderminster, U.K.). Hydroxypropyl methylcellulose acetate succinate-M grade (HPMCAS-MF) and hydroxypropyl methylcellulose phthalate (HPMCP) were donated by Shin-Etsu Chemical Company Ltd. (Tokyo, Japan). Methanol (MeOH) and phosphate-buffered saline (PBS) solution (including chemicals of sodium chloride, potassium chloride, sodium phosphate dibasic, and potassium phosphate monobasic) were purchased from Sigma-Aldrich Company Ltd. (Gillingham, U.K.). The purified water was obtained using a PKPD Millipore water purification system 7. Water resistivity was 18.2 MΩ·cm (Merck, U.K.). NMR deuterated solvent dimethyl sulfoxide-d 6 (DMSO- d 6 ) was purchased from Sigma-Aldrich Company Ltd. (Gillingham, U.K.). Methods 2.2.1 Solution 1 H Nuclear Magnetic Resonance (NMR) Spectroscopy The solution 1 H NMR spectra investigated interactions between the CXB and polymers. One dimension 1 H NMR spectra were collected using a Bruker Magnet System Ascend 400 MHz spectrometer (Bruker GmbH, Mannheim, Germany) with an acquisition time of 4 seconds, 2-second relaxation delay, 64 scans per sample at 25 °C. Pure CXB, PVP, PVPVA, HPMCAS-MF, HPMCP, and drug–polymer mixtures were dissolved in DMSO- d 6 . The drug concentration was fixed at 3 mg/mL, and the weight ratio of the drug and polymer was 1:5. 2.2.2 Construction of Drug–Polymer–Water Ternary Phase Diagram Drug–polymer–water ternary phase diagrams were constructed using a previously published method. Two approaches were utilized to obtain the binary Flory–Huggins interaction parameters. For water–polymer–drug systems, sorption isotherm experiments of water with the ingredients were collected using a DVS advantage system (Surface Measurement Systems, London, U.K.) at a temperature of 25 °C. Approximately 50–100 mg of ingredients were placed in a sample holder (mesh) within the DVS chamber. The sample environment humidity was then gradually increased from 0 to 90% RH at 10% RH intervals, using 120 min per step. The amount of water (in weight) absorbed into the sample at each water partial pressure was used to calculate the water–ingredient interaction parameter. Strong localized water–polymer bonding may occur for some partially frozen water–polymer systems such as HPMCAS. Hence, in this study, only a completely dried sample was used. Based on the DVS approach, the F–H interaction parameter may be derived using the activity of water in the mixture: 1 where the ϕ is the volume fraction of water (ϕ w ) or solute (ϕ s ), and m is the molar volume ratio of the solute over water. The solvent in this study is the water; therefore, the change of water vapor partial pressure in the DVS tests can be used to define water activity ( a w ) in . For drug–polymer F–H interaction parameters at 25 °C, the Hildebrand solubility parameter approach was utilized for the calculations ( Supporting Information ). 2.2.3 Ultraviolet/Visible (UV) Extinction Study PVP, PVPVA, HPMCAS, and HPMCP polymers with a 1 mg/mL concentration were dissolved in the pH 7.4 PBS buffer at 37 °C. UV spectra of polymer solutions were scanned using the GENESYS 180 UV/Vis spectrophotometer (Thermo Fisher Scientific, Madison, USA) connecting with a fiber optic probe coupler in the range from 200 to 800 nm, with a scanning speed of 1 nm/s. In a typical procedure, 50–250 μL of CXB-MeOH stock solution (4 mg/mL) was gradually added into a 20 mL PBS solution (pH 7.4) using a syringe pump at various flow rates. The final CXB concentrations were 10, 20, 30, 40, and 50 μg/mL, in which 10 mg of polymer was predissolved. 2.2.4 Determination of the CXB–Polymer–Water Ternary System LLPS Onset Concentration The LLPS onset concentration point was determined using the ultraviolet (UV) extinction method. Polymers of PVP, PVPVA, HPMCAS-MF, and HPMCP with concentrations of 100, 500, or 1000 μg/mL were predissolved in PBS or codissolved with the CXB in the MeOH stock solution. Four mg/mL of the CXB stock solution in the 10 mL syringe was gradually added into 20 mL PBS solution using the Aladdin SyringeONE programmable syringe pump (AL-1000, Hitchin, U.K.) with various mixing rates of the stock solution and PBS. The mixing rates were controlled by altering the pumping rates at 3, 1, or 0.5 mL/h, generating the 50 μg/mL CXB solution in 30, 15, and 5 min, respectively. The PBS solution was stirred by a magnetic stirrer at 200 rpm, and a water bath was controlled at a constant temperature of 37 °C. Values of UV extinction were measured at the interval of 0.167 μg/mL CXB concentration until a clear extinction slope difference can be observed from the initial drug concentration, which indicates the formation of the drug-rich phase in the CXB–polymer–PBS ternary system. The UV extinction was determined at the wavelength confirmed ( Supporting Information Figure S1 ), where UV absorption of CXB and polymer molecules can be insignificant. 2.2.5 Verification of the LLPS of CXB–Polymer–Water via Cryo-TEM and Total Internal Reflection Fluorescence Microscopy Cryogenic transmission electron microscopy (Cryo-EM, FEI, Thermo Fisher Scientific, Eindhoven, The Netherlands) was used to characterize the appearance of intransit nanoparticles/nanodroplets after the LLPS onset points. 3 μL of the liquid was pipetted onto a previously glow discharged, lacey carbon film EM grid, blotted for 1.2 s, and plunged frozen into liquid ethane using a Leica GP plunge freezer (Leica Microsystems, Wetzlar, Germany). The sample was kept at liquid nitrogen temperature while transferred to a Gatan 626 Cryotransfer holder (Gatan, Pleasanton, CA) and imaged using Cryo-EM. Images were acquired on a CETA camera (FEI, Thermo Fisher Scientific, Eindhoven, The Netherlands) using low-dose acquisition software. A total internal reflection fluorescence microscope (TIRFM, Lecia, Wetzlar, Germany) coupled with a 40X/0.85 NA HC PL APO objective lens was used to assess the appearance of the CXB–polymer LLPS phase. Pyrene was used as the hydrophobic fluorescence probe and dissolved with the CXB in methanol solution before pumping into the PBS using the 3 mL/h rate described above. For TIRFM fast acquisition, videos of the CXB–polymer suspensions were recorded using an Andor Zyla sCMOS camera with 4.2 megapixels (Oxford Instruments, Oxford, U.K.) at 100 ms per frame for 20 s. The excitation at 336 ± 40 nm and emission at 384 ± 40 nm were used. Samples were also recorded using the TIRFM with a polarizer at a crossed position for comparison. All videos were analyzed using ImageJ software (version 1.54f, National Institutes of Health, USA). 2.2.6 Statistical Analysis Data were analyzed using GraphPad Prism (version 9.0.0) and presented as mean ± standard deviation of three replicates. The statistical analysis was carried out using ordinary one- and two-way ANOVA. A significant difference was considered when p < 0.05. Solution 1 H Nuclear Magnetic Resonance (NMR) Spectroscopy The solution 1 H NMR spectra investigated interactions between the CXB and polymers. One dimension 1 H NMR spectra were collected using a Bruker Magnet System Ascend 400 MHz spectrometer (Bruker GmbH, Mannheim, Germany) with an acquisition time of 4 seconds, 2-second relaxation delay, 64 scans per sample at 25 °C. Pure CXB, PVP, PVPVA, HPMCAS-MF, HPMCP, and drug–polymer mixtures were dissolved in DMSO- d 6 . The drug concentration was fixed at 3 mg/mL, and the weight ratio of the drug and polymer was 1:5. Construction of Drug–Polymer–Water Ternary Phase Diagram Drug–polymer–water ternary phase diagrams were constructed using a previously published method. Two approaches were utilized to obtain the binary Flory–Huggins interaction parameters. For water–polymer–drug systems, sorption isotherm experiments of water with the ingredients were collected using a DVS advantage system (Surface Measurement Systems, London, U.K.) at a temperature of 25 °C. Approximately 50–100 mg of ingredients were placed in a sample holder (mesh) within the DVS chamber. The sample environment humidity was then gradually increased from 0 to 90% RH at 10% RH intervals, using 120 min per step. The amount of water (in weight) absorbed into the sample at each water partial pressure was used to calculate the water–ingredient interaction parameter. Strong localized water–polymer bonding may occur for some partially frozen water–polymer systems such as HPMCAS. Hence, in this study, only a completely dried sample was used. Based on the DVS approach, the F–H interaction parameter may be derived using the activity of water in the mixture: 1 where the ϕ is the volume fraction of water (ϕ w ) or solute (ϕ s ), and m is the molar volume ratio of the solute over water. The solvent in this study is the water; therefore, the change of water vapor partial pressure in the DVS tests can be used to define water activity ( a w ) in . For drug–polymer F–H interaction parameters at 25 °C, the Hildebrand solubility parameter approach was utilized for the calculations ( Supporting Information ). Ultraviolet/Visible (UV) Extinction Study PVP, PVPVA, HPMCAS, and HPMCP polymers with a 1 mg/mL concentration were dissolved in the pH 7.4 PBS buffer at 37 °C. UV spectra of polymer solutions were scanned using the GENESYS 180 UV/Vis spectrophotometer (Thermo Fisher Scientific, Madison, USA) connecting with a fiber optic probe coupler in the range from 200 to 800 nm, with a scanning speed of 1 nm/s. In a typical procedure, 50–250 μL of CXB-MeOH stock solution (4 mg/mL) was gradually added into a 20 mL PBS solution (pH 7.4) using a syringe pump at various flow rates. The final CXB concentrations were 10, 20, 30, 40, and 50 μg/mL, in which 10 mg of polymer was predissolved. Determination of the CXB–Polymer–Water Ternary System LLPS Onset Concentration The LLPS onset concentration point was determined using the ultraviolet (UV) extinction method. Polymers of PVP, PVPVA, HPMCAS-MF, and HPMCP with concentrations of 100, 500, or 1000 μg/mL were predissolved in PBS or codissolved with the CXB in the MeOH stock solution. Four mg/mL of the CXB stock solution in the 10 mL syringe was gradually added into 20 mL PBS solution using the Aladdin SyringeONE programmable syringe pump (AL-1000, Hitchin, U.K.) with various mixing rates of the stock solution and PBS. The mixing rates were controlled by altering the pumping rates at 3, 1, or 0.5 mL/h, generating the 50 μg/mL CXB solution in 30, 15, and 5 min, respectively. The PBS solution was stirred by a magnetic stirrer at 200 rpm, and a water bath was controlled at a constant temperature of 37 °C. Values of UV extinction were measured at the interval of 0.167 μg/mL CXB concentration until a clear extinction slope difference can be observed from the initial drug concentration, which indicates the formation of the drug-rich phase in the CXB–polymer–PBS ternary system. The UV extinction was determined at the wavelength confirmed ( Supporting Information Figure S1 ), where UV absorption of CXB and polymer molecules can be insignificant. Verification of the LLPS of CXB–Polymer–Water via Cryo-TEM and Total Internal Reflection Fluorescence Microscopy Cryogenic transmission electron microscopy (Cryo-EM, FEI, Thermo Fisher Scientific, Eindhoven, The Netherlands) was used to characterize the appearance of intransit nanoparticles/nanodroplets after the LLPS onset points. 3 μL of the liquid was pipetted onto a previously glow discharged, lacey carbon film EM grid, blotted for 1.2 s, and plunged frozen into liquid ethane using a Leica GP plunge freezer (Leica Microsystems, Wetzlar, Germany). The sample was kept at liquid nitrogen temperature while transferred to a Gatan 626 Cryotransfer holder (Gatan, Pleasanton, CA) and imaged using Cryo-EM. Images were acquired on a CETA camera (FEI, Thermo Fisher Scientific, Eindhoven, The Netherlands) using low-dose acquisition software. A total internal reflection fluorescence microscope (TIRFM, Lecia, Wetzlar, Germany) coupled with a 40X/0.85 NA HC PL APO objective lens was used to assess the appearance of the CXB–polymer LLPS phase. Pyrene was used as the hydrophobic fluorescence probe and dissolved with the CXB in methanol solution before pumping into the PBS using the 3 mL/h rate described above. For TIRFM fast acquisition, videos of the CXB–polymer suspensions were recorded using an Andor Zyla sCMOS camera with 4.2 megapixels (Oxford Instruments, Oxford, U.K.) at 100 ms per frame for 20 s. The excitation at 336 ± 40 nm and emission at 384 ± 40 nm were used. Samples were also recorded using the TIRFM with a polarizer at a crossed position for comparison. All videos were analyzed using ImageJ software (version 1.54f, National Institutes of Health, USA). Statistical Analysis Data were analyzed using GraphPad Prism (version 9.0.0) and presented as mean ± standard deviation of three replicates. The statistical analysis was carried out using ordinary one- and two-way ANOVA. A significant difference was considered when p < 0.05. Results and Discussions 3.1 Solution 1 H NMR Spectroscopy All solution NMR experiments were conducted in a nonaqueous environment to emphasize the drug–polymer interactions that may be more relevant to the ASD solids before rehydration . A nonaqueous solution in the NMR experiment can help differentiate the drug–polymer interactions for the drug–polymer–water ternary system before rehydration. It is important to investigate the relationship between the CXB–polymer interaction at a nonaqueous environment first and the subsequent dynamics of the LLPS. As suggested in this work, water can easily disrupt water senstive intermolecular interactions, such as the H-bonding between the drug and polymer. In contrast, hydrophobic interactions between the drug and the polymer are less affected during the dissolution of ASDs. , We argue that if the drug–polymer interactions are mainly hydrophobic, then the influence of water on the LLPS onset point may be different from systems with other types of drug–polymer interactions. The proton NMR results suggested that hydrogen bonding was formed within CXB–PVP and CXB–PVPVA combinations, supported by the downshift of hydrogens on the –NH 2 ( a–c, red labels). The hydrogen atoms on the –NH 2 group shifted to higher parts per million (ppm) in the presence of PVP, suggesting stronger hydrogen bonding between CXB and PVP than between CXB and PVPVA. This result was also reported previously in the literature. For example, IR spectroscopy indicated that the –VA groups of the PVPVA will not interact with CXB when forming ASDs. The critical interaction differences between CXB–PVP and CXB–PVPVA may also be highlighted by the level of determined drug–polymer glass transition temperatures deviating from the Gordon–Taylor equation predictions. Rask et al. found that the ASD of CXB–PVP has a more extended deviation of the glass transition temperatures rather than the CXB–PVPVA system, suggesting the CXB–PVP system generated a stronger interaction than the latter one. In comparison, hydrophobic interactions are the main form of intermolecular interactions for CXB–HPMCAS and CXB–HPMCP systems, as evidenced by the appearance of the lower ppm position of the hydrogen atoms in the pyrazole ring of the CXB ( d,e, blue labels). Similar upfield shifts have also been reported in the cases of CXB/hydroxypropyl-β-cyclodextrin and CXB/2,6-di-O-methyl-β-cyclodextrin systems. The upfield shift is caused by the shielding effect of oxygen atoms on the excipients, suggesting the CXB pyrazole ring was a critical hydrophobic group to interact with HPMCAS and HPMCP polymers in a nonaqueous environment. 3.2 CXB–Polymer–Water Ternary Phase Diagrams CXB–polymer–water ternary phase diagrams for all four polymeric additives were constructed using previously established methods . The parameters such as molecular volume, density, solubility, and F–H interaction parameters are all provided in the Supporting Information . Within the diagram, the binodal and spinodal curves were plotted to highlight the phase behaviors of the CXB–polymer–water mixture. The ternary phase diagram illustrates the relevant compositions of the CXB supersaturated solutions with polymeric additives. However, given the limited resolution of the modeling tool used and the relatively small concentrations of the CXB in water, these ternary phase diagrams were not established using the existing set of F–H interaction parameters. Instead, four ternary phase diagrams were successfully constructed with lower values of the F–H interaction parameters ( Figure S2 ). Nevertheless, this approach illustrates the importance of polymeric additives for all four ternary systems at high polymer or water compositions. The locations of the LLPS boundary indicate the formation of droplets during this process. With these boundaries, the spinodal curves sit within the binodal curves, showing the limited local stability of the mixture in these compositions. The gaps between the spinodal and binodal curves are identified as the miscibility gap, highlighting the metastability nature of the LLPS. Within the miscibility gap, it is, therefore, most likely that the appearance of the CXB–polymer–water transient phases can be observed using various analytical techniques. The starting point of such an LLPS is understood to be very close to the binodal curve of the phase diagram; it is named the LLPS onset concentration in this work. The binodal curve defines the temperature and composition of the mixture at which phase separation is thermodynamically favorable. Following the positions of the binodal curves in all four systems in and S2 , the identifiable areas are located at the three corners of the ternary phase diagram, e.g., the high polymer, high water, and high drug compositions. At these locations, the influences of the other two lower components may be illustrated by using the miscibility gaps. For example, high polymer and low drug composition areas indicate the possible phase behaviors of the drug–polymer ASD systems in water, as shown in the top areas in . A homogeneous CXB–polymer ASD system may gradually move into the unstable LLPS when exposed to moisture during dissolution. Indeed, identifying the miscibility gaps for ternary systems has been widely used to develop formulations with nanoscale artifacts for many important commercial applications. , Furthermore, it is observed that the shape of the miscibility gap is highly influenced by the polymeric additives in the CXB-water system, reflecting the differences in drug–polymer and water–polymer interaction parameters ( Table S1 ). The wide miscibility gap from selecting appropriate polymers, e.g., HPMCAS, HPMCP, and PVPVA, is expected to increase the drug concentrations and the extent of supersaturation for CXB in the aqueous medium (LLPS). In comparison, the miscibility gap for the CXB–PVP–water system is relatively small if the values of interaction parameters are reduced proportionally ( Table S1 ). The resulting ternary phase diagram appeared to have similar miscibility gaps at high water and low drug/polymer compositions, e.g., the supersaturated drug–water/polymer–water solutions. Although these constructed ternary phase diagrams are based on theoretical parameters derived using solubility parameters, the trend of such changes in the miscibility gaps reveals a reasonable outcome for the drug–polymer–water systems. With such a simple approach, one can quickly screen the polymeric additives for a given drug molecule for solubility enhancements and the likelihood of LLPS. In this work, the UV extinction measurement investigates the phase behaviors of supersaturated CXB solutions in the presence of various polymeric excipients. The supersaturated drug solution is achieved using the antisolvent/solvent-shifting method. Typically, the hydrophobic drug is dissolved in an organic solvent. The drug stock solution is then gradually added to the aqueous environment, where drugs reach the desirable supersaturation level until the LLPS occurs. This method hypothesizes that the limited fraction of organic solvent will not influence the phase behaviors of the system. The polar organic solvent in emulsion droplets will immediately disperse into the water. However, it is worth noting that due to the nature of this measurement, the appearance of the polymer–water phase separation can also cause UV extinction. 3.3 Factors Influencing the Detection of LLPS Onset Concentrations Poorly water-soluble drug molecules exposed to an aqueous environment tend to precipitate through either the classical or nonclassical nucleation pathway. UV spectroscopy is one of the most commonly used techniques to characterize the drug-rich phases generated in supersaturated solutions. Typically, when electromagnetic radiation goes through a solution, the radiation may attenuate or become extinct due to absorption or scattering. , The absorption occurs through the electrons of molecules in a solution, absorbing energy from radiation and expanding to a higher energy state. The scattering will be observed when insoluble particles are presented in the solution, where larger particles exhibit a higher level of scattering. , In the drug–water systems, the particle scattering due to drug-rich phase generation was reported to lead to a spectrum baseline distortion, forming the foundation of this measurement. A high UV extinction is commonly observed at the high absorption regions of the drug molecule, ranging from 200 to 400 nm. PVP, PVPVA, and HPMCAS showed no absorption peak at wavelengths larger than 240 nm. HPMCP PBS solution had a notable peak at a wavelength of 292 nm. To minimize the UV extinction caused by polymer absorption, a wavelength of 360 nm was selected for all experiments. UV extinction spectra of CXB–PVPVA in PBS and MeOH solutions were verified at all relevant compositions with a 10–50 μg/mL drug concentration range in a 10 μg/mL interval ( Figure S2 ). A clear step change in the relationship between drug concentrations and UV extinctions could be observed, indicating partial changes in the physical forms of the drug molecules within the solution. The occurrence of additional UV extinction was interpreted as the light scattering caused by the particles in high drug concentration samples, known as the Tyndall scattering. Therefore, the drug-rich phase onset point of the CXB–PVPVA–PBS ternary system could be estimated to be 28 μg/mL of CXB concentration. First, without the presence of polymer, the UV extinction data of pure CXB solution as a function of CXB concentration at 37 °C are shown in , with the drug stock solution pumping rates at 3, 1, and 0.5 mL/h. A wavelength of 360 nm was used to estimate the phase separation onset concentration of the CXB solution. To avoid interferences, the wavelength employed to determine the scattering intensity was far from the drug’s absorption wavelength range. The red lines were two least-squares regression curves fitted using data points at the initial stage and the subsequent appearance of supersaturated solutions with the increased CXB. The onset concentration of LLPS was derived from the intersection point of two red lines. The phase separation occurrence concentrations were determined as 7.75 ± 0.78, 8.58 ± 1.95, and 6.22 ± 1.07 μg/mL with the drug stock solution pumping rates of 3, 1, and 0.5 mL/h, respectively. No notable difference was observed with the drug stock solution pumping rate ( p > 0.05). The drug concentration at the solution’s optical turbidity point has also been highly associated with the drug’s theoretical amorphous solubility. However, the experimentally detected phase separation concentration of pure CXB was remarkably lower than the theoretically calculated drug amorphous solubility. Using various theories, the CXB amorphous solubility was estimated to be 19–22.6 μg/mL. It was suggested that the fast recrystallization speed of the CXB has resulted in the formation of small crystalline drug particles before the LLPS. In this work, white precipitations could also be observed by the optical image of the pure CXB–PBS solution in a matter of minutes . Crystalline drug suspension in the solution may contribute to the overall UV spectrum scattering and lead to the step change at the slopes. The inline UV method can be affected by the appearance of all kinds of matters in the CXB titration experiment, such as the nanocrystalline, nanodroplet, amorphous nanoparticles or mixtures of all of the above. We also tried to pause the titration when the LLPS onset point was detected. The UV extinction value remained stable for at least 20 min, suggesting the number and size of the metastable particles were stable within the individual test. Additional experiments were also conducted to highlight the differences in UV extinction values caused by the crystalline CXB and LLPS ( Figure S3 ). Seeding the CXB–PVP–water solution at a CXB concentration of 36 μg/mL with an additional 10% w/w crystalline CXB, a significant jump in the UV extinction was observed, which was higher than that caused by the LLPS. To further identify the compositions of the nanosized matter at the onset point of LLPS, the cryo-EM technique and high-resolution TIRFM (with pyrene as the hydrophobic fluorescence probe, Supporting Information Videos 1–4 ) were used. The cryo-EM and TIRFM highlighted the appearance of spherical particles ranging from 40 nm to several micrometers. Particularly in the TIRFM with polarized filter videos, no significant birefringence was observed, indicating the possible aggregation of the CXB amorphous nanodroplet following the initial LLPS at much smaller sizes . The appearance of noncrystalline nano/microparticles with the CXB–polymer–water suspension highlighted the metastability nature of the mixture, thus validating that the main cause of the UV extinction is indeed attributed to the LLPS. It should be noted that several previous research articles presented the behaviors of LLPS for CXB in PBS with the predissolved polymeric matrices, such as PVP, PVPVA, and HPMCAS. , , , These values suggest that under the experimental conditions described in this work (microfluidic pump and mixing), the onset of UV baseline change should be mainly attributed to LLPS in the CXB titration process (all experiments at various pumping rates were completed within 1 h). As mentioned, phase behaviors, e.g., LLPS in the drug–polymer–water ternary systems, were revealed to affect the solubility and permeability enhancement of ASD during oral administration. However, the polymer influence on the transient drug-rich phases is not fully understood. This section investigated various experimental conditions based on the onset point of LLPS for CXB–polymer–water ternary systems, including polymer types, polymer concentrations, drug–polymer interaction approaches, and drug–polymer mixing rates. 3.4 Effects of Polymer Types and Drug Stock Solution Pumping Rates Given the usual low LLPS onset concentrations of most poorly water-soluble drugs in the aqueous medium, polymeric excipients are often predissolved to suppress the precipitations during the experiment. In this section, different polymers of PVP, PVPVA, HPMCAS, and HPMCP with a concentration of 1 mg/mL were predissolved in the PBS buffer to inhibit the precipitations of CXB. The phase behaviors of drug–polymer solutions were monitored using the same UV extinction method. The polymer type is a critical parameter that influences the LLPS onset concentration in a CXB–polymer–water ternary system, as the UV extinction profiles depicted in depend on the drug concentrations in the medium. Compared with a pure CXB–PBS solution, it is clear that the LLPS onset concentrations have been altered for CXB when different polymeric materials are predissolved within the aqueous medium. The LLPS data varied in different polymer–water combinations. The LLPS onset concentrations for CXB ternary systems with predissolved PVP ( a) and HPMCP ( d) were recorded at 37.2 ± 0.77 and 37.6 ± 0.98 μg/mL, respectively. CXB ternary systems with polymers of PVPVA ( b) and HPMCAS ( c) exhibited lower LLPS onset concentrations, measured to be 18.0 ± 1.82 and 15.1 ± 1.33 μg/mL. In a phase diagram, drug-rich phases were suggested to be generated in the metastable region. Drug-rich phases could be determined when these transient phases reach the local minimum energy position and are kinetically stable for a short period. The kinetic influence on the determination of the LLPS point was studied when the mixing rate of the stock solution and the PBS buffer was changed. This work suggested that the drug–polymer mixing rate is not a significant parameter of the LLPS onset concentration. Specifically, examples of UV extinction profiles of the CXB–polymer–water ternary system with drug stock solution pumping rates of 3 and 0.5 mL/h at the wavelength of 360 nm are shown in and ; additional profiles of the UV extinction at a pumping rate of 1 mL/h are provided in Figure S3 . To further verify the existence of metastable CXB–polymer–water transient phases, cryo-EM microscopic analysis was carried out for several selected samples after reaching LLPS onset points . Immediately after reaching the LLPS onset points, the liquids were drawn from the sample vials and rapidly frozen to achieve amorphous ice for cryo-EM. Round-shaped condensed matter with sizes of 20–80 nm was observed in all CXB–polymer–water systems. The amount of round-shaped condensed matter may indicate the CXB LLPS onset concentrations, where more spherical particles were observed in HPMCP and PVP-based systems than in HPMCAS and PVPVA mixtures. Furthermore, signs of agglomeration were also observed in HPMCAS and HPMCP-based CXB suspensions, reflecting the possible colloidal nature of these two polymeric matrices. LLPS onset concentrations in the PBS solution with or without polymers using various mixing rates are summarized in and . Values were calculated individually at different conditions. Error bars were derived from standard deviations of those values. Blue, orange, and yellow bars represent the drug stock solution (4 mg/mL) pumping rates of 3, 1, and 0.5 mL/h, respectively. 50 μg/mL CXB solutions were generated in 5, 15, and 30 min, respectively. CXB LLPS in PVP and HPMCP aqueous solutions exhibited significantly higher concentrations than those in PVPVA and HPMCAS solutions ( p < 0.0001). No significant difference in the LLPS onset concentrations was observed when drug stock solution mixing rates were altered (two-way ANOVA, p > 0.05). The result suggested that the CXB LLPS point in the CXB–polymer–water ternary systems had a less kinetic influence within the first 30 min of the experiments. The presence of polymers influenced the LLPS onset concentrations remarkably by controlling the position of the binodal line. Samples with strong drug–polymer interactions were observed to undergo LLPS at high drug concentrations. For example, hydrogen bonding was identified within both CXB–PVP and CXB–PVPVA systems in the nonaqueous situation by 1 H NMR spectra. At the mixing rate of 3 mL/h, the LLPS onset concentration for CXB at a 1 mg/mL PVP–PBS solution was approximately two times higher than that of the PVPVA solution, . A higher CXB LLPS onset value of the CXB–PVP system was interpreted by the stronger hydrogen bonding of the drug and polymer, evidenced by a more extensive chemical shift in 1 H NMR spectra. Similarly, PVPVA and HPMCAS were reported to reduce the LLPS onset concentrations in other drug systems. It was suggested that the ibuprofen solubility was reduced in several polymer solutions, including the PVPVA, and that the bulk ibuprofen concentration was reduced with PVPVA. Miao et al. reported that the LLPS value of paclitaxel decreased from approximately 40–23 μg/mL with the HPMCAS (MF). 3.5 Role of Polymer Concentrations on the LLPS Onset The drug–polymer composition has been commonly highlighted to influence the LLPS point in supersaturated drug-water solutions. This work determined the CXB LLPS onset concentrations of several CXB–polymer–water ternary systems with different polymer concentrations. UV extinction profiles as a function of drug concentration with the polymer concentrations at 500 and 100 μg/mL are shown in and . The LLPS onset concentrations at different systems were derived from the step change of regression curve slopes (red lines). Various LLPS onset concentrations observed for the CXB–polymer–water ternary system are summarized in and . Blue, orange, and yellow bars represent the polymer concentration at 1000, 500, and 100 μg/mL, respectively. In this case, CXB LLPS onset concentrations in the ternary systems were usually not altered when reducing the polymer concentration from 1000 μg/mL to 100 μg/mL ( p > 0.05), as shown in and . However, the system with HPMCAS exhibited an abnormal LLPS concentration at a higher polymer concentration. The CXB LLPS concentration of the solution with the HPMCAS concentration of 1000 μg/mL was 12.7 ± 5.72 μg/mL. This value increased to 21.1 ± 2.27 μg/mL when the polymer concentration decreased to 500 μg/mL and then remained constant between 500 and 100 μg/mL. The UV spectra of HPMCAS PBS solution (pH = 7.4) with a serial of polymer concentrations at 37 °C are illustrated in b, where gray, orange, and blue curves represent HPMCAS concentrations of 1000, 500, and 100 μg/mL, respectively. Scattering was the sole factor contributing to the overall extinction at the LLPS determination wavelength (360 nm). It should be noted that the scattering was already observed in the HPMCAS solution at a concentration of 1000 μg/mL without the addition of CXB. This was due to the HPMCAS aggregated upon high polymer concentrations in the PBS solution (37 °C), where HPMCAS–PBS demixing occurred even without drug molecules (raised baseline in b). Similar observations on the nature of colloid formation for HPMCAS at high concentrations have already been reported in the literature. , To maintain the consistency of the experimental conditions among all polymeric carriers, the influence of HPMCAS aggregation was blanked out from the UV extinction before addition of the CXB stock solution. However, the results suggested that only approximately 12.7 μg/mL CXB was required to disrupt the existing HPMCAS aggregations in the PBS solution, resulting in a phase-separated CXB–HPMCAS colloid suspension in the PBS. 3.6 Influence of Preformed Drug–Polymer Interaction in Stock CXB Solution The LLPS concept hypothesizes that the complexity of the free energy landscape can alter the dynamics of the resulting transient phase. In this case, the rate of mixing of CXB with water and the presence of polymeric carriers should be expected to alter the resulting LLPS onset point. The Flory–Huggins model is an important theoretical approach for estimating the phase boundaries in polymer-relevant solutions, which models the interaction of the components within a lattice theory. , The entropic contribution to the free energy landscape of the system is determined by enumerating the distinct configurations of molecules and polymers within the lattice. In contrast, according to a regular solution theory, the enthalpic contribution arises from the paired interaction energies between the components. The component interaction is a dominating parameter influencing the free energy landscape and the LLPS. Previous sections studied the weaker interaction between the drug and various polymers, which can result in a lower LLPS onset concentration. Such alteration of drug–polymer interaction can also be complicated by moving the interaction from a nonaqueous state to an aqueous state. Chen et al. suggested that the drug–polymer intermolecular interaction strength in a nonaqueous environment may be weaker than in an aqueous solution. Marsac et al. found the hydrogen bonding between felodipine and PVP will be disrupted with the introduction of water. Fundamentally, the experimental approach to obtain the LLPS of a small molecule drug in water is via solvent shift, where a drug-organic solvent solution is gradually added into a polymer–water solution. Quick diffusion of the organic solvent in water results in phase separation of the drug solution due to the poor water solubility. In this situation, the drug–polymer interaction is expected to form a competitive relationship with the water–polymer interactions. In comparison, the LLPS of the drug–polymer–water bond can also be obtained by adding a drug–polymer organic solvent solution into the water medium. However, in this case, drug–polymer interaction is expected to form in the organic solvent first and then be disrupted after mixing with water. This type of drug–polymer interaction is perhaps closer to the drug–polymer interactions formed within traditional amorphous solid dispersions, providing the relevance of this experimental approach for LLPS detection. To further investigate the impacts of the preformed drug–polymer interactions on the LLPS onset concentration in PBS media, organic solutions of drug–polymer systems were first prepared (codissolving method) . In this approach, polymers and the CXB were codissolved in the MeOH, with a weight ratio of 2:1. 50 μg/mL of CXB and 100 μg/mL of polymers were expected at the end of the experiment. The stock solution mixing rate was set to 1 mL/h by the two methods. The influence of the polymer concentration on the LLPS onset concentration was negligible in this section due to the absence of a marked impact at low polymer concentrations. Extinction profiles of the CXB–polymer–water ternary system with 100 μg/mL of polymers introduced through the codissolving approach are depicted in . LLPS onset concentrations of the ternary system were calculated by using the intersection point of regression curves. LLPS onset concentrations of solutions with polymers predissolved in the PBS (pH 7.4) or codissolved with the CXB in the drug stock solution at 37 °C are shown in and . Blue bars represent the LLPS onset concentration of systems with predissolved polymers. Orange bars represent values calculated from systems when the CXB–polymer–MeOH stock solution was introduced into the pure PBS buffer. The CXB–polymer binary interaction was formed in a drug–polymer codissolving system before the organic droplet was dispersed into the water . The CXB concentrations at the LLPS point derived from the drug–polymer codissolving system were higher than those derived from the polymer–PBS predissolved method. In the case of a predissolved system, CXB needed to compete with water to interact with polymers. Notably, the LLPS onset concentration of systems in predissolved HPMCAS aqueous solution was significantly lower than the codissolved system, estimated to be 19.5 ± 3.22 and 31.2 ± 0.196 μg/mL, respectively. Similarly, the CXB LLPS onset concentration for the HPMCP codissolved system is indeed higher than the predissolved system, measured to be 44.0 ± 4.40 and 32.4 ± 4.29 μg/mL. The results suggested that the order of drug–polymer interaction is important in influencing the LLPS onset concentration of hydrophobic systems. However, this conclusion seems to not work in the hydrogen bonding-present systems (CXB–PVPVA and CXB–PVP systems). No notable difference has been observed in these samples ( p > 0.03). The drug–polymer interaction type may be another critical factor influencing the LLPS onset concentration. The hydrophobic interaction between CXB and polymers HPMCAS and HPMCP was encouraged via the codissolving method, where MeOH is the main medium. Hydrophobic interaction remains in the aqueous medium as MeOH diffuses into the water. The ternary system stays in one phase until a higher concentration of CXB is reached. Thus, the codissolving method can yield a much higher LLPS onset point in such systems than the predissolving method. In comparison, when hydrogen bonding is the dominant cause of drug–polymer interaction, it is far easier to disrupt by the water. Thus, orders of interactions between drugs and polymers (PVP, PVPVA) in an aqueous medium do not significantly affect the presence of the LLPS onset point of the system. 3.7 Understanding the Dynamics of LLPS in a CXB–Polymer–Water Phase Diagram Polymers significantly influenced the LLPS onset concentrations when using the codissolved CXB–polymer–MeOH approach. This fact demonstrated that understanding the drug–water binary system alone is inadequate for probing the drug release kinetics of ASD formulations. Without the presence of polymer additives, the drug precipitate was formed immediately without the observation of drug-rich phases. However, in most cases, the role of polymer additives and associated methodology is rather empirical for screening the polymer additives in a supersaturation study. The binary composition–temperature phase diagram perhaps helps us to understand the LLPS onset point while considering the polymer additives. Given a scenario of the drug concentration within the aqueous medium being any point between the solubility line and binodal line ( a), it is inevitable for the system to lower its energy by reducing the drug concentration in solution, moving toward point B. For the drug concentration to successfully move toward point C, polymer additives have been used to improve the kinetics stability of the drug–water binary system. , In the case of HPMCAS as the predissolved polymeric additive in an aqueous medium, the addition of CXB effectively introduced the LLPS of the HPMCAS–water binary system at relatively low concentrations (<20 μg/mL CXB in water). It has also been repeatedly suggested that a strong drug–polymer interaction can promote a significant increase of drug solubility in aqueous solutions before reaching the LLPS onset point. To better describe these differences and highlight the role of polymeric additives in enhancing the drug’s solubility in water, a drug–polymer–water ternary phase diagram should be implemented as a routine approach ( b). Due to the limited drug and polymer concentration in aqueous solution, the axis of coordinate was adjusted to highlight the region of interest with the binodal curve (LLPS onset points); volume fraction scales between 0 and 0.1 for drug, 0–0.1 for polymer, and 0.5 and 1 for water. Typically, drug–polymer systems with a strong interaction have a smaller binodal region and vice versa. The blue and green curves represent the binodal lines of the CXB–PVPVA and CXB–PVP systems, where the LLPS occurred at points D 1 and E 1 , respectively. The ternary phase diagram estimated that the drug volume fraction at the LLPS onset point of the system with a weak drug–polymer interaction was lower than that of a system with a strong interaction (φ D 1 < φ E 1 ). A higher apparent drug volume fraction can be reached in an aqueous solution with a system that has a stronger CXB–polymer interaction. As we observed in this study, the polymer concentration did not influence the LLPS onset point. Purple and red arrows represent the LLPS routes with the different predissolved polymer concentrations. Two arrows intersected with the blue line at points D 1 and D 2 and the green line at points E 1 and E 2 . For a given system, drug weight fractions of the drug-lean phases were very close to each other when altering the polymer concentration, where φ D 1 ≈ φ D 2 , φ E 1 ≈ φ E 2 . The shape of the curves in the phase diagram suggested that the effects of polymer concentrations within the system may not lead to significant changes in the drug concentration. In terms of the ASD dissolution, this shape of the binodal line suggests that the drug–polymer ratio of ASD may not potentially impact the LLPS onset concentration, thus limiting the solubility enhancement. A similar observation has also been reported in the literature in which the LLPS onset concentration of paclitaxel was not changed when increasing the HPMCAS concentration from 32 to 450 μg/mL. Drug–polymer interaction approaches also play an important role in LLPS. For systems with the water-resistant hydrophobic interaction, i.e., CXB–HPMCAS and CXB–HPMCP, it has been found that the determined CXB LLPS onset concentration from the codissolving approach was higher than that of the predissolving approach. However, for systems formed with water-sensitive hydrogen bonding, i.e., CXB–PVP and CXB–PVPVA, the LLPS onset concentrations were not significantly altered by the two different mixing approaches. Such a phenomenon was demonstrated in the ternary phase diagram, as illustrated in c, where red and black arrows represented predissolving and codissolving mixing approaches. For the codissolving scheme, hydrophobic interactions between CXB and HPMCP or HPMCAS were revealed to form in the methanol, resulting in a smaller binodal region. These systems will separate at point H with a drug volume fraction of φ H . For the predissolving scheme, similar hydrophobic interaction was harder to form in the aqueous solution, and the LLPS can be estimated at point F and lead to a smaller drug volume fraction of φ F (φ F < φ H ). In comparison, for a hydrogen bonding dominant system (CXB–PVP/PVPVA), the drug–polymer interaction was disrupted by water, irrespective of the different mixing approaches, resulting in a similar LLPS onset concentration (φ G ≈ φ F ). In the current ASD design and development framework, the polymer property was revealed as a critical factor due to its contributions to miscibility, stability, and drug release performance. − This work clarified that the presence of a polymer could also alter the dynamics of LLPS and the maximum achievable free drug concentration. In a standard dissolution study, the apparent solubility/concentration is always determined to assess the drug release performance. However, the concentration of the free drug without forming a complex with excipients was revealed to be the real driving force for improving drug absorption. Polymers that strongly interact with the drug will increase the LLPS onset concentration and the maximum achievable free drug concentration. When the drug concentration subsequently exceeds the LLPS onset concentration, the drug-rich phases are expected to reserve excess drugs, further facilitating drug absorption through the membrane. Solution 1 H NMR Spectroscopy All solution NMR experiments were conducted in a nonaqueous environment to emphasize the drug–polymer interactions that may be more relevant to the ASD solids before rehydration . A nonaqueous solution in the NMR experiment can help differentiate the drug–polymer interactions for the drug–polymer–water ternary system before rehydration. It is important to investigate the relationship between the CXB–polymer interaction at a nonaqueous environment first and the subsequent dynamics of the LLPS. As suggested in this work, water can easily disrupt water senstive intermolecular interactions, such as the H-bonding between the drug and polymer. In contrast, hydrophobic interactions between the drug and the polymer are less affected during the dissolution of ASDs. , We argue that if the drug–polymer interactions are mainly hydrophobic, then the influence of water on the LLPS onset point may be different from systems with other types of drug–polymer interactions. The proton NMR results suggested that hydrogen bonding was formed within CXB–PVP and CXB–PVPVA combinations, supported by the downshift of hydrogens on the –NH 2 ( a–c, red labels). The hydrogen atoms on the –NH 2 group shifted to higher parts per million (ppm) in the presence of PVP, suggesting stronger hydrogen bonding between CXB and PVP than between CXB and PVPVA. This result was also reported previously in the literature. For example, IR spectroscopy indicated that the –VA groups of the PVPVA will not interact with CXB when forming ASDs. The critical interaction differences between CXB–PVP and CXB–PVPVA may also be highlighted by the level of determined drug–polymer glass transition temperatures deviating from the Gordon–Taylor equation predictions. Rask et al. found that the ASD of CXB–PVP has a more extended deviation of the glass transition temperatures rather than the CXB–PVPVA system, suggesting the CXB–PVP system generated a stronger interaction than the latter one. In comparison, hydrophobic interactions are the main form of intermolecular interactions for CXB–HPMCAS and CXB–HPMCP systems, as evidenced by the appearance of the lower ppm position of the hydrogen atoms in the pyrazole ring of the CXB ( d,e, blue labels). Similar upfield shifts have also been reported in the cases of CXB/hydroxypropyl-β-cyclodextrin and CXB/2,6-di-O-methyl-β-cyclodextrin systems. The upfield shift is caused by the shielding effect of oxygen atoms on the excipients, suggesting the CXB pyrazole ring was a critical hydrophobic group to interact with HPMCAS and HPMCP polymers in a nonaqueous environment. CXB–Polymer–Water Ternary Phase Diagrams CXB–polymer–water ternary phase diagrams for all four polymeric additives were constructed using previously established methods . The parameters such as molecular volume, density, solubility, and F–H interaction parameters are all provided in the Supporting Information . Within the diagram, the binodal and spinodal curves were plotted to highlight the phase behaviors of the CXB–polymer–water mixture. The ternary phase diagram illustrates the relevant compositions of the CXB supersaturated solutions with polymeric additives. However, given the limited resolution of the modeling tool used and the relatively small concentrations of the CXB in water, these ternary phase diagrams were not established using the existing set of F–H interaction parameters. Instead, four ternary phase diagrams were successfully constructed with lower values of the F–H interaction parameters ( Figure S2 ). Nevertheless, this approach illustrates the importance of polymeric additives for all four ternary systems at high polymer or water compositions. The locations of the LLPS boundary indicate the formation of droplets during this process. With these boundaries, the spinodal curves sit within the binodal curves, showing the limited local stability of the mixture in these compositions. The gaps between the spinodal and binodal curves are identified as the miscibility gap, highlighting the metastability nature of the LLPS. Within the miscibility gap, it is, therefore, most likely that the appearance of the CXB–polymer–water transient phases can be observed using various analytical techniques. The starting point of such an LLPS is understood to be very close to the binodal curve of the phase diagram; it is named the LLPS onset concentration in this work. The binodal curve defines the temperature and composition of the mixture at which phase separation is thermodynamically favorable. Following the positions of the binodal curves in all four systems in and S2 , the identifiable areas are located at the three corners of the ternary phase diagram, e.g., the high polymer, high water, and high drug compositions. At these locations, the influences of the other two lower components may be illustrated by using the miscibility gaps. For example, high polymer and low drug composition areas indicate the possible phase behaviors of the drug–polymer ASD systems in water, as shown in the top areas in . A homogeneous CXB–polymer ASD system may gradually move into the unstable LLPS when exposed to moisture during dissolution. Indeed, identifying the miscibility gaps for ternary systems has been widely used to develop formulations with nanoscale artifacts for many important commercial applications. , Furthermore, it is observed that the shape of the miscibility gap is highly influenced by the polymeric additives in the CXB-water system, reflecting the differences in drug–polymer and water–polymer interaction parameters ( Table S1 ). The wide miscibility gap from selecting appropriate polymers, e.g., HPMCAS, HPMCP, and PVPVA, is expected to increase the drug concentrations and the extent of supersaturation for CXB in the aqueous medium (LLPS). In comparison, the miscibility gap for the CXB–PVP–water system is relatively small if the values of interaction parameters are reduced proportionally ( Table S1 ). The resulting ternary phase diagram appeared to have similar miscibility gaps at high water and low drug/polymer compositions, e.g., the supersaturated drug–water/polymer–water solutions. Although these constructed ternary phase diagrams are based on theoretical parameters derived using solubility parameters, the trend of such changes in the miscibility gaps reveals a reasonable outcome for the drug–polymer–water systems. With such a simple approach, one can quickly screen the polymeric additives for a given drug molecule for solubility enhancements and the likelihood of LLPS. In this work, the UV extinction measurement investigates the phase behaviors of supersaturated CXB solutions in the presence of various polymeric excipients. The supersaturated drug solution is achieved using the antisolvent/solvent-shifting method. Typically, the hydrophobic drug is dissolved in an organic solvent. The drug stock solution is then gradually added to the aqueous environment, where drugs reach the desirable supersaturation level until the LLPS occurs. This method hypothesizes that the limited fraction of organic solvent will not influence the phase behaviors of the system. The polar organic solvent in emulsion droplets will immediately disperse into the water. However, it is worth noting that due to the nature of this measurement, the appearance of the polymer–water phase separation can also cause UV extinction. Factors Influencing the Detection of LLPS Onset Concentrations Poorly water-soluble drug molecules exposed to an aqueous environment tend to precipitate through either the classical or nonclassical nucleation pathway. UV spectroscopy is one of the most commonly used techniques to characterize the drug-rich phases generated in supersaturated solutions. Typically, when electromagnetic radiation goes through a solution, the radiation may attenuate or become extinct due to absorption or scattering. , The absorption occurs through the electrons of molecules in a solution, absorbing energy from radiation and expanding to a higher energy state. The scattering will be observed when insoluble particles are presented in the solution, where larger particles exhibit a higher level of scattering. , In the drug–water systems, the particle scattering due to drug-rich phase generation was reported to lead to a spectrum baseline distortion, forming the foundation of this measurement. A high UV extinction is commonly observed at the high absorption regions of the drug molecule, ranging from 200 to 400 nm. PVP, PVPVA, and HPMCAS showed no absorption peak at wavelengths larger than 240 nm. HPMCP PBS solution had a notable peak at a wavelength of 292 nm. To minimize the UV extinction caused by polymer absorption, a wavelength of 360 nm was selected for all experiments. UV extinction spectra of CXB–PVPVA in PBS and MeOH solutions were verified at all relevant compositions with a 10–50 μg/mL drug concentration range in a 10 μg/mL interval ( Figure S2 ). A clear step change in the relationship between drug concentrations and UV extinctions could be observed, indicating partial changes in the physical forms of the drug molecules within the solution. The occurrence of additional UV extinction was interpreted as the light scattering caused by the particles in high drug concentration samples, known as the Tyndall scattering. Therefore, the drug-rich phase onset point of the CXB–PVPVA–PBS ternary system could be estimated to be 28 μg/mL of CXB concentration. First, without the presence of polymer, the UV extinction data of pure CXB solution as a function of CXB concentration at 37 °C are shown in , with the drug stock solution pumping rates at 3, 1, and 0.5 mL/h. A wavelength of 360 nm was used to estimate the phase separation onset concentration of the CXB solution. To avoid interferences, the wavelength employed to determine the scattering intensity was far from the drug’s absorption wavelength range. The red lines were two least-squares regression curves fitted using data points at the initial stage and the subsequent appearance of supersaturated solutions with the increased CXB. The onset concentration of LLPS was derived from the intersection point of two red lines. The phase separation occurrence concentrations were determined as 7.75 ± 0.78, 8.58 ± 1.95, and 6.22 ± 1.07 μg/mL with the drug stock solution pumping rates of 3, 1, and 0.5 mL/h, respectively. No notable difference was observed with the drug stock solution pumping rate ( p > 0.05). The drug concentration at the solution’s optical turbidity point has also been highly associated with the drug’s theoretical amorphous solubility. However, the experimentally detected phase separation concentration of pure CXB was remarkably lower than the theoretically calculated drug amorphous solubility. Using various theories, the CXB amorphous solubility was estimated to be 19–22.6 μg/mL. It was suggested that the fast recrystallization speed of the CXB has resulted in the formation of small crystalline drug particles before the LLPS. In this work, white precipitations could also be observed by the optical image of the pure CXB–PBS solution in a matter of minutes . Crystalline drug suspension in the solution may contribute to the overall UV spectrum scattering and lead to the step change at the slopes. The inline UV method can be affected by the appearance of all kinds of matters in the CXB titration experiment, such as the nanocrystalline, nanodroplet, amorphous nanoparticles or mixtures of all of the above. We also tried to pause the titration when the LLPS onset point was detected. The UV extinction value remained stable for at least 20 min, suggesting the number and size of the metastable particles were stable within the individual test. Additional experiments were also conducted to highlight the differences in UV extinction values caused by the crystalline CXB and LLPS ( Figure S3 ). Seeding the CXB–PVP–water solution at a CXB concentration of 36 μg/mL with an additional 10% w/w crystalline CXB, a significant jump in the UV extinction was observed, which was higher than that caused by the LLPS. To further identify the compositions of the nanosized matter at the onset point of LLPS, the cryo-EM technique and high-resolution TIRFM (with pyrene as the hydrophobic fluorescence probe, Supporting Information Videos 1–4 ) were used. The cryo-EM and TIRFM highlighted the appearance of spherical particles ranging from 40 nm to several micrometers. Particularly in the TIRFM with polarized filter videos, no significant birefringence was observed, indicating the possible aggregation of the CXB amorphous nanodroplet following the initial LLPS at much smaller sizes . The appearance of noncrystalline nano/microparticles with the CXB–polymer–water suspension highlighted the metastability nature of the mixture, thus validating that the main cause of the UV extinction is indeed attributed to the LLPS. It should be noted that several previous research articles presented the behaviors of LLPS for CXB in PBS with the predissolved polymeric matrices, such as PVP, PVPVA, and HPMCAS. , , , These values suggest that under the experimental conditions described in this work (microfluidic pump and mixing), the onset of UV baseline change should be mainly attributed to LLPS in the CXB titration process (all experiments at various pumping rates were completed within 1 h). As mentioned, phase behaviors, e.g., LLPS in the drug–polymer–water ternary systems, were revealed to affect the solubility and permeability enhancement of ASD during oral administration. However, the polymer influence on the transient drug-rich phases is not fully understood. This section investigated various experimental conditions based on the onset point of LLPS for CXB–polymer–water ternary systems, including polymer types, polymer concentrations, drug–polymer interaction approaches, and drug–polymer mixing rates. Effects of Polymer Types and Drug Stock Solution Pumping Rates Given the usual low LLPS onset concentrations of most poorly water-soluble drugs in the aqueous medium, polymeric excipients are often predissolved to suppress the precipitations during the experiment. In this section, different polymers of PVP, PVPVA, HPMCAS, and HPMCP with a concentration of 1 mg/mL were predissolved in the PBS buffer to inhibit the precipitations of CXB. The phase behaviors of drug–polymer solutions were monitored using the same UV extinction method. The polymer type is a critical parameter that influences the LLPS onset concentration in a CXB–polymer–water ternary system, as the UV extinction profiles depicted in depend on the drug concentrations in the medium. Compared with a pure CXB–PBS solution, it is clear that the LLPS onset concentrations have been altered for CXB when different polymeric materials are predissolved within the aqueous medium. The LLPS data varied in different polymer–water combinations. The LLPS onset concentrations for CXB ternary systems with predissolved PVP ( a) and HPMCP ( d) were recorded at 37.2 ± 0.77 and 37.6 ± 0.98 μg/mL, respectively. CXB ternary systems with polymers of PVPVA ( b) and HPMCAS ( c) exhibited lower LLPS onset concentrations, measured to be 18.0 ± 1.82 and 15.1 ± 1.33 μg/mL. In a phase diagram, drug-rich phases were suggested to be generated in the metastable region. Drug-rich phases could be determined when these transient phases reach the local minimum energy position and are kinetically stable for a short period. The kinetic influence on the determination of the LLPS point was studied when the mixing rate of the stock solution and the PBS buffer was changed. This work suggested that the drug–polymer mixing rate is not a significant parameter of the LLPS onset concentration. Specifically, examples of UV extinction profiles of the CXB–polymer–water ternary system with drug stock solution pumping rates of 3 and 0.5 mL/h at the wavelength of 360 nm are shown in and ; additional profiles of the UV extinction at a pumping rate of 1 mL/h are provided in Figure S3 . To further verify the existence of metastable CXB–polymer–water transient phases, cryo-EM microscopic analysis was carried out for several selected samples after reaching LLPS onset points . Immediately after reaching the LLPS onset points, the liquids were drawn from the sample vials and rapidly frozen to achieve amorphous ice for cryo-EM. Round-shaped condensed matter with sizes of 20–80 nm was observed in all CXB–polymer–water systems. The amount of round-shaped condensed matter may indicate the CXB LLPS onset concentrations, where more spherical particles were observed in HPMCP and PVP-based systems than in HPMCAS and PVPVA mixtures. Furthermore, signs of agglomeration were also observed in HPMCAS and HPMCP-based CXB suspensions, reflecting the possible colloidal nature of these two polymeric matrices. LLPS onset concentrations in the PBS solution with or without polymers using various mixing rates are summarized in and . Values were calculated individually at different conditions. Error bars were derived from standard deviations of those values. Blue, orange, and yellow bars represent the drug stock solution (4 mg/mL) pumping rates of 3, 1, and 0.5 mL/h, respectively. 50 μg/mL CXB solutions were generated in 5, 15, and 30 min, respectively. CXB LLPS in PVP and HPMCP aqueous solutions exhibited significantly higher concentrations than those in PVPVA and HPMCAS solutions ( p < 0.0001). No significant difference in the LLPS onset concentrations was observed when drug stock solution mixing rates were altered (two-way ANOVA, p > 0.05). The result suggested that the CXB LLPS point in the CXB–polymer–water ternary systems had a less kinetic influence within the first 30 min of the experiments. The presence of polymers influenced the LLPS onset concentrations remarkably by controlling the position of the binodal line. Samples with strong drug–polymer interactions were observed to undergo LLPS at high drug concentrations. For example, hydrogen bonding was identified within both CXB–PVP and CXB–PVPVA systems in the nonaqueous situation by 1 H NMR spectra. At the mixing rate of 3 mL/h, the LLPS onset concentration for CXB at a 1 mg/mL PVP–PBS solution was approximately two times higher than that of the PVPVA solution, . A higher CXB LLPS onset value of the CXB–PVP system was interpreted by the stronger hydrogen bonding of the drug and polymer, evidenced by a more extensive chemical shift in 1 H NMR spectra. Similarly, PVPVA and HPMCAS were reported to reduce the LLPS onset concentrations in other drug systems. It was suggested that the ibuprofen solubility was reduced in several polymer solutions, including the PVPVA, and that the bulk ibuprofen concentration was reduced with PVPVA. Miao et al. reported that the LLPS value of paclitaxel decreased from approximately 40–23 μg/mL with the HPMCAS (MF). Role of Polymer Concentrations on the LLPS Onset The drug–polymer composition has been commonly highlighted to influence the LLPS point in supersaturated drug-water solutions. This work determined the CXB LLPS onset concentrations of several CXB–polymer–water ternary systems with different polymer concentrations. UV extinction profiles as a function of drug concentration with the polymer concentrations at 500 and 100 μg/mL are shown in and . The LLPS onset concentrations at different systems were derived from the step change of regression curve slopes (red lines). Various LLPS onset concentrations observed for the CXB–polymer–water ternary system are summarized in and . Blue, orange, and yellow bars represent the polymer concentration at 1000, 500, and 100 μg/mL, respectively. In this case, CXB LLPS onset concentrations in the ternary systems were usually not altered when reducing the polymer concentration from 1000 μg/mL to 100 μg/mL ( p > 0.05), as shown in and . However, the system with HPMCAS exhibited an abnormal LLPS concentration at a higher polymer concentration. The CXB LLPS concentration of the solution with the HPMCAS concentration of 1000 μg/mL was 12.7 ± 5.72 μg/mL. This value increased to 21.1 ± 2.27 μg/mL when the polymer concentration decreased to 500 μg/mL and then remained constant between 500 and 100 μg/mL. The UV spectra of HPMCAS PBS solution (pH = 7.4) with a serial of polymer concentrations at 37 °C are illustrated in b, where gray, orange, and blue curves represent HPMCAS concentrations of 1000, 500, and 100 μg/mL, respectively. Scattering was the sole factor contributing to the overall extinction at the LLPS determination wavelength (360 nm). It should be noted that the scattering was already observed in the HPMCAS solution at a concentration of 1000 μg/mL without the addition of CXB. This was due to the HPMCAS aggregated upon high polymer concentrations in the PBS solution (37 °C), where HPMCAS–PBS demixing occurred even without drug molecules (raised baseline in b). Similar observations on the nature of colloid formation for HPMCAS at high concentrations have already been reported in the literature. , To maintain the consistency of the experimental conditions among all polymeric carriers, the influence of HPMCAS aggregation was blanked out from the UV extinction before addition of the CXB stock solution. However, the results suggested that only approximately 12.7 μg/mL CXB was required to disrupt the existing HPMCAS aggregations in the PBS solution, resulting in a phase-separated CXB–HPMCAS colloid suspension in the PBS. Influence of Preformed Drug–Polymer Interaction in Stock CXB Solution The LLPS concept hypothesizes that the complexity of the free energy landscape can alter the dynamics of the resulting transient phase. In this case, the rate of mixing of CXB with water and the presence of polymeric carriers should be expected to alter the resulting LLPS onset point. The Flory–Huggins model is an important theoretical approach for estimating the phase boundaries in polymer-relevant solutions, which models the interaction of the components within a lattice theory. , The entropic contribution to the free energy landscape of the system is determined by enumerating the distinct configurations of molecules and polymers within the lattice. In contrast, according to a regular solution theory, the enthalpic contribution arises from the paired interaction energies between the components. The component interaction is a dominating parameter influencing the free energy landscape and the LLPS. Previous sections studied the weaker interaction between the drug and various polymers, which can result in a lower LLPS onset concentration. Such alteration of drug–polymer interaction can also be complicated by moving the interaction from a nonaqueous state to an aqueous state. Chen et al. suggested that the drug–polymer intermolecular interaction strength in a nonaqueous environment may be weaker than in an aqueous solution. Marsac et al. found the hydrogen bonding between felodipine and PVP will be disrupted with the introduction of water. Fundamentally, the experimental approach to obtain the LLPS of a small molecule drug in water is via solvent shift, where a drug-organic solvent solution is gradually added into a polymer–water solution. Quick diffusion of the organic solvent in water results in phase separation of the drug solution due to the poor water solubility. In this situation, the drug–polymer interaction is expected to form a competitive relationship with the water–polymer interactions. In comparison, the LLPS of the drug–polymer–water bond can also be obtained by adding a drug–polymer organic solvent solution into the water medium. However, in this case, drug–polymer interaction is expected to form in the organic solvent first and then be disrupted after mixing with water. This type of drug–polymer interaction is perhaps closer to the drug–polymer interactions formed within traditional amorphous solid dispersions, providing the relevance of this experimental approach for LLPS detection. To further investigate the impacts of the preformed drug–polymer interactions on the LLPS onset concentration in PBS media, organic solutions of drug–polymer systems were first prepared (codissolving method) . In this approach, polymers and the CXB were codissolved in the MeOH, with a weight ratio of 2:1. 50 μg/mL of CXB and 100 μg/mL of polymers were expected at the end of the experiment. The stock solution mixing rate was set to 1 mL/h by the two methods. The influence of the polymer concentration on the LLPS onset concentration was negligible in this section due to the absence of a marked impact at low polymer concentrations. Extinction profiles of the CXB–polymer–water ternary system with 100 μg/mL of polymers introduced through the codissolving approach are depicted in . LLPS onset concentrations of the ternary system were calculated by using the intersection point of regression curves. LLPS onset concentrations of solutions with polymers predissolved in the PBS (pH 7.4) or codissolved with the CXB in the drug stock solution at 37 °C are shown in and . Blue bars represent the LLPS onset concentration of systems with predissolved polymers. Orange bars represent values calculated from systems when the CXB–polymer–MeOH stock solution was introduced into the pure PBS buffer. The CXB–polymer binary interaction was formed in a drug–polymer codissolving system before the organic droplet was dispersed into the water . The CXB concentrations at the LLPS point derived from the drug–polymer codissolving system were higher than those derived from the polymer–PBS predissolved method. In the case of a predissolved system, CXB needed to compete with water to interact with polymers. Notably, the LLPS onset concentration of systems in predissolved HPMCAS aqueous solution was significantly lower than the codissolved system, estimated to be 19.5 ± 3.22 and 31.2 ± 0.196 μg/mL, respectively. Similarly, the CXB LLPS onset concentration for the HPMCP codissolved system is indeed higher than the predissolved system, measured to be 44.0 ± 4.40 and 32.4 ± 4.29 μg/mL. The results suggested that the order of drug–polymer interaction is important in influencing the LLPS onset concentration of hydrophobic systems. However, this conclusion seems to not work in the hydrogen bonding-present systems (CXB–PVPVA and CXB–PVP systems). No notable difference has been observed in these samples ( p > 0.03). The drug–polymer interaction type may be another critical factor influencing the LLPS onset concentration. The hydrophobic interaction between CXB and polymers HPMCAS and HPMCP was encouraged via the codissolving method, where MeOH is the main medium. Hydrophobic interaction remains in the aqueous medium as MeOH diffuses into the water. The ternary system stays in one phase until a higher concentration of CXB is reached. Thus, the codissolving method can yield a much higher LLPS onset point in such systems than the predissolving method. In comparison, when hydrogen bonding is the dominant cause of drug–polymer interaction, it is far easier to disrupt by the water. Thus, orders of interactions between drugs and polymers (PVP, PVPVA) in an aqueous medium do not significantly affect the presence of the LLPS onset point of the system. Understanding the Dynamics of LLPS in a CXB–Polymer–Water Phase Diagram Polymers significantly influenced the LLPS onset concentrations when using the codissolved CXB–polymer–MeOH approach. This fact demonstrated that understanding the drug–water binary system alone is inadequate for probing the drug release kinetics of ASD formulations. Without the presence of polymer additives, the drug precipitate was formed immediately without the observation of drug-rich phases. However, in most cases, the role of polymer additives and associated methodology is rather empirical for screening the polymer additives in a supersaturation study. The binary composition–temperature phase diagram perhaps helps us to understand the LLPS onset point while considering the polymer additives. Given a scenario of the drug concentration within the aqueous medium being any point between the solubility line and binodal line ( a), it is inevitable for the system to lower its energy by reducing the drug concentration in solution, moving toward point B. For the drug concentration to successfully move toward point C, polymer additives have been used to improve the kinetics stability of the drug–water binary system. , In the case of HPMCAS as the predissolved polymeric additive in an aqueous medium, the addition of CXB effectively introduced the LLPS of the HPMCAS–water binary system at relatively low concentrations (<20 μg/mL CXB in water). It has also been repeatedly suggested that a strong drug–polymer interaction can promote a significant increase of drug solubility in aqueous solutions before reaching the LLPS onset point. To better describe these differences and highlight the role of polymeric additives in enhancing the drug’s solubility in water, a drug–polymer–water ternary phase diagram should be implemented as a routine approach ( b). Due to the limited drug and polymer concentration in aqueous solution, the axis of coordinate was adjusted to highlight the region of interest with the binodal curve (LLPS onset points); volume fraction scales between 0 and 0.1 for drug, 0–0.1 for polymer, and 0.5 and 1 for water. Typically, drug–polymer systems with a strong interaction have a smaller binodal region and vice versa. The blue and green curves represent the binodal lines of the CXB–PVPVA and CXB–PVP systems, where the LLPS occurred at points D 1 and E 1 , respectively. The ternary phase diagram estimated that the drug volume fraction at the LLPS onset point of the system with a weak drug–polymer interaction was lower than that of a system with a strong interaction (φ D 1 < φ E 1 ). A higher apparent drug volume fraction can be reached in an aqueous solution with a system that has a stronger CXB–polymer interaction. As we observed in this study, the polymer concentration did not influence the LLPS onset point. Purple and red arrows represent the LLPS routes with the different predissolved polymer concentrations. Two arrows intersected with the blue line at points D 1 and D 2 and the green line at points E 1 and E 2 . For a given system, drug weight fractions of the drug-lean phases were very close to each other when altering the polymer concentration, where φ D 1 ≈ φ D 2 , φ E 1 ≈ φ E 2 . The shape of the curves in the phase diagram suggested that the effects of polymer concentrations within the system may not lead to significant changes in the drug concentration. In terms of the ASD dissolution, this shape of the binodal line suggests that the drug–polymer ratio of ASD may not potentially impact the LLPS onset concentration, thus limiting the solubility enhancement. A similar observation has also been reported in the literature in which the LLPS onset concentration of paclitaxel was not changed when increasing the HPMCAS concentration from 32 to 450 μg/mL. Drug–polymer interaction approaches also play an important role in LLPS. For systems with the water-resistant hydrophobic interaction, i.e., CXB–HPMCAS and CXB–HPMCP, it has been found that the determined CXB LLPS onset concentration from the codissolving approach was higher than that of the predissolving approach. However, for systems formed with water-sensitive hydrogen bonding, i.e., CXB–PVP and CXB–PVPVA, the LLPS onset concentrations were not significantly altered by the two different mixing approaches. Such a phenomenon was demonstrated in the ternary phase diagram, as illustrated in c, where red and black arrows represented predissolving and codissolving mixing approaches. For the codissolving scheme, hydrophobic interactions between CXB and HPMCP or HPMCAS were revealed to form in the methanol, resulting in a smaller binodal region. These systems will separate at point H with a drug volume fraction of φ H . For the predissolving scheme, similar hydrophobic interaction was harder to form in the aqueous solution, and the LLPS can be estimated at point F and lead to a smaller drug volume fraction of φ F (φ F < φ H ). In comparison, for a hydrogen bonding dominant system (CXB–PVP/PVPVA), the drug–polymer interaction was disrupted by water, irrespective of the different mixing approaches, resulting in a similar LLPS onset concentration (φ G ≈ φ F ). In the current ASD design and development framework, the polymer property was revealed as a critical factor due to its contributions to miscibility, stability, and drug release performance. − This work clarified that the presence of a polymer could also alter the dynamics of LLPS and the maximum achievable free drug concentration. In a standard dissolution study, the apparent solubility/concentration is always determined to assess the drug release performance. However, the concentration of the free drug without forming a complex with excipients was revealed to be the real driving force for improving drug absorption. Polymers that strongly interact with the drug will increase the LLPS onset concentration and the maximum achievable free drug concentration. When the drug concentration subsequently exceeds the LLPS onset concentration, the drug-rich phases are expected to reserve excess drugs, further facilitating drug absorption through the membrane. Conclusions The LLPS onset point is a critical parameter that inherently indicates the maximum free drug concentration and generation of the drug-rich phase in a supersaturated drug solution. In previous works, the LLPS and generation of the drug-rich phase were understood using the drug–water binary phase diagram. It is unclear what the role of polymers played in generating drug-rich phases. This work systematically evaluated CXB drug-rich phases in PBS solutions combined with PVP, PVPVA, HPMCAS, and HPMCP polymers using the solvent/shifting method. The strength of the drug–polymer interaction and the orders of drug–polymer interactions were revealed to alter the dynamics of LLPS. However, parameters like the polymer concentration and the mixing rate of drug and polymer were found to be less significant for the LLPS onset concentration of CXB solutions. The general phase diagram of the drug–polymer–water ternary system was utilized to understand the LLPS onset points determined from various CXB–polymer supersaturated solutions. This study highlighted the importance of polymers in generating the metastable drug-rich transient phases by implementing LLPS of the ternary phase diagram. |
PD-L2 Expression in Breast Cancer Promotes Tumor Development and Progression | 2ce1774f-6890-4790-ae16-9f9fad33de17 | 11233179 | Anatomy[mh] | The programmed death ligand 2 (PD-L2), another member of the B7 family, was initially discovered as a gene differentially expressed between dendritic cells and activated macrophages . PD-L2 fails to bind to the cytotoxic T lymphocyte-associated antigen (CTLA)-4 and CD28, instead, it binds to PD-1, the B7-H1/PD-L1 receptor . PD-L2 expression can be induced on the surfaces of dendritic cells, macrophages, certain B cell populations, and mast cells . Moreover, it is widely suggested that PD-L2 binds to the PD-1 coinhibitory receptor to suppress immunity . Nonetheless, PD-L2 is also discovered to trigger T-cell growth and the generation of cytokines, even in non-PD-1 binding PD-L2 mutants and PD-1-deficient T-cells . Therefore, there is a general consensus that PD-L2 has a synergistic promoting or inhibiting activity. In recent years, numerous reports have revealed that PD-L2 is elevated in many cancers, including ovarian cancer , lung adenocarcinoma , gastric cancer , and esophageal squamous cell carcinoma . In contrast, only a few studies have demonstrated a connection between PD-L2 and breast cancer . Therefore, it is of significant clinical importance to explore PD-L2 expression within breast cancer. Breast cancer, a disease with molecular heterogeneity, is the most prevalently occurring malignancy among women . Early breast cancer, which is confined to the mammary gland or has only spread to the axillary lymph nodes, is regarded as treatable. The advancements in multimodal treatment have increased the likelihood of cure to ~70%–80% of cases . However, advanced (metastatic) disease cannot be cured with the existing treatments such as surgery, radiotherapy, chemotherapy, endocrine therapy, and targeted biological therapy . Thus, more effective therapeutic alternatives are requisite for breast cancer. In recent years, cancer immunotherapy has provided breakthrough therapeutics for combating cancer . Among them, anti-PD-1 or anti-PD-L1 antibodies have altered the treatment of advanced cancers, such as melanoma , lung cancer , kidney cancer , or various others . However, immunotherapy has shown limited success in breast cancer, and the effect of PD-L2 on regulating breast cancer remains largely unknown . In this study, we found that serum soluble PD-L2 (sPD-L2) concentration was elevated in breast cancer patients compared with healthy controls by enzyme-linked immunosorbent assay (ELISA), and the expression of PD-L2 was positively correlated with progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) by immunohistochemistry experiments in breast cancer patients. We also found that PD-L2 knockdown suppressed the invasion and migration of MCF-7 and MDA-MB231 cells. Furthermore, the results of the mouse xenograft tumor assay were in line with those of the in vitro cell assays. 2.1. Participants This study involved individuals from whom blood samples were obtained at the Department of General Surgery of Suzhou Municipal Hospital in Suzhou, China, during the period from January 2018 to December 2020. In addition, blood samples were also simultaneously collected from healthy blood donors to be used as controls. Tumor tissues were acquired from breast cancer patients who underwent surgery at the Department of General Surgery of Suzhou Municipal Hospital and Kunshan People's Hospital in Jiangsu Province, China, from January 2008 to December 2018, after obtaining informed consent. There were a total of 416 tumor tissues utilized in this study, and the patients had not undergone radiotherapy and chemotherapy prior to the surgery. The stage of patients was assessed according to the 8th edition of the American Joint Committee on Cancer (AJCC) staging manual . The tissues were stained with hematoxylin and eosin to confirm the pathological diagnosis. The clinical parameters were documented and can be found in . Prior to this work, the Ethics Review Board of the hospital gave its approval to our study. The tissue donors provided informed consents. 2.2. Cell Culture and Antibodies We obtained two human breast cancer cell lines, namely, MCF-7 and MDA-MB231, from the Institute of Cell Biology (Chinese Academy of Sciences, Shanghai, China) and cultured them in Dulbecco's Modified Eagle Medium (DMEM) which contained 10% fetal bovine serum (FBS; HyClone, Logan, UT, USA) along with 1% penicillin/streptomycin. For the knockdown of PD-L2 (shPD-L2), we employed the lentivirus vector T7960-1-LV3R-E11 which contains an short hairpin RNA (shRNA) sequence (GenePharma, Suzhou, China). The following oligonucleotides were utilized: shNC : 5′-TTCTCCGAACGTGTCACGT-3′; shPD-L2 : 5′-GGCCAGCATTGACCTTCAAAG-3′. Cells were incubated with the medium containing lentivirus together with 5 μ g/mL of polybrene for 48 hr and then selected with 5.0 μ g/mL puromycin for 2 weeks, and the knockdown effect of PD-L2 was validated by Western blotting. All cell lines used in this study were confirmed to be mycoplasma-free and cultivated under 37°C and 5% CO 2 conditions. The antibodies utilized in this paper are presented as follows: Anti-PD-L2 antibody (ab200377, ab187662), Anti-HER2 antibody (ab134182), and Anti-GAPDH antibody (ab181602) were procured from Abcam, Cambridge, United Kingdom. Anti-Ki67 antibody (GM7240), Anti-ER antibody (GT2056), and Anti-PR antibody (GT2160) were ordered from Gene Tech, Shanghai, China. The secondary antibody employs a goat anti-rabbit antibody with horseradish peroxidase which was obtained from Beyotime Biotechnology, Shanghai, China. 2.3. Tissue Microarray Construction For the construction of tissue microarrays, formalin-fixed and paraffin-embedded tissue blocks that contained tumor tissue were recognized by area-specialized histopathologists on hematoxylin and eosin-stained slides. Thereafter, two duplicate 0.2-mm cores were obtained from the periphery and center of the tumor and arrayed in the recipient paraffin block using the tissue puncher/arrayer (Beecher Instruments, Silver Spring, MD, USA). Subsequently, 4- μ m tissue sections were prepared and placed onto Superfrost Plus slides to conduct immunohistochemistry. 2.4. Immunohistochemistry The following is the procedure for Ki67 staining: 4- μ m sections were dewaxed in xylene, then hydrated through a series of graded ethanol baths, and rinsed in water. The activity of endogenous peroxidase (GK600510, Gene Tech) was blocked. Antigen retrieval was carried out by microwaving at full power (750 W) in citrate buffer with pH 6.0 for 10 min. The MIB-1 primary antibody (GM7240, Gene Tech) was incubated for 1 hr at room temperature. All washes and dilutions were performed using phosphate-buffered saline (PBS). Biotinylated rabbit anti-mouse immunoglobulin was applied, followed by avidin–biotin complex (GK600510, Gene Tech). Diaminobenzene (DAB, GK600510, Gene Tech) was used to develop peroxidase activity, and counterstaining was done with hematoxylin. The observer (blinded to the patient outcome) examined the stained sections using a standard light microscope with a 40x objective and a 10 × 10 eyepiece graticule. The Ki67 score was defined as the percentage of the total number of tumor cells (at least 1,000) with nuclear staining in 10 high-powered fields (40x). Estrogen receptor (ER), PR, HER2, and PD-L2 were stained using the same microwave antigen recovery staining procedure as described above. The Anti-ER antibody (GT2056, Gene Tech), Anti-PR antibody (GT2160, Gene Tech), Anti-HER2 antibody (ab134182, Abcam), and Anti-PD-L2 antibody (ab200377, Abcam) were incubated for 2 hr at room temperature. The histoscore (H-score) was used to assess ER and PR, incorporating the evaluation of the intensity of staining (0–3) and the number of cells staining (range of score 0–300). By this method, ER- or PR-positive tumors have a score of >1. For HER2 scoring, specimens were classified as positive if the immunohistochemical staining was 3+ or if the staining was 2+ and FISH (fluorescence in situ hybridization) was positive. Data on PD-L2 expression only included tumors with adequate tissue to perform immunohistochemistry. PD-L2 expression was assessed using a scoring scheme that relied on the distribution of positive tumor cells and staining intensity. In addition, the intensity factor ranging from zero (the intensity is either negative or staining, but it just exceeds the background) to two (the slide shows strong positive or dark brown staining when examined under a microscope) was used to multiply the distribution score, which estimates the percentage of cells that were positively stained. The intensity of PD-L2 immunostaining was classified as follows: the sections that scored Grade 1 were classified as low PD-L2 positivity groups, and the sections of Grade 2 were classified as groups with high PD-L2 positivity. The immunohistochemically stained sections were independently examined by two researchers who were not aware of the patients' clinicopathological features. 2.5. Western Blotting We used cold radio-Immunoprecipitation assay (RIPA) buffer with phenylmethanesulfonyl fluoride (PMSF) to homogenize cells and then centrifuged the samples to obtain the supernatants. A total of 30 μ g of proteins were separated through SDS-PAGE prior to being transferred onto the 0.22- μ m polyvinylidene fluoride (PVDF) membrane. The membrane was blocked with a blocking buffer (5% defatted milk within tris-buffered saline tween-20 (TBST) that is composed of 10 mM Tris-HCl pH 8.0, 0.05% Tween 20, and 150 mM NaCl) at ambient temperature for 1 hr. After being subjected to overnight primary antibody incubation at 4°C, the membrane was probed with horseradish peroxidase-labeled secondary antibody at ambient temperature for 1–2 hr. The enhanced chemiluminescence (ECL) Western blot detection reagents (Millipore) were utilized to visualize the membranes. 2.6. Cell Survival Assay Cells (shNC, shPD-L2) were seeded in 96-well plates at a density of 3 × 10 3 cells/well in triplicate and maintained in an incubation state overnight. Cell viability was determined with the Cell Counting Kit-8 assay (CK04, Dojindo) in accordance with specific protocols. Each clone was tested in at least three replicate experiments. The data were measured in terms of the growth percentage in comparison with the controls. The average was calculated using three or more replicates for each clone. The results were represented in the form of the growth proportion in comparison with the controls. 2.7. Wound Healing Assay Cells (shNC, shPD-L2) were plated in 6-well plates at a density of 5 × 10 5 cells per well and allowed to grow until they reached 80% confluence. The cells were scratched in a straight line using a 200- μ l sterile pipette tip, then washed three times with PBS, and cultured for 24 hr. The scratches were visualized by a fluorescence microscope at a magnification of 200x (Nikon 80I; Nikon Corporation) and analyzed using Image J software (1.80v; National Institutes of Health). 2.8. Cell Migration and Invasion Assays The Transwell migration assay was carried out by using Transwell inserts (24-well plates, Corning Costar) with a filter having an 8 μ m pore. After trypsinization into single cells, 5.0 × 10 4 cells (shNC, shPD-L2) were inoculated on top of the Transwell chamber which contained a serum-free medium. The bottom chamber was filled with 20% FBS-containing medium. After 12–16 hr, the cells were fixed with methanol on the insert membranes and stained with Giemsa solution, and nonmigrating cells on the upper surface of the membrane were gently removed. The number of migrated cells was counted in five randomly chosen fields per insert (x200) using Image J software (1.80v; National Institutes of Health). In the cell invasion assay, the upper chamber was first coated with Matrigel before cell seeding. 2.9. Tumor Xenograft Model This work obtained nude mice from Charles River in Beijing. The animal experimental protocols obtained approval from the Institutional Animal Ethics Committee of Suzhou Vocational Health College, China. The animals were randomly divided into two groups ( n = 6 for each). Mice were subjected to subcutaneous inoculation with MDA-MB231-NC and MDA-MB231-shPD-L2 cells (2 × 10 6 ) suspended in a serum-free medium (100 μ L). Approximately 1 week later, the average tumor volume was 400–500 mm 3 . Thereafter, the tumor size was determined using the Vernier caliper at 2-day intervals, and the tumor volume (mm 3 ) was calculated by multiplying the length by the square of the width and dividing by two. On the 15th day, the tumors were harvested. 2.10. Statistical Analysis All data are expressed as the mean ± SEM. Statistical analysis was carried out using GraphPad Prism 8.0 software (GraphPad Software, La Jolla, CA, USA) through Student's t -test or two-way ANOVA. The status of proteins related to the main clinical and pathologic features was compared by means of χ 2 tests and Fisher's exact test when necessary. P < 0.05 indicated statistical significance. This study involved individuals from whom blood samples were obtained at the Department of General Surgery of Suzhou Municipal Hospital in Suzhou, China, during the period from January 2018 to December 2020. In addition, blood samples were also simultaneously collected from healthy blood donors to be used as controls. Tumor tissues were acquired from breast cancer patients who underwent surgery at the Department of General Surgery of Suzhou Municipal Hospital and Kunshan People's Hospital in Jiangsu Province, China, from January 2008 to December 2018, after obtaining informed consent. There were a total of 416 tumor tissues utilized in this study, and the patients had not undergone radiotherapy and chemotherapy prior to the surgery. The stage of patients was assessed according to the 8th edition of the American Joint Committee on Cancer (AJCC) staging manual . The tissues were stained with hematoxylin and eosin to confirm the pathological diagnosis. The clinical parameters were documented and can be found in . Prior to this work, the Ethics Review Board of the hospital gave its approval to our study. The tissue donors provided informed consents. We obtained two human breast cancer cell lines, namely, MCF-7 and MDA-MB231, from the Institute of Cell Biology (Chinese Academy of Sciences, Shanghai, China) and cultured them in Dulbecco's Modified Eagle Medium (DMEM) which contained 10% fetal bovine serum (FBS; HyClone, Logan, UT, USA) along with 1% penicillin/streptomycin. For the knockdown of PD-L2 (shPD-L2), we employed the lentivirus vector T7960-1-LV3R-E11 which contains an short hairpin RNA (shRNA) sequence (GenePharma, Suzhou, China). The following oligonucleotides were utilized: shNC : 5′-TTCTCCGAACGTGTCACGT-3′; shPD-L2 : 5′-GGCCAGCATTGACCTTCAAAG-3′. Cells were incubated with the medium containing lentivirus together with 5 μ g/mL of polybrene for 48 hr and then selected with 5.0 μ g/mL puromycin for 2 weeks, and the knockdown effect of PD-L2 was validated by Western blotting. All cell lines used in this study were confirmed to be mycoplasma-free and cultivated under 37°C and 5% CO 2 conditions. The antibodies utilized in this paper are presented as follows: Anti-PD-L2 antibody (ab200377, ab187662), Anti-HER2 antibody (ab134182), and Anti-GAPDH antibody (ab181602) were procured from Abcam, Cambridge, United Kingdom. Anti-Ki67 antibody (GM7240), Anti-ER antibody (GT2056), and Anti-PR antibody (GT2160) were ordered from Gene Tech, Shanghai, China. The secondary antibody employs a goat anti-rabbit antibody with horseradish peroxidase which was obtained from Beyotime Biotechnology, Shanghai, China. For the construction of tissue microarrays, formalin-fixed and paraffin-embedded tissue blocks that contained tumor tissue were recognized by area-specialized histopathologists on hematoxylin and eosin-stained slides. Thereafter, two duplicate 0.2-mm cores were obtained from the periphery and center of the tumor and arrayed in the recipient paraffin block using the tissue puncher/arrayer (Beecher Instruments, Silver Spring, MD, USA). Subsequently, 4- μ m tissue sections were prepared and placed onto Superfrost Plus slides to conduct immunohistochemistry. The following is the procedure for Ki67 staining: 4- μ m sections were dewaxed in xylene, then hydrated through a series of graded ethanol baths, and rinsed in water. The activity of endogenous peroxidase (GK600510, Gene Tech) was blocked. Antigen retrieval was carried out by microwaving at full power (750 W) in citrate buffer with pH 6.0 for 10 min. The MIB-1 primary antibody (GM7240, Gene Tech) was incubated for 1 hr at room temperature. All washes and dilutions were performed using phosphate-buffered saline (PBS). Biotinylated rabbit anti-mouse immunoglobulin was applied, followed by avidin–biotin complex (GK600510, Gene Tech). Diaminobenzene (DAB, GK600510, Gene Tech) was used to develop peroxidase activity, and counterstaining was done with hematoxylin. The observer (blinded to the patient outcome) examined the stained sections using a standard light microscope with a 40x objective and a 10 × 10 eyepiece graticule. The Ki67 score was defined as the percentage of the total number of tumor cells (at least 1,000) with nuclear staining in 10 high-powered fields (40x). Estrogen receptor (ER), PR, HER2, and PD-L2 were stained using the same microwave antigen recovery staining procedure as described above. The Anti-ER antibody (GT2056, Gene Tech), Anti-PR antibody (GT2160, Gene Tech), Anti-HER2 antibody (ab134182, Abcam), and Anti-PD-L2 antibody (ab200377, Abcam) were incubated for 2 hr at room temperature. The histoscore (H-score) was used to assess ER and PR, incorporating the evaluation of the intensity of staining (0–3) and the number of cells staining (range of score 0–300). By this method, ER- or PR-positive tumors have a score of >1. For HER2 scoring, specimens were classified as positive if the immunohistochemical staining was 3+ or if the staining was 2+ and FISH (fluorescence in situ hybridization) was positive. Data on PD-L2 expression only included tumors with adequate tissue to perform immunohistochemistry. PD-L2 expression was assessed using a scoring scheme that relied on the distribution of positive tumor cells and staining intensity. In addition, the intensity factor ranging from zero (the intensity is either negative or staining, but it just exceeds the background) to two (the slide shows strong positive or dark brown staining when examined under a microscope) was used to multiply the distribution score, which estimates the percentage of cells that were positively stained. The intensity of PD-L2 immunostaining was classified as follows: the sections that scored Grade 1 were classified as low PD-L2 positivity groups, and the sections of Grade 2 were classified as groups with high PD-L2 positivity. The immunohistochemically stained sections were independently examined by two researchers who were not aware of the patients' clinicopathological features. We used cold radio-Immunoprecipitation assay (RIPA) buffer with phenylmethanesulfonyl fluoride (PMSF) to homogenize cells and then centrifuged the samples to obtain the supernatants. A total of 30 μ g of proteins were separated through SDS-PAGE prior to being transferred onto the 0.22- μ m polyvinylidene fluoride (PVDF) membrane. The membrane was blocked with a blocking buffer (5% defatted milk within tris-buffered saline tween-20 (TBST) that is composed of 10 mM Tris-HCl pH 8.0, 0.05% Tween 20, and 150 mM NaCl) at ambient temperature for 1 hr. After being subjected to overnight primary antibody incubation at 4°C, the membrane was probed with horseradish peroxidase-labeled secondary antibody at ambient temperature for 1–2 hr. The enhanced chemiluminescence (ECL) Western blot detection reagents (Millipore) were utilized to visualize the membranes. Cells (shNC, shPD-L2) were seeded in 96-well plates at a density of 3 × 10 3 cells/well in triplicate and maintained in an incubation state overnight. Cell viability was determined with the Cell Counting Kit-8 assay (CK04, Dojindo) in accordance with specific protocols. Each clone was tested in at least three replicate experiments. The data were measured in terms of the growth percentage in comparison with the controls. The average was calculated using three or more replicates for each clone. The results were represented in the form of the growth proportion in comparison with the controls. Cells (shNC, shPD-L2) were plated in 6-well plates at a density of 5 × 10 5 cells per well and allowed to grow until they reached 80% confluence. The cells were scratched in a straight line using a 200- μ l sterile pipette tip, then washed three times with PBS, and cultured for 24 hr. The scratches were visualized by a fluorescence microscope at a magnification of 200x (Nikon 80I; Nikon Corporation) and analyzed using Image J software (1.80v; National Institutes of Health). The Transwell migration assay was carried out by using Transwell inserts (24-well plates, Corning Costar) with a filter having an 8 μ m pore. After trypsinization into single cells, 5.0 × 10 4 cells (shNC, shPD-L2) were inoculated on top of the Transwell chamber which contained a serum-free medium. The bottom chamber was filled with 20% FBS-containing medium. After 12–16 hr, the cells were fixed with methanol on the insert membranes and stained with Giemsa solution, and nonmigrating cells on the upper surface of the membrane were gently removed. The number of migrated cells was counted in five randomly chosen fields per insert (x200) using Image J software (1.80v; National Institutes of Health). In the cell invasion assay, the upper chamber was first coated with Matrigel before cell seeding. This work obtained nude mice from Charles River in Beijing. The animal experimental protocols obtained approval from the Institutional Animal Ethics Committee of Suzhou Vocational Health College, China. The animals were randomly divided into two groups ( n = 6 for each). Mice were subjected to subcutaneous inoculation with MDA-MB231-NC and MDA-MB231-shPD-L2 cells (2 × 10 6 ) suspended in a serum-free medium (100 μ L). Approximately 1 week later, the average tumor volume was 400–500 mm 3 . Thereafter, the tumor size was determined using the Vernier caliper at 2-day intervals, and the tumor volume (mm 3 ) was calculated by multiplying the length by the square of the width and dividing by two. On the 15th day, the tumors were harvested. All data are expressed as the mean ± SEM. Statistical analysis was carried out using GraphPad Prism 8.0 software (GraphPad Software, La Jolla, CA, USA) through Student's t -test or two-way ANOVA. The status of proteins related to the main clinical and pathologic features was compared by means of χ 2 tests and Fisher's exact test when necessary. P < 0.05 indicated statistical significance. 3.1. sPD-L2 Is Overexpressed in the Serum of Breast Cancer Patients To assess the relationship between the PD-L2 level and breast cancer, serum sPD-L2 levels were detected among healthy controls and breast cancer patients by means of ELISA. As depicted in , the concentration of sPD-L2 was increased among breast cancer patients (6926.45 ± 78.38 pg/mL) as compared to that in healthy controls (2452.83 ± 57.53 pg/mL; P < 0.0001). Moreover, breast cancer patients with high concentrations of sPD-L2 had higher Ki67 values (≥30%) and tumor grade . As a nuclear protein, the expression of Ki67, which is measured via immunohistochemistry, is a marker of proliferation. In conclusion, PD-L2 is likely to be associated with the occurrence of breast cancer. 3.2. PD-L2 Expression in Breast Cancer Tissues Subsequently, in order to clarify the relationship of PD-L2 with the progression of breast cancer, the PD-L2 levels within tumors were analyzed. Immunohistochemistry disclosed that PD-L2 was positively expressed in 329 (79.09%) out of 416 tumors, was lowly expressed in 192 (46.15%), and was highly expressed in 137 (32.93%) of the tumors . The immunohistochemical results indicated that PD-L2 mainly resided on the cancer cell membrane and in the cytoplasm. 3.3. Correlations between PD-L2 Level and Clinicopathological and Molecular Characteristics summarizes the patient characteristics based on the PD-L2 expression status. The PD-L2 levels in breast cancer samples were examined, and as a result, it showed a positive correlation with HER2 ( P =0.0179) and PR ( P =0.0123). Taken together, PD-L2-positive tumors exhibited biological aggressiveness. However, the PD-L2 expression was not related to additional clinicopathological factors like age, tumor size ( P =0.4088), tumor grade ( P =0.3513), lymph node metastasis ( P =0.3508), or estrogen receptor expression ( P =0.1769). 3.4. PD-L2 Silencing Impeded MCF-7 and MDA-MB231 Cell Growth To explore the potential biological effect of PD-L2 on breast cancer cells, PD-L2 knockdown MCF-7 and MDA-MB231 stable cells were constructed. Gene knockdown efficiency of shPD-L2 lentiviral vector was confirmed by Western blotting . The growth curve indicated that PD-L2 knockdown impaired the proliferation of MCF-7 and MDA-MB231 cells (Figures and ). As revealed by the scratch assay, PD-L2 knockdown MCF-7 and MDA-MB231 cells had less migration than empty vector-transfected counterparts (Figures and ). Consistently, the Transwell assay showed that PD-L2 knockdown significantly decreased migration of both cell lines . The invasion assay demonstrated that PD-L2 knockdown blocked the invasion of these two cell lines . In total, low PD-L2 levels inhibit breast cancer cell proliferation, migration, and invasion. 3.5. PD-L2 Knockdown Inhibited Tumor Growth In Vivo Subsequently, we evaluated the function of PD-L2 in tumor growth in vivo. Animals were subjected to subcutaneous injection of MDA-MB231-shPD-L2 and MDA-MB231-shNC cells. The tumor volume was observed at 2-day intervals. The results indicated that there was no significant difference in body weight between the two groups . In addition, tumors with PD-L2 knockdown showed a significantly decreased growth rate compared to the controls . To assess the relationship between the PD-L2 level and breast cancer, serum sPD-L2 levels were detected among healthy controls and breast cancer patients by means of ELISA. As depicted in , the concentration of sPD-L2 was increased among breast cancer patients (6926.45 ± 78.38 pg/mL) as compared to that in healthy controls (2452.83 ± 57.53 pg/mL; P < 0.0001). Moreover, breast cancer patients with high concentrations of sPD-L2 had higher Ki67 values (≥30%) and tumor grade . As a nuclear protein, the expression of Ki67, which is measured via immunohistochemistry, is a marker of proliferation. In conclusion, PD-L2 is likely to be associated with the occurrence of breast cancer. Subsequently, in order to clarify the relationship of PD-L2 with the progression of breast cancer, the PD-L2 levels within tumors were analyzed. Immunohistochemistry disclosed that PD-L2 was positively expressed in 329 (79.09%) out of 416 tumors, was lowly expressed in 192 (46.15%), and was highly expressed in 137 (32.93%) of the tumors . The immunohistochemical results indicated that PD-L2 mainly resided on the cancer cell membrane and in the cytoplasm. summarizes the patient characteristics based on the PD-L2 expression status. The PD-L2 levels in breast cancer samples were examined, and as a result, it showed a positive correlation with HER2 ( P =0.0179) and PR ( P =0.0123). Taken together, PD-L2-positive tumors exhibited biological aggressiveness. However, the PD-L2 expression was not related to additional clinicopathological factors like age, tumor size ( P =0.4088), tumor grade ( P =0.3513), lymph node metastasis ( P =0.3508), or estrogen receptor expression ( P =0.1769). To explore the potential biological effect of PD-L2 on breast cancer cells, PD-L2 knockdown MCF-7 and MDA-MB231 stable cells were constructed. Gene knockdown efficiency of shPD-L2 lentiviral vector was confirmed by Western blotting . The growth curve indicated that PD-L2 knockdown impaired the proliferation of MCF-7 and MDA-MB231 cells (Figures and ). As revealed by the scratch assay, PD-L2 knockdown MCF-7 and MDA-MB231 cells had less migration than empty vector-transfected counterparts (Figures and ). Consistently, the Transwell assay showed that PD-L2 knockdown significantly decreased migration of both cell lines . The invasion assay demonstrated that PD-L2 knockdown blocked the invasion of these two cell lines . In total, low PD-L2 levels inhibit breast cancer cell proliferation, migration, and invasion. Subsequently, we evaluated the function of PD-L2 in tumor growth in vivo. Animals were subjected to subcutaneous injection of MDA-MB231-shPD-L2 and MDA-MB231-shNC cells. The tumor volume was observed at 2-day intervals. The results indicated that there was no significant difference in body weight between the two groups . In addition, tumors with PD-L2 knockdown showed a significantly decreased growth rate compared to the controls . PD-L2 was discovered to be a new member of the B7 family among the genes that were differentially expressed in dendritic cells and activated macrophages in a gene library . Initially, it was thought that PD-L2 expression was restricted to antigen-presenting cells such as dendritic cells and macrophages . Nonetheless, it has now been discovered that PD-L2 is expressed in various tumor, stromal, and immune cells, according to microenvironmental stimuli . Compared with PD-L1, PD-L2 displays the higher affinity to PD-1 . As Tanegashima et al. discovered, the individual or simultaneous expression of PD-L2 with PD-L1 in cancer cells inhibited antitumor immunity, which was related to the resistance to anti-PD-L1 treatment that was found in preclinical animal models. The mechanisms by which PD-L2 modulates tumor immunity are still unclear, but these results suggest the importance of PD-L2 for evading antitumor immunity. Furthermore, the blockade of PD-L1 and PD-1 or PD-L2 should be taken into consideration in order to attain the optimal immunotherapy for PD-L2-positive tumors. PD-L2 is highly expressed in various human cancer types, such as ovarian cancer , lung adenocarcinoma , gastric cancer , and esophageal squamous cell carcinoma . By conducting immunohistochemistry, we demonstrated that positive PD-L2 expression was presented in 329 (79.09%) out of 416 breast cancer samples, weak expression was seen in 192 (46.15%) of the patients, and strong expression was observed in 137 (32.93%) of the patients. sPDL2 is mainly thought to be produced by the cleavage of membrane-bound PD-L2, similar to its cousin PD-L1 . According to a recent study, the expression of sPD-L2 significantly increased among nonsmall cell lung cancer patients when compared to that of normal donors . In another study, it was found that platinum resistance in advanced epithelial ovarian carcinoma was associated with sPDL2 . Therefore, sPD-L2 is capable of enhancing cancer invasion via its interaction with the membrane-bound PD-1 on immune cells . Based on our results, by means of ELISA, the serum sPD-L2 expression increased among breast cancer patients. The sPD-L2 level among the patients with breast cancer was 6926.45 ± 78.38 pg/mL, while the concentration in healthy controls was 2452.83 ± 57.53 pg/mL. Moreover, the patients with breast cancer having high concentrations of sPD-L2 had higher Ki67 values (≥30%) and tumor grades. These findings imply that sPD-L2 can serve as a biomarker for the diagnosis and prognosis prediction of breast cancer. The upregulation of PD-L2 has been noticed to be associated with the poor prognostic outcome of specific tumor types, while different results are uncovered in additional cancer types. Hence, it is of great importance to examine the precise effect of PD-L2 on tumor tissue. In our study, the PD-L2 expression significantly increased in 79.09% of breast cancer tissues, demonstrating a notable correlation with HER2 and PR. Nevertheless, the PD-L2 expression was not related to other clinicopathological factors such as age, tumor size ( P =0.4088), tumor grade ( P =0.3513), lymph node metastasis ( P =0.3508), or estrogen receptor expression ( P =0.1769). This is not in line with the findings in serum where the expression levels of sPD-L2 correlate with the tumor grade. Such a discrepancy in the results is likely associated with the differences in patient basic characteristics as well as the PD-L2 immunohistochemical methods, thresholds for positive expression, and scoring systems. For instance, in studies regarding PD-L2 and tumors, positive PD-L2 expression shows significantly different prognostic significance . Our study possesses certain limitations. Firstly, we had a relatively small number of clinical samples and relatively few indicators for analyzing PD-L2 and clinicopathological features. Secondly, the expression profiles of PD-L2 and sPD-L2 were obtained from two separate sample banks. We will continue to gather clinical samples by using tissue-matched plasma samples prior to surgery in order to assess the correlation between the tissue expression of PD-L2 and the serum expression of sPD-L2 with the clinicopathological features of the patients. To summarize, we determined the oncogenic impact of the immune checkpoint PD-L2 on the progression of breast cancer. The high expression of PD-L2 within breast cancer cells boosts cell growth, invasion, and migration. Correspondingly, knocking down PD-L2 inhibited tumor cell growth, migration, and invasion both in vivo and in vitro . Given the extensive upregulation of PD-L2 within breast cancer and its pro-oncogenic effect on the development of breast cancer, targeting PD-L2 represents a potential antibreast cancer treatment. |
Über Geschmack lässt sich nicht streiten? | 8b74bb3b-ac70-472a-868d-4c3c65d28952 | 10272244 | Pharmacology[mh] | Wer heutzutage in Deutschland ein Studium der Pharmazie aufnimmt, durchläuft im Fach Pharmazeutische Biologie auch einen atavistisch anmutenden Kursus, der sich unter den Student*innen allerdings allgemeiner Beliebtheit erfreut: Es geht darum, über 100 verschiedene Arten von Wurzeln, Blättern, Kräutern, Früchten und so weiter mit allen Sinnen wahrzunehmen, auch – oder gerade – ihren Geschmack zu prüfen und die Eindrücke genau zu protokollieren, damit diese sogenannten „Drogen“ daraufhin einwandfrei identifiziert werden können. Es bedarf kaum einer historischen Vorbildung für die Einsicht, dass diese Art der wissenschaftlichen Praxis ihre Wurzeln in einer Zeit hat, zu der instrumentelle Analytik noch nicht zur Verfügung stand und der Heilkundige alle Möglichkeiten organoleptischer Prüfung auszuschöpfen hatte, um die Qualität einer Droge zu bestimmen. Manche mögen heute zweifeln, wie ein Geschmacksurteil, über das sich laut einem Sprichwort nicht streiten lässt (was bedeutet, dass die Subjektivität des Urteils jedem zugestanden wird), zu objektivierbaren Ergebnissen führen soll. Doch ein solcher Zweifel wäre antiken Ärzten wie Galen nicht nur fremd (Singer & van der Eijk : 19), sondern ist auch, wie ich hoffe zeigen zu können, in Bezug auf die Pharmakognosie weitgehend unbegründet. Galen entfaltet seine Arzneimittellehre in mehreren Schriften, wobei die in ihrer letzten Edition knapp 900 Seiten zählende Abhandlung Über die Wirkungspotenziale der einfachen Arzneimittel systematisch gesehen den ersten Platz einnimmt. Die Schrift gliedert sich ähnlich wie moderne Lehrbücher der Pharmakologie in eine einleitende allgemeine (Bücher I–V) und eine spezielle (VI–XI) Arzneimittellehre – ein Katalog mit den von Dioskurides beschriebenen Vegetabilia, Mineralia und Animalia. Im letzten Buch (XI) dieser Schrift kommt Galen bei der Beurteilung von Fetten und Talgen auch auf seine Vorgänger zu sprechen. Durch längere Lagerung komme es zur Qualitätsveränderung; die Fette würden im Geschmack „beißender“ (δριμύτερα), nicht jedoch „adstringierender“, was einige Autoren zu Galens Missfallen behauptet haben (XII 329.9–17 K.): Einige von denjenigen [medizinischen Autoren], die die Bedeutungen der Wörter korrumpieren (τινὲς δὲ τῶν διαφθειρόντων τὰ σημαινόμενα τῶν ὀνομάτων), nennen alle derartigen [Stoffe] nicht „beißend“ (δριμέα), sondern „adstringierend“ (στύφοντα), bis hin zum Pfeffer – als ob es keinen Unterschied machte, entweder „adstringierend“ oder „beißend“ zu sagen. Und wenn man sie wiederum nach Galläpfeln, Myrten, Mispeln, Granatapfelschalen, unreifen Oliven und Gerbersumach fragen würde, dann behaupten sie, dass auch diese adstringieren, obgleich wir ja hierbei die gegensätzlichste Wahrnehmung haben zu derjenigen, die uns durch Pfeffer, Feuerwurzel-Bertram und Senf […] entsteht. Zu den die Wortbedeutung korrumpierenden Autoren zählt, wie Galen im Folgenden (XII 350.12 K.) deutlich macht, insbesondere der Anazarbeer Dioskurides – also seine pharmakologische Hauptquelle. Man kann sich sicherlich fragen, ob es nicht Pedanterie, vielleicht auch Polemik oder die von diesem Autor bekannte „Gier nach ausschließlicher Leserbindung“ in einem „Agon aller gegen alle“ (Asper : 33) ist, über die Geschmacksurteile längst verstorbener Ärzte zu streiten. Damit würde man allerdings ein wesentliches Charakteristikum von Galens Pharmakologie (und auch denjenigen der Folgezeit) verkennen. Für Galen war die Sinneswahrnehmung zunächst ein valides und notwendiges Mittel, das in der medizinischen Forschung stets hinzuzuziehen und an dem die theoretischen Überlegungen jederzeit zu prüfen sind. Genau genommen macht Galen auch nicht den Fehler, Dioskurides absprechen zu wollen, einen bestimmten Geschmackseindruck gehabt zu haben (denn darüber lässt sich wirklich nicht streiten), sondern er kritisiert nur die Nachlässigkeit des Sprachgebrauchs bei einem Autor, der als Nicht-Muttersprachler gar in seinem Vorwort den Leser bittet, das Augenmerk nicht auf seine Ausdrucksfähigkeit zu legen (μὴ τὴν ἐν λόγοις δύναμιν ἡμῶν σκοπεῖν, Dsc. Prooem 5). Um eine entsprechend exakte Terminologie, die insbesondere auch den Sprachgebrauch seiner eigenen Zeit berücksichtigt, war Galen folglich äußerst bemüht. Das gesamte vierte und fünfte Buch seiner Schrift (XI 619–788 K.) ist im Wesentlichen mit der Frage beschäftigt, wie die Mischung ( krasis ) der Pharmaka mit ihren wärmenden, kühlenden, trocknenden und feuchtmachenden Eigenschaften ( dynameis ) einerseits und den Geschmacksqualitäten (Buch IV) sowie den spezifischen Arzneimittelwirkungen (Buch V) andererseits zusammenhängen. Prinzipiell gehen seine Überlegungen von den Prämissen aus, dass sowohl die Geschmacksqualitäten als auch die Arzneimittelwirkungen ihre Ursache in einer bestimmten krasis der Elementarqualitäten haben und dass ferner zunächst „die Geschmacksqualität eines Pharmakons die Äußerungsform seiner medizinischen Wirkung im oder am Körper darstellt“ (Harig : 69). Entsprechend ausgearbeitet ist seine Geschmackslehre, die in der Tradition Platos ( Timaios 28: 65c–66c) und des Peripatos (Theophrast De caus. pl. VI 4.1) steht und insgesamt neun Qualitäten unterscheidet: 1. adstringierend (στῦφον/ styphon , 136×), 2. herb-sauer (στρυφνόν/ stryphnon , 27×), 3. herb (αὐστηρόν/ austēron , 15×), 4. stechend-sauer (ὀξύ/ oxy , 7×), 5. süß (γλυκύ/ glyky , 18×), 6. fettig (λιπαρόν/ liparon , 3×), 7. salzig (ἁλυκόν/ halykon oder ἁλμυρόν/ halmyron , 4×), 8. bitter (πικρόν/ pikron , 114×) und 9. beißend(-scharf, δριμύ/ drimy , 110×). Der Geschmackssinn erhält bei Galen insofern seine besondere Bedeutung, als er vor allen anderen Sinnen geeignet ist, Hinweise zur dynamis eines Arzneimittels zu liefern. Die von Harig genauer untersuchte Ableitung der Geschmacksqualitäten kann an dieser Stelle nur gerafft wiedergegeben werden: Die adstringierenden, das heißt „zusammenziehenden“ Qualitäten – man denke beispielsweise an einen gerbstoffreichen Wein (1.–3.) – werden von der kalten Primärqualität abgeleitet, die bei (3.) allerdings weniger stark ausgeprägt sei. Geschmack (4.) ist ebenfalls kalt, allerdings von besonders „feinteiliger Substanz“. Die typischen Geschmacksqualitäten der Nahrungsmittel (5.) süß und (6.) fettig würden in ihrer Temperierung der Wärme des menschlichen Körpers entsprechen und daher als angenehm empfunden. Die übrigen drei Qualitäten seien entsprechend Ausdruck der warmen Elementarqualität, an der sie einen graduell ansteigenden Anteil haben, wobei die äußerst warme, scharfe Qualität (9.) aufgrund ihrer gleichzeitigen „Feinteiligkeit“ besonders beißend sei (man denke an Pfeffer, Zwiebeln und Knoblauch). In der obigen Aufzählung ist ferner die Anzahl an Vegetabilia angegeben, denen bei Galen die jeweilige Qualität zugeschrieben wird (nach Haars : 194). Es zeigt sich schon hier: Medizin muss bitter schmecken! Um beurteilen zu können, in welchem Umfang die Geschmacksangaben in der speziellen Pharmakologie Galens ( De simpl. med. Bücher VI–XI) auf eigenen Forschungen basieren oder aber auf seine Vorgänger zurückzuführen sind, sollen im Folgenden insbesondere die Autoren betrachtet werden, die Galen heranzog. Da diese Autoren unsere Hauptquellen zur antiken Arzneimittellehre darstellen, dürfte die folgende Untersuchung auch insgesamt repräsentativ für diesen Zweig der griechischen Heilkunde sein. Die zu Galens Zeit und weit darüber hinaus berühmteste und ausführlichste Arzneimittellehre war diejenige des Anazarbeers Dioskurides aus der zweiten Hälfte des ersten nachchristlichen Jahrhunderts. Galen (XI 794 K.) kannte sie gut und lobt sie ausdrücklich, wie auch ein ähnliches, in der ersten Hälfte desselben Jahrhunderts entstandenes griechisches Werk des Sextius Niger (Wellmann , Scarborough ), das als gemeinsame Quelle sowohl bei Dioskurides als auch bei Plinius eingeflossen ist (im Folgenden abgekürzt: e S. N., nach Wellmann 1907–1914). Aus den Übereinstimmungen beider Autoren kann man seine Lehre herausschälen, weswegen wir im Folgenden auch Plinius betrachten, obwohl Galen ihn nicht erwähnt – wie überhaupt keinen ausschließlich in lateinischer Sprache schreibenden Autor (vgl. Nutton : 395, Anm. 89). Hierzu wurde das dritte Buch von De materia medica (hg. Wellmann ) ausgewertet, worin unter anderem Heilpflanzen der wichtigen Familien Apiaceae, Asteraceae und Lamiaceae behandelt sind. Um den deskriptiven Charakter dieser Angaben zu zeigen, müssen einige Beispiele genügen: Von den 158 Kapiteln, in denen meist eine Arzneipflanze (zuweilen auch mehrere ähnliche Arten) abgehandelt wird, finden sich in 78 – also fast der Hälfte – Geschmacksangaben. Stichproben der anderen Bücher zeigen Ähnliches: In 129 Kapiteln des ersten und 192 des vierten Buches finden sich Angaben zu 81 (hier werden viele aromatische Pflanzen und Obstarten behandelt) beziehungsweise 79 Pflanzen (sie fehlen bei vielen giftigen Arten). Die häufigsten Geschmacksqualitäten der Simplicia sind beißend-scharf (δριμύς/ drimys , 29×), brennend (πυρωτικός/ pyrōtikos , 7×), bitter (πικρός/ pikros , 21×), süß (γλυκύς/ glykys , 8×), adstringierend (Formen von στύφειν/ styphein , 6×), sauer (στρυφνός/ stryphnos , 2×), salzig (ὕφαλμος/ hyphalimos , 1×) und aromatisch (Formen von ἀρωματίζειν/ arōmatizein , 6×). Ein genauerer Vergleich des dritten Buches mit Plinius ergibt ferner, dass Dioskurides bezüglich der Geschmacksangaben fast immer ausführlicher ist. Dies überrascht nicht, denn man würde von vornherein annehmen, dass ein drogenkundliches Werk, das für Fachkreise bestimmt ist, eher auf die Geschmacksprüfung zur Identitätsfeststellung eingeht als eine naturkundliche Enzyklopädie für eine breitere Leserschaft. Zu den Angaben in 34 Kapiteln bei Dioskurides findet man etwa in den Parallelstellen bei Plinius überhaupt keine Entsprechung. So handelt Dsc. III 48 ausführlich die Panazee des Herakles ab, mit ihren „wohlriechenden, brennend-scharfen Samen“ und „Wurzeln, die eine dicke, im Geschmack leicht bittere Rinde besitzen“, von welchen diejenigen besser seien, die „im Geschmack brennend und aromatisch sind“, während sich unter den davon gewonnenen „Gummiharzen dasjenige mit dem bitteren Geschmack auszeichne“. Diese Angaben fehlen hingegen bei Plinius XII 127 (e S. N.), wie auch, dass der Samen von ami im Geschmack „origanisierend“ sei (so die Wortneubildung von Dsc. III 62: ὀριγανίζον τῇ γεύσει – fehlt bei Pl. XX 163f. e S. N.). Auch die Geschmacksbeschreibung der berühmten Droge Silphium bietet nur Dsc. III 80: „Es zeichnet sich derjenige Saft aus, der weder lauchartig noch unangenehm im Geschmack ist [… D]er Kyrenäische Saft […] ist auch im Geruch so ausgesprochen milde, dass auch nach dem Kosten kein Mundgeruch entsteht“. In zwei Fällen erwähnt Plinius die Heilpflanze nicht. In 25 anderen Fällen stimmen diese bei beiden Autoren genau überein (wenn man eine gewisse Toleranz durch die Übersetzung der griechischen Quelle ins Lateinische bei Plinius zulässt). So sagt Dsc. III 1 von dem Baumpilz agarikon , dass er zu Beginn süß schmeckt, beim weiteren Kauen aber bitter (γεύσει δὲ […] κατ’ ἀρχὰς μὲν γλυκάζοντα, εἶτα ἐξ ἀναδόσεως ἔμπικρα), was wir ebenso bei Plinius lesen ( initio gustus dulcis mox in amaritudinem transit , Pl. XXV 103, e S. N.). Auf dieses Urteil werden wir unten bei Galen noch zurückkommen. Gerade da, wo der Geschmackstest wichtig ist, um Drogen zu unterscheiden, stimmen beide überein: Das Kraut Kretischer Diktam nennen sie (nach meinem Empfinden zu Recht) „beißend“ (Dsc. III 32: πόα […] δριμεῖα λίαν – acre gustu , Pl. XXV 92), und zwar schon in einer sehr geringen Menge ( minima portione accendit os , PI. XXV 93 ), während die Verfälschung, das Pseudodiktamnon, weniger beißend sei (Dsc. III 32.2: ἧττον δριμύ). Beide unterscheiden ferner die wilde von der kultivierten Raute anhand ihres beißenderen Geschmacks (Dsc. III 45: τὸ ὄρειον καὶ ἄγριον τοῦ ἡμέρου δριμύτερον – ruta […] Silvestris […] ad omnia acrior , Pl. XX 131 e S. N.). Weitgehende Einigkeit herrscht auch bei der Beurteilung der berühmten Droge Euphorbium, deren Geschmacksprüfung die antiken Autoren jedoch vor eine große Herausforderung stellte – vgl. Dsc. III 82: „Wähle aber den scharfen [Saft] aus. Allerdings ist er schwer mit dem Geschmackssinn nach der Einnahme zu beurteilen, denn wenn die Zunge einmal gebissen wurde, bleibt der brennende Geschmack eine ganze Weile, sodass alles [ihr] Zugeführte Euphorbium zu sein scheint“. Plinius, der mit seiner Quelle fälschlich den aus der Pflanze chamelaia bereiteten Saft für Euphorbium hält, sagt: „[E]r brennt, wenn auch nur leicht gekostet, lange im Mund und nach einer Weile noch heftiger.“ Die nach Circe benannte Pflanze hat Wurzeln, welche Dsc. III 119 als „wohlriechend und wärmend“ beschreibt (κιρκαία […] ῥίζας […] εὐώδεις, θερμαντικάς). Dass „wärmend“ hier als Geschmack gemeint ist, bezeugt Plinius: Circaea […] radice […] odorata, gustus calidi , Pl. XXVII 60 (e S. N.). In elf Fällen bieten beide Autoren voneinander abweichende Geschmacksangaben und in 80 Kapiteln fehlen Geschmacksangaben bei beiden, wobei die Gründe hierfür unklar sind. In sieben Fällen, wo bei Dioskurides Angaben fehlen, findet man solche bei Plinius, und zwar jeweils in Abschnitten, die aus Sextius Niger stammen. Was lernen wir daraus? Dass Dioskurides und Plinius ihrer gemeinsamen Quelle, dem römischen Arzt Sextius, auch in Hinblick auf die Geschmacksqualitäten viel verdanken, steht außer Frage. Unklar ist allerdings der Umfang, in dem Dioskurides ihn für Informationen herangezogen hat, die bei Plinius fehlen. Hier muss man davon ausgehen, dass der Anazarbeer eine andere Quelle hatte oder eigene Beobachtungen oder Geschmäcke darlegt. So bemerkt Dsc. III 54 über den Samen von Tordylon, dass er „leicht beißend und aromatisch sei“ (σπερμάτιον […] ὑπόδριμυ, ἀρωματίζον) – eine Angabe, die bei Plinius (XXIV 177 e S. N.) fehlt; ja er schreibt gar, dass er über die Pflanze nichts anderes in Erfahrung bringen (und das heißt wohl: bei Sextius lesen) konnte, als was er hierzu berichtet ( neque aliud de ea proditum invenio quam […]). Dies ist freilich nur ein winziges, kaum belastbares Indiz, und da die Schrift des Sextius nicht überliefert ist, wird eindeutige Klarheit nicht zu gewinnen sein. Durch die Quellenuntersuchung von Wellmann haben wir Anlass zu glauben, dass die pharmakognostischen Angaben des Sextius Niger unter anderem auf Krateuas, den Leibarzt von Mithridates VI. (ca. 132–63 v. Chr.), zurückgehen. Er verfasste nach den wenigen Zeugnissen, die überliefert wurden, ein bebildertes, alphabetisch geordnetes Kräuterbuch, das vielleicht als Vorbild des Wiener Dioskurides (Österreichische Nationalbibliothek, Cod. med. gr. 1 [Abb. ]) gedient haben könnte (hierfür gibt es jedoch keinen Beweis, siehe auch Cronier ), ferner ein als „Wurzelschneiderbuch“ ( rhizotomikon ) betiteltes Werk mit Pflanzenbeschreibungen (aber ohne Bilder) und schließlich eine Schrift, in der auch mineralische Drogen behandelt wurden und die vielleicht Teil eines größeren Werks nach Art der dioskurideischen De materia medica gewesen ist (Wellmann ; Jacques ). Aus dem ersten Werk sind zehn Fragmente in dem Wiener Prachtcodex erhalten und ediert (bei Wellmann : 144–146). Sie haben jeweils eine Arzneipflanze mit ihren Indikationen zum Gegenstand, wobei im Unterschied zu den korrespondierenden Kapiteln bei Dioskurides keine Geschmacksqualitäten genannt werden. Dies mag uns nicht überraschen, denn die Geschmacksqualitäten waren (bei Dioskurides und wohl auch vor ihm) Bestandteil des deskriptiven botanischen Teils, der ja in einem bebilderten Kräuterbuch entbehrlich war. Eine größere, wenn auch immer noch sehr dürftige Textmenge stellen die überlieferten Zeugnisse zu seiner Lehre dar (bei Wellmann : 139–144). In diesen Testimonien, die fast ausschließlich die Synonyme der Pflanzen betreffen, finden sich nur wenige ausführlichere Zitate, die auch die Beschreibung der Gewächse, niemals aber ihren Geschmack zum Gegenstand haben. Eine Ausnahme ist allein die Stelle aus der pseudogalenischen Schrift „Über die Heilkräfte der Centaurea“, welche Krateuas im dritten Buch seines Werks behandelt haben soll (test. A 20). Wellmann hat in diesem Testimonium zudem die ganze Beschreibung der Centaurea mitsamt der Angabe „auch ist ihre Wurzel im Geschmack süß und leicht beißend“ ( et est [sc. radix centaureae] in gustu dulcis et subacris ) ausgehoben und legt damit nahe, dass hier Krateuas paraphrasiert wird – was wir freilich nicht wissen können. Zumindest sehe ich aber keine Gründe, die dagegensprächen (vgl. auch Nutton : 152). Abschließend lässt sich festhalten, dass das überlieferte Krateuas-Material kein Urteil in der Frage zulässt, ob der Autor Geschmacksangaben systematisch erhoben hat oder nicht. Ähnliches gilt auch für Diokles von Karystos (4. Jh. v. Chr.), obgleich wir hier deutlich mehr überlieferten Text besitzen. In dessen zuletzt von Philip van der Eijk ( –2001) gesammelten Fragmenten begegnen uns gelegentlich Geschmacksqualitäten, meist jedoch in diätetischen Zusammenhängen (fr. 183a.104.119f.125, fr. 187.11–13 usw.). Diokles empfiehlt beispielsweise (das schon oben genannte) „Silphium, das am besten riecht und am bittersten [schmeckt]“ (σίλφιον δέ […] τὸ εὐωδέστερον καὶ πικρότατον, fr. 187.24f.). Aus den Übereinstimmungen von Pflanzensynonyma und -beschreibungen zwischen Dioskurides, Nikander und Theophrasts Historia plantarum (insbesondere Buch IX) schloss Wellmann , dass Diokles’ rhizotomikon das älteste Kräuterbuch der Griechen darstellt. Diesem Wurzelschneiderbuch, das in unserem Zusammenhang am interessantesten wäre, kann jedoch nur ein einziges Fragment (fr. 204, ohne Geschmacksangabe) sicher zugeordnet werden. Bekannt ist Diokles ferner für das von Galen (VI 455f. K.) überlieferte sogenannte „große Methodenfragment“ (fr. 176), in welchem davor gewarnt wird, von Substanzen (im Kontext bei Galen: Lebensmitteln) ähnlichen Geschmacks notwendigerweise auf ähnliche (Nähr‑)Kräfte zu schließen („Diejenigen, die annehmen, dass Substanzen mit ähnlichem Geschmack […] alle auch die gleichen Kräfte besitzen, glauben dies zu Unrecht“; οἱ μὲν οὖν ὑπολαμβάνοντες τὰ τοὺς ὁμοίους ἔχοντα χυλοὺς […] πάντα τὰς αὐτὰς ἔχειν δυνάμεις οὐ καλῶς οἴονται) – eine Warnung, die sich insbesondere gegen undifferenzierte Generalisierungen richtet, nicht gegen Ursachenforschung überhaupt (van der Eijk –2001: 331f.). Bevor wir uns Galen zuwenden, wollen wir kurz die Ergebnisse der vorangegangenen Untersuchung festhalten: Geschmacksqualitäten hatten bei den Autoren, die Galen heranzog, primär die Funktion, die Droge zu charakterisieren, ihre Qualität zu bestimmen, sie von anderen ähnlichen Sorten abzugrenzen oder Fälschungen zu erkennen. Die meisten Geschmacksangaben finden wir bei Dioskurides, nämlich zu durchschnittlich 50 Prozent der behandelten Simplicia . Zumindest ein großer Teil dieser Angaben findet sich auch bereits bei seiner wichtigsten Quelle, Sextius Niger. Dass auch dessen Quellen – nämlich Krateuas und möglicherweise Diokles – in größerem Umfang Geschmacksangaben erfasst oder weitertradiert haben, lässt sich zwar nicht mehr oder nur noch anhand weniger Fragmente für einzelne Beispiele zeigen, doch dürfte nach dem Charakter dieser Literaturgattung die Beweislast eher bei dem liegen, der das Gegenteil behaupten wollte. Nach dem bereits Gesagten dürften wir erwarten, dass sich bei Galen ein deutlich größerer Umfang an Geschmacksangaben findet als bei Dioskurides. Von den drei Büchern (VI–VIII), die sich thematisch (von der alphabetischen Anordnung abgesehen) mit den untersuchten Büchern des Dioskurides gut vergleichen lassen, finden sich Geschmacksangaben in 287 der insgesamt 474 Kapitel, also in 60 Prozent der Kapitel. Die Quote ist zwar um immerhin 10 Prozentpunkte höher als bei Dioskurides, doch noch nicht so hoch, wie angesichts des in Buch IV in Aussicht gestellten Forschungsprogramms zu erwarten wäre. Damit in Zusammenhang steht ferner eine Beobachtung, die Harig in seiner Monografie zur Intensitätsbestimmung, also der Einordnung der Drogen nach ihrer in vier Grade abgestuften Wirkstärke, gemacht hat. Wie er ( : 143–145) zeigen konnte, liegen numerische Angaben nur für ein Drittel des galenischen Arzneischatzes vor, weswegen man den speziellen Teil wohl als unvollendet bezeichnen muss. Dennoch sagt die Quote allein wenig über die Arbeitsweise des Pergameners aus. Vergleicht man seine Ausführungen mit denen des Dioskurides, fällt nämlich auf, dass dieser nicht nur Angaben bietet, die bei jenem fehlen, sondern dass er mitunter auch zu anderen Einschätzungen oder differenzierteren Urteilen kommt. Dioskurides gibt beispielsweise den Geschmack der agrōstis -Wurzel allein als „stark süß“ (γλυκείας ἰσχυρῶς, Dsc. IV 31) an, während Galen bemerkt, dass die agrōstis eine essbare Wurzel habe, „die, immer wenn sie weich ist, einerseits wässrig-süß schmeckt, andererseits aber etwas Beißendes und leicht Herbsaures hat“ (ἄγρωστις ἐδώδιμον ἔχει τὴν ῥίζαν, ἔστ’ ἂν ᾖ μαλακή, γλυκεῖα μὲν ὑδατώδης, δριμὺ δέ τι καὶ ὑπόστρυφνον ὀλίγον ἔχουσα, XI 810.15–17 K.). Der schon oben als Beispiel angeführte Baumpilz agarikon habe laut Dioskurides zu Beginn einen süßlichen, nach dem Kauen aber einen bitteren Geschmack (Dsc. III 1). Galen erscheint er ebenfalls „beim ersten Kosten irgendwie süß, wenig später aber leicht bitter und mit der Zeit führt er den Eindruck einer gewissen beißenden Schärfe und geringen Adstringenz herbei“ (κατὰ μὲν τὴν πρώτην γεῦσιν γλυκεῖά τις, ὑπόπικρος δὲ ὀλίγον ὕστερον φαινομένη καί τινος ἐν τῷ χρόνῳ δριμύτητος ἔμφασιν ἐπάγουσα, καὶ βραχείας στύψεως, XI 813.12–15 K.). Der Detailreichtum dieser Beschreibung erscheint manchen Interpreten zwar befremdlich („rather obscure“, Totelin : 62), doch belegt er eindrücklich, für wie wichtig Galen diesen Aspekt in der Pharmakologie offensichtlich erachtete und man darf wohl vermuten, dass er dieselbe Akribie bei allen 474 Pflanzenlemmata würde angewendet haben, wenn er die Möglichkeit dazu gehabt hätte. Aber auch ungeachtet der Tatsache, dass er seinen Katalog in dieser Hinsicht nicht vervollständigen konnte, sind die genannten Beispiele keine Einzelfälle und es lassen sich hier noch viele weitere Angaben finden, die über das von Dioskurides Mitgeteilte hinausgehen, seine Einschätzungen ergänzen oder modifizieren – und zwar allein bei den Vegetabilia in über 100 Fällen. Was bedeutet dies für die Arbeitsweise Galens? Offensichtlich konnte er sich für viele Angaben nicht allein auf seine Hauptquelle stützen, die in diesem Belang nicht immer differenziert genug, teilweise auch unpräzise oder gar fehlerhaft war. Konnte er also auf eine andere Quelle zurückgreifen? Von den Autoren, die laut Galen (XI 789–798 K.) über Simplicia geschrieben haben, kommen eigentlich nur Sextius Niger, vielleicht Krateuas oder Diokles infrage. Wir haben indes keinen Anlass zu glauben, dass Galen Geschmacksqualitäten aus Sextius übernahm. Weder sind dessen Angaben, sofern wir sie durch Vergleich mit Plinius erschließen konnten, ausführlicher oder differenzierter, noch bezieht sich Galen in den sieben Fällen, wo Plinius Geschmacksangaben aus Sextius schöpft, die bei Dioskurides fehlen, eindeutig auf den Text des Sextius. Über die anderen beiden Autoren lässt sich, wie oben skizziert, wenig sagen. Es sollte aber zu denken geben, dass in Galens Zitaten aus den Werken dieser Autoren Geschmacksqualitäten keine Erwähnung finden. Hätte er denn überhaupt die Materia medica des Dioskurides als die „vollendetste“ (τελεώτατα, XI 794.11 K.) Darstellung ihrer Art loben können, wenn die anderen Autoren umfangreichere oder präzisere Angaben boten? Die Antwort mag sich jeder selbst geben. Wie die Dinge liegen, kommt vornehmlich eine Option in Betracht: Galen hat die abweichenden oder über Dioskurides hinausgehenden Geschmacksangaben selbst erhoben. Dass diese Vermutung nicht abwegig ist, dürfte aus den vielen Bemerkungen Galens, wo dieser um botanische und pharmazeutische Autopsie wirbt, eigentlich selbstverständlich sein. Dass Galen, dem als Leibarzt des Kaisers die wertvollen Magazine mit Arzneidrogen aus dem ganzen Reich offenstanden und der zudem viele Reisen unternommen hat, die allein der Beschaffung von Arzneimitteln dienten, dass also dieser Arzt wohl mehr als jeder andere Zeitgenosse auch die Möglichkeiten zum genauen Studium der Heilmittel hatte, wird wohl niemand bezweifeln. Da es nun plausibel erscheint, dass ein Großteil der protokollierten Geschmacksangaben tatsächlich auf das Konto Galens gehen, können wir als nächstes die eigentliche Tätigkeit des Schmeckens genauer untersuchen, und zwar unter der Bedingung der Möglichkeit, dass es sich um galenische Forschung handelt. Hier sind wir natürlich an das gebunden, was uns der Zufall der Überlieferung als Quellenmaterial zur Verfügung stellt. Die Ausgangslage ist jedoch nicht so hoffnungslos, wie ein Neuzeithistoriker annehmen könnte, der es gewohnt ist, sein Material aus so vielfältigen Quellen wie Labortagebüchern, unpublizierten Forschungsdaten, Korrespondenzen etc. speisen zu können. Denn auch das macht die Faszination des Autors Galen aus, dass sein gewaltiges Œuvre von biografischen Anmerkungen durchzogen ist und unsere Hauptquelle, die Heilpflanzenkataloge der Bücher VI bis VIII, weniger einen durchkomponierten und abgeschlossenen Traktat darstellen als vielmehr ein verschriftlichtes work in progress . Die folgenden Angaben sind für das Verständnis der pharmakologischen Forschung Galens daher sehr erhellend, wobei nicht verschwiegen werden darf, dass uns eine unabhängige Quelle zur Verifikation fehlt: Galen bezog die Heilpflanzen entweder von vertrauensvollen Kontaktpersonen „teils aus Groß-Syrien, Palästina, Ägypten, Kappadokien, andere aus Pontus, ebenso aus Makedonien und den Ländern im Westen, wo Kelten, Iberer und in dem gegenüberliegenden Lande Mauretanier wohnen“ ( De antidotis XIV 8.14–9.2 K.) oder bei in Rom ansässigen, aber oft weniger vertrauensvollen Händlern (Fälschungen waren sehr verbreitet, vgl. XIV 7.6 K.). Außerdem konnte Galen als Leibarzt unter anderem Mark Aurels (reg. 161–180) auf die kaiserlichen Magazine zurückgreifen, die mit wertvollen Spezereien insbesondere aus Kreta, dem „botanical Eden“ (Nutton : 252), gefüllt waren. Von dort kamen alljährlich im Sommer die Kräuter der kaiserlichen Sammler in großen Körben an und wurden von Drogenhändlern unter anderem anhand des Geschmacks in ihrer Qualität beurteilt ( De antidotis XIV 9–11). In Rom verfügte Galen entsprechend über einen der größten Arzneimittelvorräte seiner Zeit (Boudon-Millot : 305, Anm. 6). Eine dritte Möglichkeit bestand darin, die Heilmittel selbst zu sammeln. So solle ein Arzt, wenn er sich gerade auf dem Land aufhalte und seine Medikamente nicht bei sich führe, „fähig sein, alles zu finden, was er zur Therapie benötige an Blüten, Früchten, Wurzeln, Rinden, Milchsäften, Blättern und Presssäften, Bäumen usw.“ ( Opt. med. cogn . 12.3: CMG Suppl. Or. IV 125). In Latium unternahm Galen zu diesem Zweck botanische Exkursionen – im Wissen um die natürlichen Wuchsorte und die besten Erntezeitpunkte vieler Heilpflanzen in der Umgebung Roms ( De antidotis XIV 30f. K.). Aber auch von den Pharmaka, die bei Händlern zu erwerben waren, solle man sich möglichst eine Kenntnis ihrer Herkunft – das heißt bei Arzneipflanzen ihres natürlichen Standorts, bei mineralischen Heilmitteln ihrer Lagerstätten – verschaffen ( De antidotis XIV 7–9 K.). Diese Studien haben möglichst unter Anleitung eines erfahrenen Lehrers zu erfolgen; keinesfalls hinreichend sei ein bloßes Bücherstudium und das Betrachten von Pflanzenabbildungen. Auf keinen Fall solle man es nämlich „den Kapitänen gleichtun, die die Steuermannskunst nur aus Büchern erlernten“ (XI 797.1f. K.). Es sei hierzu viel Erfahrung nötig und es „genügt nicht, die Begutachtung (διάγνωσις) der Ingredienzien der Heilmittel einmalig, zweimalig oder dreimalig besorgt zu haben, sondern sehr viele Male“ ( De antidotis XIV 6.14f. K.). Wenn man keine persönliche Erfahrung mit einem Arzneimittel habe, so solle man dies auch offen bekennen ( De san. tu. VI 196.10–14 K.) – eine Maxime, der Galen selbst beispielsweise im Zusammenhang mit der im dritten Buch bei Dioskurides behandelten Pflanze asklēpias vorbildlich nachkommt (ΧΙ 840.11 Κ.). Diese Informationen mögen genügen, um eine Vorstellung von Galens Umfeld und seinen Arbeitsmöglichkeiten zu erhalten. Wie er bei der Prüfung der Heilmittel konkret vorgeht, lässt sich mehreren Passagen aus Buch IV entnehmen. Besonders deutlich wird er in Kapitel 4 und 7 (XI 632f. und 642 K., siehe hierzu auch Harig : 81–85 und Singer : 183f.): Die Prüfung der Pharmaka geschieht nicht beliebig, sondern bedarf einer genauen Vorbereitung. Zunächst muss die Testperson gesund und in bestem körperlichem Zustand (XI 641.15f. K.) sein. Die eigentliche Geschmacksprüfung muss vorher – oder überhaupt immer wieder – sorgfältig eingeübt werden (γυμνάζειν ἐπιμελῶς, XI 632.7f. K.). Dazu beginnt man mit Substanzen, die nur eine einzige Qualität aufweisen (632.9 K., vgl. 643.2 K.: μιᾶς μόνης μετέχει ποιότητος) – etwa Zwiebeln oder Knoblauch (632.10 K.), die nur die beißend-scharfe Qualität besitzen. Diese kostet man kontinuierlich (συνεχῶς), zerkaut sie dabei so gründlich wie möglich (ἐπὶ πλεῖστον bzw. ἐπιμελῶς Par. gr. 2279) und versucht, sich den Geschmackseindruck exakt (ἀκριβῶς) einzuprägen (632.11–13 K.). Danach führt man dasselbe Prozedere mit bloß adstringierenden Drogen wie Galläpfeln und Gerbersumach (ἐπὶ κηκῖδος [κικίδος Kühn] τε καὶ ῥοῦ), dann mit bitteren und süßen Substanzen durch. Als Vergleichsmaß (μέσον) für den qualitätslosen (ἄποιον) Geschmack wird Wasser empfohlen (632.16f. K.). Wer in dieser Weise möglichst reine Geschmackseindrücke memoriert hat, kann sich an Drogen mit komplexen Qualitäten versuchen. Auch wenn wir nicht wissen, wie Galen bei der eigentlichen Testphase der Simplicia im Detail vorging, wie viele Drogen er in einem Durchgang untersuchte, ob er vielleicht personelle Unterstützung hatte und sich über die Eindrücke austauschte, so darf zumindest vermutet werden, dass er sich zu den (mehrfach) untersuchten, also auch „geschmeckten“ Heilmitteln Aufzeichnungen machte und diese vielleicht geordnet nach Herkunft (Vegetabilia, Mineralia, Animalia) und nach den gängigen Bezeichnungen (halb-)alphabetisch sortiert in einer Art Kartei ablegte, aus der später das Konzept für den Katalog erstellt wurde. Man könnte auch vermuten, dass er aus dem Gedächtnis und seiner täglichen Praxis heraus seine Erfahrungen als Marginalien zur Hauptquelle hinzufügte, doch war die Materia medica von Dioskurides ja nicht alphabetisch gegliedert, weswegen eine irgendwie geartete vorherige Disponierung des Materials sicherlich zu erfolgen hatte. Seine Bemerkungen zum agarikon lesen sich ferner, als habe er die Geschmacksbeschreibung unmittelbar während oder kurz nach der Durchführung notiert – gewissermaßen noch mit dem Geschmack auf der Zunge und jedenfalls nicht aus dem Gedächtnis. Oder wäre es wahrscheinlich, dass ihm beim Abfassen des Kapitels über den Lorbeer zufällig einfällt, dass die „Wurzelrinde zwar weniger beißend und warm, dafür aber bitterer [ist] und auch eine bestimmte Adstringenz [besitzt]“ (XI 863.3–5 K.)? Ist es denkbar, dass ihm, beim Buchstaben Iota angelangt, wieder in den Sinn kommt, dass die Farnpflanze ἵππουρις/ hippouris „eine mit Bitterkeit verbundene adstringierende Qualität“ (XI 889f. K.) aufweist? Oder dass ein anderes Farngewächs δρυοπτερίς/ dryopteris „von süßer, beißender, leicht bitterer, bezüglich der Wurzel auch von herb-saurer Qualität“ (XI 865 K.), die Wurzel von ἐρυθρόδανον/ erythrodanon „herb-sauer und bitter“ (XI 878 K.) ist, oder der Sadebaum βράθυ/ brathy „Anteil an einer beißenden Qualität, ferner auch an einer Bitterkeit und Adstringenz“ (XI 853f. K.) hat? In all den genannten Fällen bietet Dioskurides weniger Informationen, und wenn man nicht voraussetzen möchte, dass Galen sich die entsprechenden Drogen vorher alphabetisch sortiert bereitgelegt hat, um den Katalog von Alpha bis Omega abfassen zu können, wird die Existenz einer Art Zettelkasten für Protokollnotizen äußerst wahrscheinlich. Ein weiteres Beispiel seines methodischen Vorgehens ist die detaillierte Analyse der Zitronatzitrone (XII 77 K.). Hinsichtlich des Geschmacks nimmt er eine dreifache Gliederung der Fruchtwand vor, die sich ja auch im Längsschnitt schon makroskopisch aufdrängt (vgl. den Kommentar und die Zeichnung von Citrus medica L. bei Haars : 308f.). So sei im Bereich des Samens die stechend-saure Qualität und trocknende Wirkpotenz vorherrschend, in der äußeren Rinde aber die beißende, während das essbare Fleisch zäh und schleimig sei, der Samen selbst aber bitter. Spätestens in diesem Kontext stellt sich die Frage, wie zutreffend seine Schilderungen sind, insbesondere wenn er von Dioskurides abweicht. Ein Vergleich mit modernen Angaben liegt allerdings nicht im Rahmen dieser Untersuchung. Es muss genügen zu bemerken, dass ein Vergleich da, wo er methodisch möglich ist, die galenischen Angaben in der Regel bestätigt (Haars : 466f.). Es wäre angesichts der starken Rezeption, die diese Art der Pharmakologie erfahren hat, auch erstaunlich, wenn es sich anders verhielte. Doch wenden wir uns noch einmal seiner Methode zu, die vielleicht am Beispiel der Zitronatzitrone am besten zum Ausdruck kommt: Zunächst werden verschiedene Pflanzenteile unterschieden und diese dann separat hinsichtlich ihres Geschmacks untersucht. Bis hierhin bewegt Galen sich durchaus im Rahmen der traditionellen, deskriptiven Pharmakognosie, auch wenn er vielleicht noch mehr an Details interessiert ist, als wir das bei Dioskurides (d. h. Sextius Niger) gesehen haben. Doch damit nicht genug: Nach Feststellung des Geschmacks, der „stechend-sauren Qualität“ des inneren Fruchtteils bemerkt Galen auch ein „trocknendes Vermögen“ und folgert aus beidem, „dass sie zum dritten Grad sowohl von den trocknenden als auch kühlenden Mitteln gehört“ (ὡς τῆς τρίτης εἶναι τάξεως ἀπὸ τῶν ξηραινόντων τε καὶ ψυχόντων, XII 77.6f. K.). Diese Folgerung hat nun nicht mehr den Charakter einer Protokollnotiz und lässt sich weder inhaltlich noch formal hinsichtlich der Verknüpfung in der vorhergehenden pharmazeutischen Literatur finden. Um zu verstehen, welche Bedeutung die Geschmackswahrnehmung für den pharmakologischen Wissenserwerb für Galen besessen hat, müssen wir im nächsten Teil also auf ebenjene Beziehung genauer eingehen. Neben der quantitativen Zunahme, Präzisierung und Korrektur von Geschmacksangaben lässt sich bei Galen auch eine Veränderung der Kontextualisierung dieser Angaben beobachten. Nur in wenigen Fällen stehen diese nämlich wie in der vorhergehenden pharmakognostischen Literatur isoliert da (z. B. XII 52.2f. K.). Meist sind sie eingebettet in eine wohldurchdachte Argumentation, die letztlich den Aufweis der durch die Elementarqualitäten und ihr Mischungsverhältnis hervorgerufenen pharmakologischen Effekte zum Ziel hat. Hierfür seien einige Beispiele genannt: „Die Wurzelrinde der Kaper hat vorherrschend eine bittere Qualität, als zweites eine beißende und danach eine saure. Damit ist klar, dass sie aus verschiedenen und streitenden Wirkungspotenzialen zusammengesetzt ist.“ Dioskurides (II 173.2) bietet übrigens keine Geschmacksbeschreibung dieser Droge. Für Galen ist der Geschmack aber nicht nur Indikator für komplexe Zusammensetzungen und Mischungen der Elementarqualitäten, sondern kann auch zur Bestimmung ihrer Intensität herangezogen werden: „Das kleine polion ist nämlich auch beißender und bitterer als das große, sodass es aus dem dritten Grad der trocknenden und dem ausgefüllten zweiten Grad der wärmenden Mittel ist“ . Die Beobachtung, dass sich Unterschiede von wilden und kultivierten Pflanzen auch im Geschmack manifestieren, wurde auch von Dioskurides öfters gemacht (u. a. Dsc. III 45, siehe oben). Dass die Ursache für den oft kräftigeren Geschmack aber die höhere Intensität der Elementarqualitäten ist, findet sich erst bei Galen: „Die wilde Kichererbse ist in allen Belangen stärker als die kultivierte, deshalb ist sie wärmender und trocknender in dem Maße, wie sie auch beißender und bitterer ist.“ Öfter noch als das Verhältnis zu den ersten Qualitäten wird der Zusammenhang mit den Arzneimittelwirkungen beleuchtet. So sei der Meerfenchel „irgendwie salzig im Geschmack und zugleich mit einer geringen Bitterkeit, und deshalb ist sein Wirkpotenzial zugleich reinigend und trocknend“ . Die Pflanze chamaidrys hingegen habe „vorherrschend die bittere Qualität, sie ist aber auch irgendwie beißend. Aus diesem Grund offensichtlich erweicht sie die Milz regelrecht und treibt den Urin und die Monatsblutung.“ Sison sei „warm und leicht bitter im Geschmack und deshalb harntreibend“ und der Pfirsichbaum habe „in den Trieben und Blättern vorherrschend eine bittere Qualität und aus diesem Grund töten seine zerriebenen und um den Bauchnabel herum gelegten Blätter die Eingeweidewürmer“. Von der Ulme wird berichtet, dass „die Rinde in höherem Grade bitter und adstringierend“ ist, „sodass sie auch den Aussatz heilt“. Es ließen sich noch beliebig viele weitere Beispiele anführen, doch ist die Eigentümlichkeit der galenischen Wissenschaftsprosa hiermit wohl bereits hinlänglich deutlich geworden. Geschmacksangaben sind stets in größere Aussagezusammenhänge durch ätiologische Begründungen oder logische Folgerungen integriert. Es werden kausale Verbindungen hergestellt zwischen den Aussagen, dass Droge x eine bestimmte Geschmacksqualität – im Folgenden G(x) – oder eine bestimmte Elementarqualität oder -wirkung E(x) zukommt, die zuweilen auch hinsichtlich ihrer Intensität I E (x) bestimmt wird, und den speziellen medizinischen Effekten oder partikularen Wirkungen P (x) der Heilpflanzen. Galen schöpft hierbei die vielfältigen Möglichkeiten der griechischen Sprache voll aus, um seine Argumentation wissenschaftlich zu fundieren. Da der Sprachverwendung Galens im Hinblick auf die epistemologische Bewertung der Geschmacksangaben eine große Bedeutung zukommt, soll diese im Folgenden für die Bücher VI bis VIII genauer untersucht werden. Hierzu wurden die Formulierungen zunächst nach ihrer Aussageabsicht eingeteilt, was allerdings in einigen Fällen reine Interpretationsfrage ist. Die folgende Gliederung versteht sich daher nur als Annäherung und Überblick mit Fokus auf der Funktion von Geschmacksurteilen und nicht als eine Untersuchung zur Argumentation Galens insgesamt. Als systematisch vorrangig ist die Verhältnisbestimmung G(x) zu E(x) in den Blick zu nehmen, die nach unserer Zählung in mindestens 48 Fällen vorliegt. Es lassen sich grundsätzlich folgende Aspekte feststellen: 33 × E(x) unbekannt; aus G(x) als Indikator wird auf E(x) zurückgeschlossen. Die Korrelation wird ausgedrückt durch: (*aus G[x] wird I E [x] bestimmt. Die Proportionalität wird ausgedrückt durch:) 9× kausale Adverbien wie διὰ τοῦτο „aus diesem Grund“ (XI 864.9 K., 889.9 K., vgl. auch XII 20.18f. K., XII 58.9 K., 114.15 K., 156.15 K.*), διὸ καὶ (XI 841.12 K.*, XII 44.7 K., 60.5 K.*, 77.9 K.*) 7× Evidenz anzeigende Ausdrücke wie ἐξ ὧν δῆλον ὡς „daraus erhellt, dass“ (XI 810.18 K., 834.11 K., 879.13 K.) oder sogar ἐξ ὧν ἁπάντων εὔδηλον (XI 813.15f. K.), ὅθεν δῆλον ὡς (XI 820.1 K.), ἐξ ὧν ἁπάντων δῆλον ὡς (XI 821.8 K.), καὶ δηλονότι (XI 851.16 K.) 7× quantifizierende Vergleiche wie εἰς ὅσον καὶ „in dem Maße wie“ (XI 866.8 K.*), τοσούτῳ … ὅσον καὶ „in einem solchen Maße … wie auch“ (XI 868.4 K.*) oder εἰς ὅσον … εἰς τοσοῦτον (XI 686.12f. K.*, 869.5 K.*, XII 24.14f. K.*, 123.18f. K.), ὅσον περ (XI 877.8 K.*) 6× konsekutive Konjunktionen wie ὥστε καὶ „sodass auch …“ (XI 833.13 K.*, XII 47.15 K.) oder einfach ὡς (XII 122.14 K.*, 107.5 K.*, 107.15 K.*, 125.18 K.*) 5× qualifizierende Vergleiche wie ὥσπερ τῇ γεύσει … οὕτως καὶ τῇ κράσει/„wie im Geschmack … so auch in der Mischung [der E]“ (XI 856.14f. K., vgl. XII 52.18 K., vgl. XII 99.5f. K., vgl. auch 112.7 K., vgl. 131.10 K.) 10× Gegebene E(x) werden durch G(x) als Verifikatoren bestätigt. Die Kausalität wird ausgedrückt durch: 8× Partikel γὰρ (XI 812.1 K., 856.11 K., 865.3 K., XII 36.8 K., 52.18 K., 70.10 K., 106.4 K., 128.18 K.) 1× ὅτι καὶ „weil auch“ (XI 815.7 K.) 1× konsekutive Konjunktion ὥστε (XII 14.17 K.) 6× Beobachtete G(x) werden durch E(x) erklärt. G(x) als Explananda. Die Kausalität wird ausgedrückt durch: 3× kausale Adverbien wie καὶ διὰ τοῦτο „und aus diesem Grund …“ (XII 69.14 K.), διὸ καὶ „weswegen auch …“ (XII 92.10, 118.12 K.) 3× Präpositionale Wendungen: ἀπό + E im Genitiv in der Bedeutung „ausgehend von E …“, XII 130.16f. K., oder mit διά + E im Akk. in der Bedeutung „aufgrund der E …“, XII 134.4 K., 135.13 K. Wie man sieht, geht es Galen überwiegend darum, in einem heuristischen, induktiven Verfahren aus gegebenen G(x) auf die sie hervorrufenden E(x) zu schließen, wobei den G(x) eine Indikatorfunktion zukommt (Fall 1). In den anderen beiden Fällen sind die E(x) bereits bekannt – etwa aus der Tradition oder durch ein vorangegangenes heuristisches Verfahren –, und die G(x) werden in einem nachträglichen deduktiven Verfahren eingeführt. Im zweiten Fall geht es also darum, die – wodurch auch immer – gegebenen E(x) durch festgestellte G(x) zu bestätigen, wobei ihre Funktion als Verifikatoren impliziert wird. Im dritten Fall geht es darum, die gegebenen G(x) mittels der E(x) zu erklären. Da Galen das induktive und deduktive Verfahren in den einzelnen Kapiteln nicht kombiniert, kommt seine Argumentation nicht sofort in den Verdacht, zirkulär zu sein. Für alle drei Muster hat Galen bestimmte Ausdrucksweisen, die er präferiert. Die andere Gruppe von Aussagen, die betrachtet werden muss, betrifft die Relation der Geschmacksqualitäten zu den ebenfalls empirisch beobachteten und tradierten Arzneimittelwirkungen. 4. 29× aus G(x) wird P (x) gefolgert: Die Kausalität wird ausgedrückt durch: 15× konsekutive Konjunktionen wie ὥστε καὶ „sodass auch …“ (XI 819.12 K., 851.1 K., 855.5 K., 886.18 K., XII 13.8 K., 41.16 K., 42.2 K., 52.11 K., 109.7 K., 131.6 K., 155.14 K.), oder einfach ὡς (XII 56.16 K., 93.18f. K., 97.18 K., 143.17 K.) 8× folgernde Partikel „folglich, daher, aus diesem Grund“ wie τοιγαροῦν (XI 891.8 K., XII 102.13 K., 106.15 K., 121.3 K., 126.3 K.), τοίνυν (XI 888.14 K.), ἄρα (XI 859.3 K., 863.5 K.) 6× Evidenz anzeigende Ausdrücke wie ᾧ καὶ δῆλον „daraus erhellt, dass …“ (XII 9.12f. K.), δῆλον ὅτι καὶ … usw. (XI 824.6 K., 879.13 K.), (ἐξ ὧν) δηλονότι (XII 77.16 K., 153.9 K.), εὐλογῶς (XII 84.15f. K.) 5. 71× Gegebene P (x) wird durch G(x) begründet. Die Kausalität wird ausgedrückt durch: 44× kausale Adverbien wie ὅθεν „wodurch, weshalb“ (XI 878.9 K., 884.1 K., XII 9.1 K., 40.12 K., 41.9 K., 54.12 K., 69.2 K., 88.14 K., 103.2 K., 120.10 K., 136.2 K.), διὸ καὶ „weswegen auch …“ (XI 819.3 K., 826.15 K., 830.12 f. K., 858.4f. K., 865.17 K., 884.10 K., XII 43.14 K., 44.7 K., 89.5 K., 92.6 K., 96.11 K., 121.7 K., 123.15 K., 154.2 K., 157.18 K.), καὶ διὰ τοῦτο „und aus diesem Grund“ (XI 834.19f. K., 837.17 K., 845.5 K., 853.2f. K., 853.12 K., 861.6 K.,862.5 K., 864.9 K., 880.4 K., XII 49.10 K., 54.4 K., 55.8 K., 68.16 K., 76.11 K., 93.7f. K., 128.8 K., vgl. 131.6 K., 147.2 K.) 14× präpositionale Wendung διά + G im Akk. „aufgrund“ (XI 877.2 K., 879.4 K., XII 11.14 K., 12.18 K., 41.2–5 K., 60.8 K., 64.13 K., 79.4 K., 80.12 K., 85.14 K., 113.13f. K., 127.6 K., 136.16 K., 142.7 K.) 5× Partizipialkonstruktion mit kausalem Sinn, z. B. XI 856.7 K., 867.13 K., XII 57.1 f. K., 152.16 K., 153.15 K. 4× Partikel γάρ „denn“ (XI 815.16 K., XII 23.17f. K., 68.6 K., 109.17 K.) 3 × G im instrumentalen Dativ in der Bedeutung „durch (oder vermittels) G wirkt die Droge …“: XI 883.11f. K., XII 9.15f. K., 101.1 K. 1× Konstruktion mit dem Relativadverb ᾗ „insofern“ (XI 859.6f. K.) 6. 10 × G(x) und P (x) korrelieren in einem nicht näher bestimmten kausalen Verhältnis. Die Korrelation wird ausgedrückt durch: 7× (ὅμοιον) κατά τε τὴν γεῦσιν καὶ κατὰ τὴν ἐνέργειαν (auch δύναμιν), XI 890.17f. K., XII 15.5f. K., XII 19.10–18 K., vgl. auch XII 27.1f. K., 107.17f. K., 116.4 K., 155.1–3 K. 2× korrelatives μέν – δέ: „einerseits … andererseits“, Bsp. καὶ γευομένη μέν … καὶ τοῖς ἔργοις δὲ, XI 888.5f. K., vgl. auch XII 93.10f. K. 1× quantitativer Vergleich: εἰς ὅσον … εἰς τοσοῦτον (XII 91.17 K.) Die Untersuchung der Verhältnisbestimmung G(x) zu P (x) ergibt zunächst – für uns unerwartet –, dass diese mit 110 Instanzen mehr als doppelt so häufig auftritt wie E(x) zu G(x). Dies ist insofern überraschend, als es ja Galen vornehmlich um den Aufweis der E(x) beziehungsweise I E (x) geht und G und P nur indirekt über E zusammenhängen. Man könnte daher vermuten, dass Galen hier mit verkürzten Syllogismen operiert, indem er den Zusammenhang mit E impliziert, und dass er G nicht tatsächlich als Ursache für P ansieht (dies bedürfte einer eingehenden Untersuchung, die hier nicht geleistet werden kann). Der insgesamt häufigste Fall (5) ist, dass ähnlich dem deduktiven Verfahren von 2) oder 3) gegebene P (x) durch gegebene G(x) eine Begründung oder Bestätigung erfahren, was in den meisten Fällen (5a) durch kausale Adverbien beziehungsweise Konjunktionen geschieht. Auch hinter den Formulierungen von Fall (5) steht eine ähnliche Intention, indem aus einer gegebenen G(x) eine P (x) gefolgert wird. Das letzte Muster schließlich erfasst Aussagen, die G(x) und P (x) miteinander wechselseitig korrelieren, ohne dass die Abhängigkeit thematisiert wird. Abgesehen von der ebenfalls häufig auftretenden Verhältnisbestimmung von E(x) zu P (x) unter Nichtbeachtung von G(x), die für unser Thema aber nicht unmittelbar von Bedeutung ist und daher hier nicht untersucht wurde, haben wir die wichtigsten G(x)-Argumentationsmuster in den Büchern VI bis VIII damit erfasst. Grundsätzlich zeigt sich, dass die Feststellung von Geschmacksqualitäten eine zentrale Stellung in Galens pharmakologischen Analysen einnimmt, insofern diese mit allen anderen Variablen in Verbindung gebracht werden und ihre Verknüpfung, wie wir gesehen haben, durch ganz unterschiedliche Arten von Junktoren auch wechselseitig geschieht. Die Geschmackswahrnehmung erfüllt dadurch nicht nur die Funktion, Arzneimittelwirkungen und ihre zugrunde liegenden Wirkungspotenziale anzuzeigen, sondern dient auch dazu, letztere zu bestätigen und liefert überhaupt auch den Anstoß für pharmakologische Erklärungsversuche. Insofern können wir sagen, dass die Geschmackswahrnehmung eine dreifache Funktion für den Erwerb von Wissen in der speziellen Pharmakologie Galens erfüllt: 1.) eine Indikatorfunktion, 2.) eine Kontroll- oder Verifikatorfunktion und 3.) eine Funktion als Lieferant erklärungsbedürftiger Phänomene (Explananda), die Galen zum Anlass ätiologischer Betrachtungen nimmt. In Friedrich August Flückigers (1828–1894) im Jahr 1867 erschienenem Lehrbuch Pharmakognosie des Pflanzenreichs , mit dem er die gleichnamige Disziplin nach herrschender Meinung auf eine wissenschaftliche Grundlage stellte, begegnet uns eine Einteilung, die uns eigentümlich bekannt vorkommt: Schon im Inhaltsverzeichnis werden Wurzeln und Rhizome klassifiziert in „aromatische, von schleimigem oder süßem Geschmacke, adstringierende, bitterliche oder bittere, von kratzendem oder scharf-brennendem Geschmacke“. Genauso geht er dann auch bei den Rinden, Blättern und Früchten vor – und stellt somit in einer Zeit, als chemische Methoden längst eine Art Deutungshoheit in der Pharmazie reklamiert hatten, „an die Sinnesorgane höchste Ansprüche“ (Haug : 289). Zwar mag die Einteilung aus pragmatischen Überlegungen erfolgt sein. Dennoch sollte es uns zu denken geben, dass in dieser Disziplin, der im 19. Jahrhundert ein ganz anderes Verständnis von Wissenschaft zugrunde liegt, eine Systematik maßgeblich auf Grundlage der Geschmackswahrnehmung entwickelt werden und paradigmatisch Eingang in ein Lehrbuch für Pharmazeut*innen finden konnte. Die Geschmacksprüfung der Vegetabilia ist, wie oben ausgeführt, noch immer ein Gegenstand des Pharmaziestudiums. Auch der Autor des vorliegenden Artikels hat auf diese Weise die Leistungsfähigkeit der gustatorischen Methode erfahren, was – und das soll nicht verschwiegen werden – sein Verständnis der galenischen Pharmakologie und ihrer wissenschaftsgeschichtlichen Bedeutung nicht unwesentlich beeinflusst hat. Obwohl Galens Argumentation, wie wir ansatzweise gesehen haben, unter den Verdacht fällt, zirkulär zu sein und aus heutiger Sicht auf vollkommen falschen Prämissen beruht, kann man ihr eine gewisse Überzeugungskraft nicht absprechen. Hierbei haben möglicherweise gerade die Geschmacksurteile einen großen Anteil, suggerieren sie doch dem/der Leser*in, dass man sich stets auf festem empirischem Boden bewegt. So mag auch die kühnste Hypothese akzeptiert werden, wenn die Argumentation mit der Feststellung abschließt, dass Absinth bitter schmeckt. Doch der starke Einbezug der Geschmackswahrnehmung ist viel mehr als nur ein rhetorischer Trick. Mit ihrer Hilfe gewinnt der Pharmakologe überhaupt erst ein vollständiges Bild, was die Wirkung eines Heilmittels ausmacht, und kann (innerhalb bestimmter Grenzen) zu einer Einteilung gelangen, die sich in der Therapie fruchtbar machen lässt (etwa durch die Einteilung Adstringentia – Bitterstoffdrogen – Kaustika). Es könnte aufschlussreich sein zu untersuchen, ob das starke Eintreten für die Geschmacksprüfung der Simplicia bei Galen implizit oder explizit seinen Niederschlag in den drogenkundlichen Werken bis in die Neuzeit gefunden und wie Galens Lesern seine Lehre geschmeckt hat. Die Schwierigkeiten, die sich etwa durch die Übersetzung der Geschmacksqualitäten ergeben, sollten hierfür jedenfalls kein unüberwindbares Hindernis darstellen. Mit ihnen waren die Rezipienten des Galenismus im Lauf der Geschichte regelmäßig konfrontiert und haben sie durch unterschiedliche hermeneutische Anstrengungen zu bewältigen versucht. Dabei wurden die Pharmaka auch immer wieder neu geschmeckt und die Ergebnisse mit den antiken Autoren verglichen. Das pharmakologische System Galens bietet in seiner historischen Entwicklung somit tiefe Einblick in eine andere, sinnliche Forschungspraxis, in der nur Fortschritte macht, wer bereit ist, über Geschmack zu streiten. |
The FIGO ovulatory disorders classification system | 8cbaba02-a4bb-4b77-9d21-a5a1852a3b49 | 10086853 | Gynaecology[mh] | INTRODUCTION Ovulatory disorders are common in girls and women of reproductive age and are associated with episodic or chronic dysfunction of the hypothalamic–pituitary–ovarian (H‐P‐O) axis. , These disorders may adversely affect quality of life when they manifest with infertility or as aberrations in menstrual function. Menstrual symptoms may include altered frequency or regularity of flow, as well as prolonged or heavy menstrual bleeding (HMB), or even a complete absence of menstrual blood flow, referred to as amenorrhea. Reproductive function may be adversely impacted as chronic anovulation is a common cause of infertility. While there are numerous known causes and contributors to ovulatory disorders, the entire spectrum of mechanisms of pathogenesis remains to be fully elucidated. Ovulatory disorders are often associated with underlying endocrinopathies, neoplasms, psychological and psychiatric conditions, and the use of specific pharmacologic agents. Optimally effective research, teaching, and clinical management of ovulatory disorders has been impeded by the absence of a comprehensive, internationally recognized and utilized structured classification system. The WHO system for ovulatory disorders was first presented as a monograph in 1973 and has been modified over time in various reviews and book chapters by single authors rather than international consensus. Some 50 years later, much more is known about ovulatory disorders. As a result, the International Federation of Gynecology and Obstetrics (FIGO) has undertaken a process whereby the global community of stakeholders involved with ovulatory disorders has designed a new system to better meet the needs of investigators, clinicians, and medical educators worldwide. The development of the system started with the formation of an Ovulatory Disorders Steering Committee (ODSC) comprising members of FIGO's Committee on Menstrual Disorders (MDC) (now the Committee on Menstrual Disorders and Related Health Impacts, or MDRHI) and Committee on Reproductive Medicine, Endocrinology, and Infertility. The involvement of the MDRHI reflects the common and important impact of ovulatory disorders on menstrual bleeding experience, an entity referred to as AUB‐O in FIGO System 2 (see below). BACKGROUND AND RATIONALE 2.1 Defining ovulatory disorders In the reproductive years—and in the absence of pregnancy, the process of lactation, or the use of pharmacological agents such as contraceptive steroids—the normal woman releases a mature oocyte from a Graafian follicle in a relatively predictable and cyclical fashion. However, a consensus definition of ovulatory disorders, sometimes called ovulatory dysfunction, has been lacking. The notion of anovulation or absent ovulation is but one manifestation, but there exists a spectrum of chronic or episodic conditions or circumstances that also disrupt the predictable and cyclical ovulatory process. Previously, infrequent ovulation has been termed “oligo‐ovulation,” which typically, but not always, manifests with some combination of infrequent and irregular onset of menstruation as defined in FIGO AUB System 1 (FIGO discontinued the term oligomenorrhea). However, and recognizing that many women with ovulatory disorders may have normal‐length menstrual cycles, no clear definition of infrequent ovulation has been adopted, and this was not addressed in the joint “Committee Opinion” on Infertility Workup for the Women's Health Specialist produced by the American College of Obstetricians and Gynecologists and the American Fertility Society. Furthermore, while an occasional failure to ovulate is expected and may not contribute to infertility, it may well cause an episode of delayed onset of menses and even HMB. This circumstance begs the inclusion of intermittent anovulation in a broad‐based, all‐encompassing definition of ovarian dysfunction. An additional consideration is other aberrations in ovulatory function, such as the luteinized unruptured follicle (LUF) , and the luteal out of phase (LOOP) events 9 that represent, respectively, mechanical failure to release the mature oocyte and the premature recruitment of follicles in the luteal phase, each of which could be candidates for inclusion in the definition of ovulatory dysfunction. As a result of these considerations, it is apparent that there is an unmet need for both a revised definition of ovulatory disorders and a consensus classification system designed to guide research, education, and clinical care across disciplines. 2.2 Existing “system” and its value and limitations The original WHO classification presented three types of ovulatory dysfunction. Group I included “women with amenorrhea and with little or no evidence of endogenous estrogen activity, including patients with (a) hypogonadotrophic ovarian failure, (b) complete or partial hypopituitarism, or (c) pituitary‐hypothalamic dysfunction.” Group II was described as “Women with a variety of menstrual cycle disturbances (including amenorrhea) who exhibit distinct estrogen activity (urinary estrogens usually <10 mcg/24 h), whose urinary and serum gonadotrophins are in the normal range and fluctuating, and who may also have fairly regular spontaneous menstrual bleeds (i.e. 24–38 days apart) but without ovulation.” Group III was described as “Females with primary ovarian failure (sic, now known as primary ovarian insufficiency; POI) associated with low endogenous estrogen activity and pathologically elevated serum and urinary gonadotrophins.” This classification illustrates the now‐outdated assay methodology of the time. A second monograph was published in 1976, which presented an algorithm based upon whether the serum prolactin concentration was elevated or normal, the response to a progestagen challenge test to assess estrogenization, and whether the serum follicle‐stimulating hormone (FSH) concentration was elevated or normal. The results of these assays were to be used to define seven groups: Group I: Hypothalamic pituitary failure Group II: Hypothalamic pituitary dysfunction Group III: Ovarian failure Group IV: Congenital or acquired genital tract disorders Group V: Hyperprolactinemia, with a space‐occupying lesion Group VI: Hyperprolactinemia, with no detectable space‐occupying lesion Group VII: Non‐functioning hypothalamic/pituitary tumors Over the last 40 years, numerous descriptions of the WHO classification have appeared in various monographs and book chapters in textbooks on gynecology, infertility, and reproductive endocrinology. Multiple authors have modified the classification without any evidence of further scientific discussion or consensus development. Interestingly, the UK NICE Guidelines on the investigation and management of infertility, first published in 2004, describe three groups with reference to the WHO Manual for the Standardized Investigation and Diagnosis of the Infertile Couple , published in 1993. Yet this WHO manual does not contain any classification of ovulatory disorders. Nonetheless, the NICE classification encompasses the three groups that most authors refer to currently, namely: Group I: Low gonadotropins and estradiol Group II: “Gonadotropin disorder” and normal estradiol Group III: High gonadotropins and low estradiol In this classification, Group I essentially refers to hypogonadotropic hypogonadism and pituitary insufficiency but also includes hyperprolactinemia. Group II is often referred to as “hypothalamic/pituitary dysfunction,” and most consider this group to primarily comprise women with polycystic ovary syndrome (PCOS), while Group III is consistently primary ovarian insufficiency (POI). However, it is essential to appreciate that hormone levels do not obey clear rules. For example, in those with hypothalamic amenorrhea who are underweight, levels of serum luteinizing hormone (LH) are usually suppressed, while levels of FSH are often in the normal range. , In addition, women with PCOS often have levels of FSH and LH in the normal range. Furthermore, anovulation is only one extreme of ovulatory dysfunction that includes a spectrum of manifestations that range from isolated episodes to chronic ovulatory failure. Since the first iterations of the WHO classification, there have been significant advances in understanding the control of ovulation and the pathophysiology of ovulatory disorders, together with improvements in assay technology and genomics. Consequently, there exists a need for a more comprehensive and updated classification. 2.3 The FIGO Systems for Abnormal Uterine Bleeding ( AUB ) in the Reproductive Years In 2011, and again in 2018, FIGO published its two systems for describing nongestational AUB in the reproductive years, including System 2, the classification system known as “PALM‐COEIN” that categorizes causes of AUB in non‐gravid women of reproductive age, including those with ovulatory disorders (AUB‐O). These systems were developed and designed using a rigorous Delphi process, with the participants including international experts and representation from multiple and diverse stakeholder organizations, including national and subspecialty societies and journals and the US Food and Drug Administration. The overall process also included an examination of the available population databases dealing with menstruation that resulted in new, evidence‐based definitions for normal and abnormal menstrual metrics that are now known as the FIGO AUB System 1. , , The process has been iterative, with periodic revisions of systems that reside in what is described as a “living document.” The whole process has been underpinned and continues to be supported by FIGO and the FIGO Committee on Menstrual Disorders (MDC), which, since 2022, has been known as the Committee on Menstrual Disorders and Related Health Impacts. FIGO AUB System 1 describes non‐gestational normal and AUB in the reproductive years and addresses the features of menstruation, that is, frequency, regularity, duration, and perceived volume of menstrual blood loss in addition to the presence of bleeding between periods (intermenstrual bleeding) as well as unscheduled bleeding associated with the use of gonadal steroids for contraception. The latter is now encompassed by the increasingly used term “contraceptive‐induced menstrual bleeding changes” (CiMBC). Notably, System 1 is currently based upon data from studies of women aged 18–45 years, as evidence from adolescent girls and women in the late reproductive years is less well defined. The second system, FIGO AUB System 2, describes potential causes or contributors to symptoms of AUB that are categorized in System 1. The nine categories, arranged according to the acronym PALM‐COEIN, are as follows: Polyp (AUB‐P); Adenomyosis (AUB‐A); Leiomyoma (AUB‐L); Malignancy and hyperplasia (AUB‐M); Coagulopathy (AUB‐C); Ovulatory dysfunction (AUB‐O); Endometrial disorders (AUB‐E); Iatrogenic (AUB‐I); and Not otherwise classified (AUB‐N). For the present context, ovulatory disorders (AUB‐O) incorporate a range of disturbances in normal ovulatory function ranging from irregular to infrequent to absent ovulation. To date, in the context of management of patients with AUB, the diagnosis of ovulatory disorders has been based mainly on a detailed menstrual history to meet the parameters that comprise FIGO System 1. In the 2018 revisions of the two FIGO systems, the recommendation was made that treatments that may interfere with the H‐P‐O axis and associated with AUB be placed within the “AUB‐I" category. The rationale and methodology for developing a sub‐classification system for AUB‐O are now presented. Defining ovulatory disorders In the reproductive years—and in the absence of pregnancy, the process of lactation, or the use of pharmacological agents such as contraceptive steroids—the normal woman releases a mature oocyte from a Graafian follicle in a relatively predictable and cyclical fashion. However, a consensus definition of ovulatory disorders, sometimes called ovulatory dysfunction, has been lacking. The notion of anovulation or absent ovulation is but one manifestation, but there exists a spectrum of chronic or episodic conditions or circumstances that also disrupt the predictable and cyclical ovulatory process. Previously, infrequent ovulation has been termed “oligo‐ovulation,” which typically, but not always, manifests with some combination of infrequent and irregular onset of menstruation as defined in FIGO AUB System 1 (FIGO discontinued the term oligomenorrhea). However, and recognizing that many women with ovulatory disorders may have normal‐length menstrual cycles, no clear definition of infrequent ovulation has been adopted, and this was not addressed in the joint “Committee Opinion” on Infertility Workup for the Women's Health Specialist produced by the American College of Obstetricians and Gynecologists and the American Fertility Society. Furthermore, while an occasional failure to ovulate is expected and may not contribute to infertility, it may well cause an episode of delayed onset of menses and even HMB. This circumstance begs the inclusion of intermittent anovulation in a broad‐based, all‐encompassing definition of ovarian dysfunction. An additional consideration is other aberrations in ovulatory function, such as the luteinized unruptured follicle (LUF) , and the luteal out of phase (LOOP) events 9 that represent, respectively, mechanical failure to release the mature oocyte and the premature recruitment of follicles in the luteal phase, each of which could be candidates for inclusion in the definition of ovulatory dysfunction. As a result of these considerations, it is apparent that there is an unmet need for both a revised definition of ovulatory disorders and a consensus classification system designed to guide research, education, and clinical care across disciplines. Existing “system” and its value and limitations The original WHO classification presented three types of ovulatory dysfunction. Group I included “women with amenorrhea and with little or no evidence of endogenous estrogen activity, including patients with (a) hypogonadotrophic ovarian failure, (b) complete or partial hypopituitarism, or (c) pituitary‐hypothalamic dysfunction.” Group II was described as “Women with a variety of menstrual cycle disturbances (including amenorrhea) who exhibit distinct estrogen activity (urinary estrogens usually <10 mcg/24 h), whose urinary and serum gonadotrophins are in the normal range and fluctuating, and who may also have fairly regular spontaneous menstrual bleeds (i.e. 24–38 days apart) but without ovulation.” Group III was described as “Females with primary ovarian failure (sic, now known as primary ovarian insufficiency; POI) associated with low endogenous estrogen activity and pathologically elevated serum and urinary gonadotrophins.” This classification illustrates the now‐outdated assay methodology of the time. A second monograph was published in 1976, which presented an algorithm based upon whether the serum prolactin concentration was elevated or normal, the response to a progestagen challenge test to assess estrogenization, and whether the serum follicle‐stimulating hormone (FSH) concentration was elevated or normal. The results of these assays were to be used to define seven groups: Group I: Hypothalamic pituitary failure Group II: Hypothalamic pituitary dysfunction Group III: Ovarian failure Group IV: Congenital or acquired genital tract disorders Group V: Hyperprolactinemia, with a space‐occupying lesion Group VI: Hyperprolactinemia, with no detectable space‐occupying lesion Group VII: Non‐functioning hypothalamic/pituitary tumors Over the last 40 years, numerous descriptions of the WHO classification have appeared in various monographs and book chapters in textbooks on gynecology, infertility, and reproductive endocrinology. Multiple authors have modified the classification without any evidence of further scientific discussion or consensus development. Interestingly, the UK NICE Guidelines on the investigation and management of infertility, first published in 2004, describe three groups with reference to the WHO Manual for the Standardized Investigation and Diagnosis of the Infertile Couple , published in 1993. Yet this WHO manual does not contain any classification of ovulatory disorders. Nonetheless, the NICE classification encompasses the three groups that most authors refer to currently, namely: Group I: Low gonadotropins and estradiol Group II: “Gonadotropin disorder” and normal estradiol Group III: High gonadotropins and low estradiol In this classification, Group I essentially refers to hypogonadotropic hypogonadism and pituitary insufficiency but also includes hyperprolactinemia. Group II is often referred to as “hypothalamic/pituitary dysfunction,” and most consider this group to primarily comprise women with polycystic ovary syndrome (PCOS), while Group III is consistently primary ovarian insufficiency (POI). However, it is essential to appreciate that hormone levels do not obey clear rules. For example, in those with hypothalamic amenorrhea who are underweight, levels of serum luteinizing hormone (LH) are usually suppressed, while levels of FSH are often in the normal range. , In addition, women with PCOS often have levels of FSH and LH in the normal range. Furthermore, anovulation is only one extreme of ovulatory dysfunction that includes a spectrum of manifestations that range from isolated episodes to chronic ovulatory failure. Since the first iterations of the WHO classification, there have been significant advances in understanding the control of ovulation and the pathophysiology of ovulatory disorders, together with improvements in assay technology and genomics. Consequently, there exists a need for a more comprehensive and updated classification. The FIGO Systems for Abnormal Uterine Bleeding ( AUB ) in the Reproductive Years In 2011, and again in 2018, FIGO published its two systems for describing nongestational AUB in the reproductive years, including System 2, the classification system known as “PALM‐COEIN” that categorizes causes of AUB in non‐gravid women of reproductive age, including those with ovulatory disorders (AUB‐O). These systems were developed and designed using a rigorous Delphi process, with the participants including international experts and representation from multiple and diverse stakeholder organizations, including national and subspecialty societies and journals and the US Food and Drug Administration. The overall process also included an examination of the available population databases dealing with menstruation that resulted in new, evidence‐based definitions for normal and abnormal menstrual metrics that are now known as the FIGO AUB System 1. , , The process has been iterative, with periodic revisions of systems that reside in what is described as a “living document.” The whole process has been underpinned and continues to be supported by FIGO and the FIGO Committee on Menstrual Disorders (MDC), which, since 2022, has been known as the Committee on Menstrual Disorders and Related Health Impacts. FIGO AUB System 1 describes non‐gestational normal and AUB in the reproductive years and addresses the features of menstruation, that is, frequency, regularity, duration, and perceived volume of menstrual blood loss in addition to the presence of bleeding between periods (intermenstrual bleeding) as well as unscheduled bleeding associated with the use of gonadal steroids for contraception. The latter is now encompassed by the increasingly used term “contraceptive‐induced menstrual bleeding changes” (CiMBC). Notably, System 1 is currently based upon data from studies of women aged 18–45 years, as evidence from adolescent girls and women in the late reproductive years is less well defined. The second system, FIGO AUB System 2, describes potential causes or contributors to symptoms of AUB that are categorized in System 1. The nine categories, arranged according to the acronym PALM‐COEIN, are as follows: Polyp (AUB‐P); Adenomyosis (AUB‐A); Leiomyoma (AUB‐L); Malignancy and hyperplasia (AUB‐M); Coagulopathy (AUB‐C); Ovulatory dysfunction (AUB‐O); Endometrial disorders (AUB‐E); Iatrogenic (AUB‐I); and Not otherwise classified (AUB‐N). For the present context, ovulatory disorders (AUB‐O) incorporate a range of disturbances in normal ovulatory function ranging from irregular to infrequent to absent ovulation. To date, in the context of management of patients with AUB, the diagnosis of ovulatory disorders has been based mainly on a detailed menstrual history to meet the parameters that comprise FIGO System 1. In the 2018 revisions of the two FIGO systems, the recommendation was made that treatments that may interfere with the H‐P‐O axis and associated with AUB be placed within the “AUB‐I" category. The rationale and methodology for developing a sub‐classification system for AUB‐O are now presented. METHODOLOGY The approach selected was based on RAND Delphi methodology, extensively used for consensus development processes, including classification systems for medical conditions. The two FIGO systems for AUB in the reproductive years, the sub‐classification systems for leiomyomas (AUB‐L) and adenomyosis (AUB‐A), now undergoing validation, have all been developed using a version of this process. , , The project was submitted to and approved by the FIGO Executive, and FIGO's Education Communication and Advocacy Consortium (ECAC) approved the results before submission of the manuscript. 3.1 Ovulatory Disorders Steering Committee The first step was to form an Ovulatory Disorders Steering Committee (ODSC) comprising members of FIGO's MDC (now MDRHI) and Committee on Reproductive Medicine, Endocrinology, and Infertility. The chairs of each of these committees collaborated to form the ODSC by identifying eight members from their committees, adding an external member who had a leadership position in the Global PCOS Alliance. The resulting nine‐member committee had diverse reach and comprised one from each of the continents of Africa, Asia, and North America, and two from each of the European Union, the United Kingdom, and South America. The ODSC met at regular intervals between June and December 2020 to identify and engage stakeholders and develop and test the consensus process. The scope of the ODSC also included review and analysis of the results of the various rounds and the design and testing of subsequent Delphi rounds. 3.2 Stakeholder and participant identification The first task of the ODSC was to identify and engage the appropriate stakeholders necessary for the Delphi process. The chosen categories included the following: National obstetrical and gynecological societies Subspecialty societies representing reproductive endocrinologists Specialty (obstetrics and gynecology) and subspecialty (reproductive endocrinology and infertility) journals Recognized experts in ovulatory disorders not participating in categories 1–3 Lay organizations interested in infertility, AUB, or PCOS Descriptive letters were created and customized for the various categories describing the rationale for the process and a synopsis of the methodology. Via the FIGO record of member countries, each of the national obstetrical and gynecological societies was contacted and invited by email to support the process by naming a representative. The ODSC identified the spectrum of subspecialty societies on the six continents and contacted leadership to explain the process and solicit support. The descriptive letter was sent electronically to both the society headquarters and the identified participant. A similar process involved the editorial offices of relevant specialty and subspecialty journals. The ODSC then identified recognized experts based on a combination of personal knowledge of the field and a search of the literature, subtracting those identified by national societies, subspecialty societies, or journals for representation. Finally, the ODSC sought to identify lay organizations that could represent women and adolescent girls who may have ovulatory disorders. These groups were generally contacted directly, and if there was interest and an indication of commitment, a lay‐based version of the letter was sent. 3.3 The Delphi consensus process 3.3.1 | Background and scoring system The Delphi process was developed by the RAND Corporation as a method for determining multi‐stakeholder expert consensus in a semi‐anonymous fashion that minimizes the impact of interpersonal issues on the outcome. Originally designed to forecast the impact of technology on warfare, it has subsequently been utilized across a number of disciplines including health care. Versions of the Delphi Process were used previously in the development of the FIGO AUB systems , , and are generally similar to the original RAND system comprising a series of survey rounds designed to be administered in a web‐based or live environment with electronic scoring. Members of the ODSC did not participate in the Delphi process as participants. The scoring system has nine levels (1–9), with “1” being the most substantial disagreement with a statement, “9” the strongest agreement, and “5” representing neutrality. Scores in the top tertile (7, 8, and 9) indicated “agreement” with a statement, while those in the bottom tertile (1, 2, and 3) were indications of disagreement. As a result, the remaining scores (4, 5, and 6) comprised the “neutral” category, with “4” leaning to disagreement and “6” leaning to agreement. The minimum requirement for consensus agreement was a mean score of at least 7 (scores of 6.5–6.9 were rounded to 7), with no more than 15% in the disagreement category. Conversely, “disagreement” was defined as a mean score of 3 or less (scores of 3.1–3.4 were rounded to 3), with no more than 15% in the agreement category. For each statement or question in a survey, there is a field to allow for free‐text comments by the participants. 3.3.2 | Participant orientation meeting Before distributing the first round of surveys, two orientation meetings for the participants were held to ensure that the appropriate contact information was in the study database and systems and that all understood the survey mechanisms. The two meetings were held on the Zoom platform (Zoom Video Communications Inc, San Jose, CA, USA), with dates and times selected to facilitate flexibility for the diverse group of participants, particularly considering the spectrum of world time zones involved. Included in the messaging of this meeting was the understanding that Delphi participant answers would remain confidential and that all distributions would be anonymized. Demonstrations of the functionality of the system were provided. A session was recorded and uploaded to an accessible server for individuals who could not attend either of the live, web‐based meetings and to provide a resource for all participants who wished to review the instructions on their own time. It is to be noted that the lay component of the process was planned to occur after the medical stakeholders had developed a draft system. 3.3.3 | Conduct of the first round The first round of the Delphi process was designed to identify the participants' age, gender, location, expertise, and constituency and evaluate general opinions, the latter using statements intended to elicit an “agree” or “disagree” response. These statements were crafted in a fashion that invited and measured opinions regarding the clinical relevance of ovulatory disorders, the need for a well‐designed classification system, and the broad categories that should be included if such a system was to be designed. The draft set of questions was created by the Chair of the ODSC, reviewed by the committee members in meetings using the Zoom platform, and then tested on the web‐based survey instrument SurveyMonkey (Momentive, San Mateo, CA, USA). The final version of the first round was distributed to the stakeholders via their identified email addresses within the web‐based survey system. The ODSC Chair, who also functioned as the Facilitator, kept track of responses and sent out reminder emails at intervals of 7–10 days until there were no additional responses. The data were then exported to an Excel (Microsoft Corp, Everett, WA, USA) workbook comprising spreadsheets containing the survey template that automatically calculated means and the percentage of answers in the agree (7–9), neutral (4–6), and disagree (1–3) categories. The free‐text comments made by the participants were also included in the spreadsheet. The ODSC reviewed these data as a prelude to the design of the second round. The aggregate anonymized results were sent to each participant along with a copy of their responses for comparative purposes. 3.3.4 | Conduct of the second round The second‐round survey was constructed, in part, based upon the first‐round results. Some “neutral” responses that had marginal scores close to 3 or 7, or defined principally by the outliers, were reviewed in particular because, in such circumstances, it was possible that rewording a question or providing appropriately representative evidence would result in a change in the participant's opinion. It was also possible that “re‐asking” the question in the context of individual participant understanding of the group response might result in changes in individual responses. This information allowed the ODSC to construct a second survey round that eliminated items with defined agreement or disagreement but included reworded statements and new statements seeking to refine and expand the criteria that the participants thought necessary. The distribution of the second‐round survey was confined to those participating in and responding to the first round. The web‐based system, distribution, and follow‐up reminder technique were again employed. The data were retrieved, exported into the same Excel workbook with worksheet templates, and analyzed by the ODSC. Similarly, the participants received an anonymized summary of the participant responses to each of the items and a copy of their answers for comparison. At this point, the committee had enough information to design a draft system that addressed and included the elements identified in the first two Delphi rounds. This was conducted iteratively until a draft acceptable to all ODSC members was created. 3.3.5 | Conduct of the third round As a prelude to the live stakeholder meeting, a short clarifying third round was created, tested, distributed, and the results analyzed by the ODSC, conducted in a fashion similar to that of the first two rounds. Included in this round was a version of the draft system with solicitation of preliminary opinions from the participants. As was the case for the first two rounds, each participant was provided an anonymized copy of the results of the previous round and a copy of their responses, all for review before the live participant meeting. 3.3.6 | Participant meeting All medical participants and the ODSC were invited to participate in the stakeholder meeting held live on the Zoom platform. Here, the overall results of the survey rounds were presented, including those items where consensus one way or the other had not been reached. The draft system was also reviewed. An open discussion was invited, and preliminary polls were taken using the system available on the Zoom platform. 3.3.7 | Post‐meeting and fourth survey round The ODSC undertook the post‐meeting analysis. Subsequently, a short fourth‐round poll was conducted to reach a consensus on the remaining elements and include individuals who could not participate in the live meeting. 3.3.8 | Lay round The lay round was designed to query the lay representatives, both for their perception of a need for a classification system and their opinions of the system developed by the expert and representative participants. A separate survey was designed that included some of the items in the medical participant rounds but presented in a fashion accessible by a lay audience. There was a focus on their opinions of clarity and utility in the context of discussion and counseling involving healthcare practitioners and patients. The draft lay‐round elements were reviewed and revised by the ODSC, uploaded to the SurveyMonkey platform, tested, and then distributed to the participants in a fashion similar to that used for the medical participant rounds. The results were reviewed and analyzed by the ODSC, who considered these opinions in revising the system and constructing the manuscript and the design of materials for the lay audience. Ovulatory Disorders Steering Committee The first step was to form an Ovulatory Disorders Steering Committee (ODSC) comprising members of FIGO's MDC (now MDRHI) and Committee on Reproductive Medicine, Endocrinology, and Infertility. The chairs of each of these committees collaborated to form the ODSC by identifying eight members from their committees, adding an external member who had a leadership position in the Global PCOS Alliance. The resulting nine‐member committee had diverse reach and comprised one from each of the continents of Africa, Asia, and North America, and two from each of the European Union, the United Kingdom, and South America. The ODSC met at regular intervals between June and December 2020 to identify and engage stakeholders and develop and test the consensus process. The scope of the ODSC also included review and analysis of the results of the various rounds and the design and testing of subsequent Delphi rounds. Stakeholder and participant identification The first task of the ODSC was to identify and engage the appropriate stakeholders necessary for the Delphi process. The chosen categories included the following: National obstetrical and gynecological societies Subspecialty societies representing reproductive endocrinologists Specialty (obstetrics and gynecology) and subspecialty (reproductive endocrinology and infertility) journals Recognized experts in ovulatory disorders not participating in categories 1–3 Lay organizations interested in infertility, AUB, or PCOS Descriptive letters were created and customized for the various categories describing the rationale for the process and a synopsis of the methodology. Via the FIGO record of member countries, each of the national obstetrical and gynecological societies was contacted and invited by email to support the process by naming a representative. The ODSC identified the spectrum of subspecialty societies on the six continents and contacted leadership to explain the process and solicit support. The descriptive letter was sent electronically to both the society headquarters and the identified participant. A similar process involved the editorial offices of relevant specialty and subspecialty journals. The ODSC then identified recognized experts based on a combination of personal knowledge of the field and a search of the literature, subtracting those identified by national societies, subspecialty societies, or journals for representation. Finally, the ODSC sought to identify lay organizations that could represent women and adolescent girls who may have ovulatory disorders. These groups were generally contacted directly, and if there was interest and an indication of commitment, a lay‐based version of the letter was sent. The Delphi consensus process 3.3.1 | Background and scoring system The Delphi process was developed by the RAND Corporation as a method for determining multi‐stakeholder expert consensus in a semi‐anonymous fashion that minimizes the impact of interpersonal issues on the outcome. Originally designed to forecast the impact of technology on warfare, it has subsequently been utilized across a number of disciplines including health care. Versions of the Delphi Process were used previously in the development of the FIGO AUB systems , , and are generally similar to the original RAND system comprising a series of survey rounds designed to be administered in a web‐based or live environment with electronic scoring. Members of the ODSC did not participate in the Delphi process as participants. The scoring system has nine levels (1–9), with “1” being the most substantial disagreement with a statement, “9” the strongest agreement, and “5” representing neutrality. Scores in the top tertile (7, 8, and 9) indicated “agreement” with a statement, while those in the bottom tertile (1, 2, and 3) were indications of disagreement. As a result, the remaining scores (4, 5, and 6) comprised the “neutral” category, with “4” leaning to disagreement and “6” leaning to agreement. The minimum requirement for consensus agreement was a mean score of at least 7 (scores of 6.5–6.9 were rounded to 7), with no more than 15% in the disagreement category. Conversely, “disagreement” was defined as a mean score of 3 or less (scores of 3.1–3.4 were rounded to 3), with no more than 15% in the agreement category. For each statement or question in a survey, there is a field to allow for free‐text comments by the participants. 3.3.2 | Participant orientation meeting Before distributing the first round of surveys, two orientation meetings for the participants were held to ensure that the appropriate contact information was in the study database and systems and that all understood the survey mechanisms. The two meetings were held on the Zoom platform (Zoom Video Communications Inc, San Jose, CA, USA), with dates and times selected to facilitate flexibility for the diverse group of participants, particularly considering the spectrum of world time zones involved. Included in the messaging of this meeting was the understanding that Delphi participant answers would remain confidential and that all distributions would be anonymized. Demonstrations of the functionality of the system were provided. A session was recorded and uploaded to an accessible server for individuals who could not attend either of the live, web‐based meetings and to provide a resource for all participants who wished to review the instructions on their own time. It is to be noted that the lay component of the process was planned to occur after the medical stakeholders had developed a draft system. 3.3.3 | Conduct of the first round The first round of the Delphi process was designed to identify the participants' age, gender, location, expertise, and constituency and evaluate general opinions, the latter using statements intended to elicit an “agree” or “disagree” response. These statements were crafted in a fashion that invited and measured opinions regarding the clinical relevance of ovulatory disorders, the need for a well‐designed classification system, and the broad categories that should be included if such a system was to be designed. The draft set of questions was created by the Chair of the ODSC, reviewed by the committee members in meetings using the Zoom platform, and then tested on the web‐based survey instrument SurveyMonkey (Momentive, San Mateo, CA, USA). The final version of the first round was distributed to the stakeholders via their identified email addresses within the web‐based survey system. The ODSC Chair, who also functioned as the Facilitator, kept track of responses and sent out reminder emails at intervals of 7–10 days until there were no additional responses. The data were then exported to an Excel (Microsoft Corp, Everett, WA, USA) workbook comprising spreadsheets containing the survey template that automatically calculated means and the percentage of answers in the agree (7–9), neutral (4–6), and disagree (1–3) categories. The free‐text comments made by the participants were also included in the spreadsheet. The ODSC reviewed these data as a prelude to the design of the second round. The aggregate anonymized results were sent to each participant along with a copy of their responses for comparative purposes. 3.3.4 | Conduct of the second round The second‐round survey was constructed, in part, based upon the first‐round results. Some “neutral” responses that had marginal scores close to 3 or 7, or defined principally by the outliers, were reviewed in particular because, in such circumstances, it was possible that rewording a question or providing appropriately representative evidence would result in a change in the participant's opinion. It was also possible that “re‐asking” the question in the context of individual participant understanding of the group response might result in changes in individual responses. This information allowed the ODSC to construct a second survey round that eliminated items with defined agreement or disagreement but included reworded statements and new statements seeking to refine and expand the criteria that the participants thought necessary. The distribution of the second‐round survey was confined to those participating in and responding to the first round. The web‐based system, distribution, and follow‐up reminder technique were again employed. The data were retrieved, exported into the same Excel workbook with worksheet templates, and analyzed by the ODSC. Similarly, the participants received an anonymized summary of the participant responses to each of the items and a copy of their answers for comparison. At this point, the committee had enough information to design a draft system that addressed and included the elements identified in the first two Delphi rounds. This was conducted iteratively until a draft acceptable to all ODSC members was created. 3.3.5 | Conduct of the third round As a prelude to the live stakeholder meeting, a short clarifying third round was created, tested, distributed, and the results analyzed by the ODSC, conducted in a fashion similar to that of the first two rounds. Included in this round was a version of the draft system with solicitation of preliminary opinions from the participants. As was the case for the first two rounds, each participant was provided an anonymized copy of the results of the previous round and a copy of their responses, all for review before the live participant meeting. 3.3.6 | Participant meeting All medical participants and the ODSC were invited to participate in the stakeholder meeting held live on the Zoom platform. Here, the overall results of the survey rounds were presented, including those items where consensus one way or the other had not been reached. The draft system was also reviewed. An open discussion was invited, and preliminary polls were taken using the system available on the Zoom platform. 3.3.7 | Post‐meeting and fourth survey round The ODSC undertook the post‐meeting analysis. Subsequently, a short fourth‐round poll was conducted to reach a consensus on the remaining elements and include individuals who could not participate in the live meeting. 3.3.8 | Lay round The lay round was designed to query the lay representatives, both for their perception of a need for a classification system and their opinions of the system developed by the expert and representative participants. A separate survey was designed that included some of the items in the medical participant rounds but presented in a fashion accessible by a lay audience. There was a focus on their opinions of clarity and utility in the context of discussion and counseling involving healthcare practitioners and patients. The draft lay‐round elements were reviewed and revised by the ODSC, uploaded to the SurveyMonkey platform, tested, and then distributed to the participants in a fashion similar to that used for the medical participant rounds. The results were reviewed and analyzed by the ODSC, who considered these opinions in revising the system and constructing the manuscript and the design of materials for the lay audience. The Delphi process was developed by the RAND Corporation as a method for determining multi‐stakeholder expert consensus in a semi‐anonymous fashion that minimizes the impact of interpersonal issues on the outcome. Originally designed to forecast the impact of technology on warfare, it has subsequently been utilized across a number of disciplines including health care. Versions of the Delphi Process were used previously in the development of the FIGO AUB systems , , and are generally similar to the original RAND system comprising a series of survey rounds designed to be administered in a web‐based or live environment with electronic scoring. Members of the ODSC did not participate in the Delphi process as participants. The scoring system has nine levels (1–9), with “1” being the most substantial disagreement with a statement, “9” the strongest agreement, and “5” representing neutrality. Scores in the top tertile (7, 8, and 9) indicated “agreement” with a statement, while those in the bottom tertile (1, 2, and 3) were indications of disagreement. As a result, the remaining scores (4, 5, and 6) comprised the “neutral” category, with “4” leaning to disagreement and “6” leaning to agreement. The minimum requirement for consensus agreement was a mean score of at least 7 (scores of 6.5–6.9 were rounded to 7), with no more than 15% in the disagreement category. Conversely, “disagreement” was defined as a mean score of 3 or less (scores of 3.1–3.4 were rounded to 3), with no more than 15% in the agreement category. For each statement or question in a survey, there is a field to allow for free‐text comments by the participants. Before distributing the first round of surveys, two orientation meetings for the participants were held to ensure that the appropriate contact information was in the study database and systems and that all understood the survey mechanisms. The two meetings were held on the Zoom platform (Zoom Video Communications Inc, San Jose, CA, USA), with dates and times selected to facilitate flexibility for the diverse group of participants, particularly considering the spectrum of world time zones involved. Included in the messaging of this meeting was the understanding that Delphi participant answers would remain confidential and that all distributions would be anonymized. Demonstrations of the functionality of the system were provided. A session was recorded and uploaded to an accessible server for individuals who could not attend either of the live, web‐based meetings and to provide a resource for all participants who wished to review the instructions on their own time. It is to be noted that the lay component of the process was planned to occur after the medical stakeholders had developed a draft system. The first round of the Delphi process was designed to identify the participants' age, gender, location, expertise, and constituency and evaluate general opinions, the latter using statements intended to elicit an “agree” or “disagree” response. These statements were crafted in a fashion that invited and measured opinions regarding the clinical relevance of ovulatory disorders, the need for a well‐designed classification system, and the broad categories that should be included if such a system was to be designed. The draft set of questions was created by the Chair of the ODSC, reviewed by the committee members in meetings using the Zoom platform, and then tested on the web‐based survey instrument SurveyMonkey (Momentive, San Mateo, CA, USA). The final version of the first round was distributed to the stakeholders via their identified email addresses within the web‐based survey system. The ODSC Chair, who also functioned as the Facilitator, kept track of responses and sent out reminder emails at intervals of 7–10 days until there were no additional responses. The data were then exported to an Excel (Microsoft Corp, Everett, WA, USA) workbook comprising spreadsheets containing the survey template that automatically calculated means and the percentage of answers in the agree (7–9), neutral (4–6), and disagree (1–3) categories. The free‐text comments made by the participants were also included in the spreadsheet. The ODSC reviewed these data as a prelude to the design of the second round. The aggregate anonymized results were sent to each participant along with a copy of their responses for comparative purposes. The second‐round survey was constructed, in part, based upon the first‐round results. Some “neutral” responses that had marginal scores close to 3 or 7, or defined principally by the outliers, were reviewed in particular because, in such circumstances, it was possible that rewording a question or providing appropriately representative evidence would result in a change in the participant's opinion. It was also possible that “re‐asking” the question in the context of individual participant understanding of the group response might result in changes in individual responses. This information allowed the ODSC to construct a second survey round that eliminated items with defined agreement or disagreement but included reworded statements and new statements seeking to refine and expand the criteria that the participants thought necessary. The distribution of the second‐round survey was confined to those participating in and responding to the first round. The web‐based system, distribution, and follow‐up reminder technique were again employed. The data were retrieved, exported into the same Excel workbook with worksheet templates, and analyzed by the ODSC. Similarly, the participants received an anonymized summary of the participant responses to each of the items and a copy of their answers for comparison. At this point, the committee had enough information to design a draft system that addressed and included the elements identified in the first two Delphi rounds. This was conducted iteratively until a draft acceptable to all ODSC members was created. As a prelude to the live stakeholder meeting, a short clarifying third round was created, tested, distributed, and the results analyzed by the ODSC, conducted in a fashion similar to that of the first two rounds. Included in this round was a version of the draft system with solicitation of preliminary opinions from the participants. As was the case for the first two rounds, each participant was provided an anonymized copy of the results of the previous round and a copy of their responses, all for review before the live participant meeting. All medical participants and the ODSC were invited to participate in the stakeholder meeting held live on the Zoom platform. Here, the overall results of the survey rounds were presented, including those items where consensus one way or the other had not been reached. The draft system was also reviewed. An open discussion was invited, and preliminary polls were taken using the system available on the Zoom platform. The ODSC undertook the post‐meeting analysis. Subsequently, a short fourth‐round poll was conducted to reach a consensus on the remaining elements and include individuals who could not participate in the live meeting. The lay round was designed to query the lay representatives, both for their perception of a need for a classification system and their opinions of the system developed by the expert and representative participants. A separate survey was designed that included some of the items in the medical participant rounds but presented in a fashion accessible by a lay audience. There was a focus on their opinions of clarity and utility in the context of discussion and counseling involving healthcare practitioners and patients. The draft lay‐round elements were reviewed and revised by the ODSC, uploaded to the SurveyMonkey platform, tested, and then distributed to the participants in a fashion similar to that used for the medical participant rounds. The results were reviewed and analyzed by the ODSC, who considered these opinions in revising the system and constructing the manuscript and the design of materials for the lay audience. RESULTS 4.1 Medical expert participants A total of 88 invitations were sent to the responding national gynecological and obstetrical societies, experts at large, and the delegated representatives of journals and subspecialty societies. Ultimately, 46 individuals from all six continents responded and participated in the first Delphi round; approximately half were from Europe (Figure ), with age and gender distribution demonstrated in Figure . Of these, 28 (61%) were men and 18 (39%) were women. Over half of the participants (59%) were national society representatives, and 19% were experts at large (Figure ). Participants were asked about their principal role, and 72% responded “clinical care,” with the rest distributed across clinical research, teaching, and epidemiology. The secondary roles included clinical research, reported by 36%, and education by 24%, with some reporting bench research, administrative duties, and editorial responsibilities (Figure ). 4.2 Results of rounds 1–3 The results from rounds 1, 2, and 3 are shown in Tables , , and , respectively. In round 1, of 37 items, there was consensus on all but five. There was general support for the stated definition of ovulatory disorders and the rationale for a consensus classification system to support research, teaching, and clinical care. Respondents neither supported nor disagreed with the statement “The WHO classification system, in its current form, would meet the needs for a contemporary classification system for ovulatory disorders.” There was broad support for a spectrum of potential causes of ovulatory disorders except for idiopathic mechanisms and LOOP cycles. 9 The ODSC took these results and developed and tested the second Delphi round before distributing it to the 46 respondents in the first round. There were 41 respondents with the results of the 22 items shown in Table . The results of the second round suggested that there would be support for an anatomically based system (hypothalamus, pituitary, ovarian) with a separate category for PCOS. There was general support for this concept, with a mean score of 7.1. The survey also explored the notion of distinguishing chronic from isolated or intermittent ovulatory disorders, and this concept received consensus support with a mean score of 7.5 with no respondent disagreeing. Importantly, no consensus was reached on the question of using the Rotterdam Criteria to define PCOS, as 22.0% were in disagreement despite a mean overall score of 6.7. The second round was also designed to clarify some items from the first round and to identify more granular concepts relating to the pathogenesis of ovulatory disorders. There was a lack of consensus regarding the role of ovarian neoplasms, bacterial and viral infections, and the concept of infectious or inflammatory causes in general. There was also no consensus on the role of an absent surge of LH and LOOP events. While “menopause” as an etiology had a mean score otherwise sufficient to indicate agreement, 15% of the respondents disagreed, thereby preventing the attainment of consensus. With these data, the ODSC devised a draft system based upon anatomy that included a separate component for PCOS. Before distributing to the participants, and as a prelude to the live virtual meeting of the participants in the Delphi process, a five‐item third round was developed, tested, and distributed. Included in the distribution to the participants was evidence describing and evaluating LOOP events and the potential role of ovarian neoplasms and infectious or inflammatory disorders in the pathogenesis of ovulatory dysfunction. Related items were modified, and the results from the 38 respondents are displayed in Table . There was now consensus support for the inclusion of menopause and LOOP events, but lack of agreement on the role of ovarian neoplasms and infectious or other inflammatory disorders in the genesis of ovulatory dysfunction. 4.3 Live meeting For the live meeting, the ODSC distributed the draft system and an Excel workbook comprising a summary of the results of the three rounds and how the consensus agreements attained were integrated into the design. The live meeting was conducted on August 25, 2021, using the Zoom video platform. The meeting agenda included a review of the rationale for the process and the results of the three Delphi rounds, summarizing areas of agreement and focusing on the few places where consensus had not been reached. A total of 22 respondents could attend, so it was impossible to survey them officially. Still, there was a strong indication of support for the system based upon an in‐meeting electronic poll. The formal process was the subject of the fourth round. 4.4 Results of round 4 For this round, the ODSC sought the participants' opinions on the draft system and tried to resolve some of the remaining items upon which there was a persisting lack of consensus. For this four‐item survey, there were 39 respondents, with the results displayed in Table . There was support for the presented system by 95% of the respondents (mean score 8.0), with disagreement of only 2.6%. The fourth round also saw agreement that there should be a category for ovarian neoplasms. Although more than 60% supported the notion of inflammatory or infectious mechanisms, these items failed to achieve the predetermined criteria for consensus. There were some valuable comments about the specific graphical depiction of the system that will be discussed subsequently in the context of the results of the lay round. 4.5 Results of the lay round The lay round, as planned, was conducted following the deliberations of the experts and society, and journal representatives and the development of the draft FIGO Ovulatory Disorders Classification System. The results of the 11‐item survey sent to 17 individuals can be seen in Table . The first three items were designed to obtain demographic data; all 10 respondents were women representing organizations from Africa, Europe, and North America with an age distribution of 25–54 years. There was general agreement on the definition of ovulatory disorders and their potential role in the genesis of infertility. However, there was no consensus on the contribution of ovulatory disorders to symptoms of AUB. While there was agreement that girls and women often do not understand the causes of ovulatory disorders, there was uncertainty regarding reasons unknown to healthcare providers and other medical professionals. There was a clear consensus that a well‐conceived system of classifying ovulatory disorders would improve the design and interpretation of research and facilitate communication between patients and healthcare practitioners. However, the support for the draft system was mixed with a mean score of 4.9 and only 33% agreeing that the system was “understandable” and one that could provide “a platform upon which a lay audience” could “gain insight into the possible causes of ovulatory disorders.” The comments from the participants were illuminating (Table ) and, in some instances, mirrored comments from the other participants. Respecting these comments, the ODSC altered the graphical representation of the system without changing the content, placing the PCOS panel at the bottom, allowing for the use of the acronym “HyPO‐P.” In addition, a draft lay version of the major elements of the system was developed with lay language that was nonetheless compatible with the medical version (Supplementary Material). This draft was distributed to lay participants and their comments were generally incorporated into the text, and into modifications of the graphical content. Medical expert participants A total of 88 invitations were sent to the responding national gynecological and obstetrical societies, experts at large, and the delegated representatives of journals and subspecialty societies. Ultimately, 46 individuals from all six continents responded and participated in the first Delphi round; approximately half were from Europe (Figure ), with age and gender distribution demonstrated in Figure . Of these, 28 (61%) were men and 18 (39%) were women. Over half of the participants (59%) were national society representatives, and 19% were experts at large (Figure ). Participants were asked about their principal role, and 72% responded “clinical care,” with the rest distributed across clinical research, teaching, and epidemiology. The secondary roles included clinical research, reported by 36%, and education by 24%, with some reporting bench research, administrative duties, and editorial responsibilities (Figure ). Results of rounds 1–3 The results from rounds 1, 2, and 3 are shown in Tables , , and , respectively. In round 1, of 37 items, there was consensus on all but five. There was general support for the stated definition of ovulatory disorders and the rationale for a consensus classification system to support research, teaching, and clinical care. Respondents neither supported nor disagreed with the statement “The WHO classification system, in its current form, would meet the needs for a contemporary classification system for ovulatory disorders.” There was broad support for a spectrum of potential causes of ovulatory disorders except for idiopathic mechanisms and LOOP cycles. 9 The ODSC took these results and developed and tested the second Delphi round before distributing it to the 46 respondents in the first round. There were 41 respondents with the results of the 22 items shown in Table . The results of the second round suggested that there would be support for an anatomically based system (hypothalamus, pituitary, ovarian) with a separate category for PCOS. There was general support for this concept, with a mean score of 7.1. The survey also explored the notion of distinguishing chronic from isolated or intermittent ovulatory disorders, and this concept received consensus support with a mean score of 7.5 with no respondent disagreeing. Importantly, no consensus was reached on the question of using the Rotterdam Criteria to define PCOS, as 22.0% were in disagreement despite a mean overall score of 6.7. The second round was also designed to clarify some items from the first round and to identify more granular concepts relating to the pathogenesis of ovulatory disorders. There was a lack of consensus regarding the role of ovarian neoplasms, bacterial and viral infections, and the concept of infectious or inflammatory causes in general. There was also no consensus on the role of an absent surge of LH and LOOP events. While “menopause” as an etiology had a mean score otherwise sufficient to indicate agreement, 15% of the respondents disagreed, thereby preventing the attainment of consensus. With these data, the ODSC devised a draft system based upon anatomy that included a separate component for PCOS. Before distributing to the participants, and as a prelude to the live virtual meeting of the participants in the Delphi process, a five‐item third round was developed, tested, and distributed. Included in the distribution to the participants was evidence describing and evaluating LOOP events and the potential role of ovarian neoplasms and infectious or inflammatory disorders in the pathogenesis of ovulatory dysfunction. Related items were modified, and the results from the 38 respondents are displayed in Table . There was now consensus support for the inclusion of menopause and LOOP events, but lack of agreement on the role of ovarian neoplasms and infectious or other inflammatory disorders in the genesis of ovulatory dysfunction. Live meeting For the live meeting, the ODSC distributed the draft system and an Excel workbook comprising a summary of the results of the three rounds and how the consensus agreements attained were integrated into the design. The live meeting was conducted on August 25, 2021, using the Zoom video platform. The meeting agenda included a review of the rationale for the process and the results of the three Delphi rounds, summarizing areas of agreement and focusing on the few places where consensus had not been reached. A total of 22 respondents could attend, so it was impossible to survey them officially. Still, there was a strong indication of support for the system based upon an in‐meeting electronic poll. The formal process was the subject of the fourth round. Results of round 4 For this round, the ODSC sought the participants' opinions on the draft system and tried to resolve some of the remaining items upon which there was a persisting lack of consensus. For this four‐item survey, there were 39 respondents, with the results displayed in Table . There was support for the presented system by 95% of the respondents (mean score 8.0), with disagreement of only 2.6%. The fourth round also saw agreement that there should be a category for ovarian neoplasms. Although more than 60% supported the notion of inflammatory or infectious mechanisms, these items failed to achieve the predetermined criteria for consensus. There were some valuable comments about the specific graphical depiction of the system that will be discussed subsequently in the context of the results of the lay round. Results of the lay round The lay round, as planned, was conducted following the deliberations of the experts and society, and journal representatives and the development of the draft FIGO Ovulatory Disorders Classification System. The results of the 11‐item survey sent to 17 individuals can be seen in Table . The first three items were designed to obtain demographic data; all 10 respondents were women representing organizations from Africa, Europe, and North America with an age distribution of 25–54 years. There was general agreement on the definition of ovulatory disorders and their potential role in the genesis of infertility. However, there was no consensus on the contribution of ovulatory disorders to symptoms of AUB. While there was agreement that girls and women often do not understand the causes of ovulatory disorders, there was uncertainty regarding reasons unknown to healthcare providers and other medical professionals. There was a clear consensus that a well‐conceived system of classifying ovulatory disorders would improve the design and interpretation of research and facilitate communication between patients and healthcare practitioners. However, the support for the draft system was mixed with a mean score of 4.9 and only 33% agreeing that the system was “understandable” and one that could provide “a platform upon which a lay audience” could “gain insight into the possible causes of ovulatory disorders.” The comments from the participants were illuminating (Table ) and, in some instances, mirrored comments from the other participants. Respecting these comments, the ODSC altered the graphical representation of the system without changing the content, placing the PCOS panel at the bottom, allowing for the use of the acronym “HyPO‐P.” In addition, a draft lay version of the major elements of the system was developed with lay language that was nonetheless compatible with the medical version (Supplementary Material). This draft was distributed to lay participants and their comments were generally incorporated into the text, and into modifications of the graphical content. PROPOSED HyPO‐P SYSTEM 5.1 Rationale and development The system was designed to align with the results of the Delphi process (see Supplementary Table ). There was support for a design that grouped the causes of ovulatory disorders anatomically, a logical extension of the former WHO classification but more precise and more accessible than one based primarily on hormone assays. It was, therefore, rational to design this classification system according to the levels of the H‐P‐O axis as reflected in the second Delphi round (Table , question 1). It was also considered essential to allow for the designation of any element that is known or suspected to alter the functionality of the organ in a fashion that could contribute to the genesis of ovulatory dysfunction, whether related to demonstrable histopathology, abnormal laboratory assays, iatrogenic mechanisms, or even functional disorders without measurable laboratory features. However, it was recognized that an important cause of ovulatory disorders is PCOS since it affects 8%–13% of women of reproductive age. It is a complex and heterogeneous condition with comprehensive international guidelines for diagnosis, investigation, and management , , that cannot be confined to an ovarian origin. Therefore, it was determined that PCOS constitutes a class apart from the anatomical categorization, a notion that was supported in the second round of the Delphi process (Table , question 2). Therefore, the proposed FIGO classification now includes ovulatory disorders categorized into four groups as follows: Type I: Hypothalamic; Type II: Pituitary; Type III: Ovarian; and Type IV: PCOS (Figure ). The system can be referred to by the acronym “HyPO‐P,” where the “P” is separated from the other three categories recognizing that it does not reside in a single anatomic location. The new system provides practical utility and a second layer, or sub‐classification, for each of the three anatomically defined entities, including discrete pathophysiological categories. These can be remembered using the acronym “GAIN‐FIT‐PIE” (Figure ). A detailed description of every known or suspected cause of ovulatory dysfunction is beyond the scope of the present paper. Still, the new classification is presented with references to some of the many included conditions. Supplementary Table shows the linkages between various potential causes or categories of causes and the elements in the FIGO Ovulatory Disorders Classification System. Rationale and development The system was designed to align with the results of the Delphi process (see Supplementary Table ). There was support for a design that grouped the causes of ovulatory disorders anatomically, a logical extension of the former WHO classification but more precise and more accessible than one based primarily on hormone assays. It was, therefore, rational to design this classification system according to the levels of the H‐P‐O axis as reflected in the second Delphi round (Table , question 1). It was also considered essential to allow for the designation of any element that is known or suspected to alter the functionality of the organ in a fashion that could contribute to the genesis of ovulatory dysfunction, whether related to demonstrable histopathology, abnormal laboratory assays, iatrogenic mechanisms, or even functional disorders without measurable laboratory features. However, it was recognized that an important cause of ovulatory disorders is PCOS since it affects 8%–13% of women of reproductive age. It is a complex and heterogeneous condition with comprehensive international guidelines for diagnosis, investigation, and management , , that cannot be confined to an ovarian origin. Therefore, it was determined that PCOS constitutes a class apart from the anatomical categorization, a notion that was supported in the second round of the Delphi process (Table , question 2). Therefore, the proposed FIGO classification now includes ovulatory disorders categorized into four groups as follows: Type I: Hypothalamic; Type II: Pituitary; Type III: Ovarian; and Type IV: PCOS (Figure ). The system can be referred to by the acronym “HyPO‐P,” where the “P” is separated from the other three categories recognizing that it does not reside in a single anatomic location. The new system provides practical utility and a second layer, or sub‐classification, for each of the three anatomically defined entities, including discrete pathophysiological categories. These can be remembered using the acronym “GAIN‐FIT‐PIE” (Figure ). A detailed description of every known or suspected cause of ovulatory dysfunction is beyond the scope of the present paper. Still, the new classification is presented with references to some of the many included conditions. Supplementary Table shows the linkages between various potential causes or categories of causes and the elements in the FIGO Ovulatory Disorders Classification System. USE OF THE FIGO OVULATORY DISORDERS CLASSIFICATION SYSTEM 6.1 Clinical application 6.1.1 | Identifying individuals with ovulatory disorders The new system is designed for clinicians, educators, and investigators, including those involved in basic, translational, clinical, and epidemiological research. Depending on the audience, educators may focus only on the four primary categories or add the detail afforded by the second GAIN‐FIT‐PIE stratification. To be categorized by the system, the individual or patient must be identified as having an ovulatory disorder. Several potential clinical “entry points” are based on suspicion or knowledge about the presence of an ovulatory disorder that range from delayed menarche to infrequent or irregular menstruation through to presentation with primary or secondary infertility or hirsutism or other features or findings associated with PCOS. The term “ovulatory disorder” is not synonymous with the term “anovulation.” Instead, ovulatory disorders are considered to exist on a spectrum ranging from episodic to chronic (Figure ). Individuals may present with a chronic problem or may experience a singular episode where an anovulatory “cycle” manifests with delayed onset of HMB. Especially in the late reproductive years, women may experience regular, predictable cycles of normal length but experience HMB as the development of follicles in the luteal phase contribute to high premenstrual estradiol levels, a process known as a LOOP cycle. 9 Individuals with primary amenorrhea deserve special attention, and details regarding their investigation are beyond the scope of the present paper. However, in general, primary amenorrhea is said to be present when menstruation has not yet occurred by the age of 14 years in the absence of secondary sexual characteristics (when it is called delayed puberty) or 16 years in the presence of secondary sexual characteristics. Associated symptoms such as cyclical pelvic pain may suggest the presence of ovulation in association with a Müllerian anomaly or other obstruction that should be appropriately investigated without delay. Most, but certainly not all, ovulatory disorders are suggested by the presence of symptoms of AUB, ranging from complete absence (amenorrhea) to infrequent or irregular onset of menstrual blood flow. Secondary amenorrhea is generally defined as the cessation of menstruation for 6 months consecutively after at least one previous spontaneous menstrual bleed. Using data from extensive epidemiological studies, FIGO has previously determined that for those aged 18–45 years, and using the 5%–95% percentiles from large‐scale population studies, the normal frequency of menses is 24–38 days. Those with a cycle length of fewer than 24 days are deemed “frequent” while those whose cycle length is more than 38 days “infrequent,” a term designed to replace oligomenorrhea. , , , , Even in this category, regularity varies by age; for those aged either 18–25 or 42–45 years, the difference between the shortest and longest cycle should be 9 days or less, while for those aged 26–41 years, it is 7 days or less. Regardless, those with infrequent or irregular menstrual bleeding should be considered to have an ovulatory disorder. Diagnosing the presence of an ovulatory disorder at the extremes of reproductive age can be challenging, depending on the perception of what is normal. For postmenarcheal girls aged under 18 years, infrequent menstrual bleeding or irregular menstrual cycles suggesting ovulatory dysfunction are common, with available evidence suggesting that the individual's “normal” cycle length may not be established until the sixth year after menarche. , , During this pubertal transition, ovulatory dysfunction impacts about 50% of adolescent girls in the first year after menarche with a cycle length that is typically in the range of 21–45 days , but sometimes is as short as 20 days or may even exceed 60 days. In the years after menarche, these variations change such that 6 years later, the range is similar to those of adults. These issues can be explored in detail elsewhere. , However, it should be remembered that while common, and even “normal,” the individual's experience with this transition can be disruptive at a vulnerable time in their social, psychological, and physical development. A somewhat similar experience exists at the opposite end of the reproductive age spectrum, beyond the age of 45 years, as women enter what has been called the menopausal transition, where cycle length typically becomes more infrequent or irregular before culminating in amenorrhea as ovarian secretion of estradiol declines and ultimately ceases. However, this experience is perhaps even less orderly than that of the post‐menarcheal period, as there may be highly variable endocrine changes resulting in unpredictable impacts on menstrual function . Again, what is common, and often portrayed as “normal”, can be highly disruptive, particularly when coupled with other symptoms. Women who present with infertility may have accompanying menstrual symptoms typical of ovulatory disorders. However, women with cyclically normal onset of menstrual bleeding may not be ovulating, or at least not ovulating regularly, as the frequency of single‐cycle anovulation in the context of normal regular cycles is in the range of 3.7%–26.7%. , , Consequently, further evaluation beyond a detailed history will be necessary to identify those with ovulatory disorders. The optimal way to assess for ovulation and, by extension, confirm ovulatory disorders may vary according to the clinical circumstance. The menstrual history of regular, predictable cycles between 24 and 38 days remains a helpful tool, and reflects the overall experience better than evaluation of endocrine or imaging parameters from a single cycle does. While patients and clinicians have traditionally used measurement of basal body temperature, interpretation can be difficult, so this approach should be used with caution. , If available, ovulation predictor kits that measure the levels of luteinizing hormone in urine samples generally accurately reflect levels of serum luteinizing hormone and are a valuable tool for detecting ovulation in a given cycle. Simply measuring progesterone in the predicted luteal phase may provide satisfactory evidence supporting ovulatory function, particularly when the first day of the next menstrual period is known. Such an approach may be helpful in circumstances such as hirsutism, where the incidence of anovulation in women with cyclically predictable menstrual cycles is higher. There are other, less common ovulatory disorders that may require more complex evaluation to determine if they are present in a given individual. For example, identifying LUF cycles, somewhat common in infertile women, requires both confirmation of the LH surge and the performance of serial ultrasound to demonstrate failed rupture of the dominant follicle. It should be remembered that scrutiny of a single cycle may not reflect the overall experience for a given individual. 6.1.2 | Categorization in the FIGO Ovulatory Disorders Classification System The new system recognizes three basic strata once an ovulatory disorder has been diagnosed. The first level is categorization by one of the four primary categories as follows: Type I: Hypothalamus; Type II: Pituitary; Type III: Ovary; and Type IV: PCOS. The second level requires assignment to the known or suspected anatomically based abnormality as directed by the GAIN‐FIT‐PIE acronym. The third or tertiary level identifies a specific entity causing or contributing to the ovulatory disorder. Categorizing into these levels requires that the clinician perform whatever investigations deemed appropriate to localize the site and the presumed underlying mechanism contributing to ovulatory dysfunction. For example, the individual with infrequent and irregular menses, galactorrhea, elevated prolactin, and a magnetic resonance image demonstrating a pituitary tumor would categorize as a type 2 – N (pituitary neoplasm). The same might be said about an individual with irregular and infrequent menstruation, mild hirsutism, and sonographic evidence of at least one symmetrically enlarged ovary (≥10 ml) or an ovary with more than 20 follicles without a dominant follicle or corpus luteum, a circumstance that dictates a type 4 – PCOS classification. Use of the 20‐follicle threshold is utilized only when the patient is examined with an endovaginal ultrasound transducer with a high frequency bandwidth of at least 8 MHz. , It is recognized that the precision in determining the anatomic location and the mechanism of pathogenesis is somewhat aspirational and will vary to a degree by the disorder and the resources available to the clinician. Further discussion of the detection, characterization, and management of ovulatory disorders is beyond the spectrum of the present study, which is designed to provide a structure for clinical care, investigation, and education. Clinical application 6.1.1 | Identifying individuals with ovulatory disorders The new system is designed for clinicians, educators, and investigators, including those involved in basic, translational, clinical, and epidemiological research. Depending on the audience, educators may focus only on the four primary categories or add the detail afforded by the second GAIN‐FIT‐PIE stratification. To be categorized by the system, the individual or patient must be identified as having an ovulatory disorder. Several potential clinical “entry points” are based on suspicion or knowledge about the presence of an ovulatory disorder that range from delayed menarche to infrequent or irregular menstruation through to presentation with primary or secondary infertility or hirsutism or other features or findings associated with PCOS. The term “ovulatory disorder” is not synonymous with the term “anovulation.” Instead, ovulatory disorders are considered to exist on a spectrum ranging from episodic to chronic (Figure ). Individuals may present with a chronic problem or may experience a singular episode where an anovulatory “cycle” manifests with delayed onset of HMB. Especially in the late reproductive years, women may experience regular, predictable cycles of normal length but experience HMB as the development of follicles in the luteal phase contribute to high premenstrual estradiol levels, a process known as a LOOP cycle. 9 Individuals with primary amenorrhea deserve special attention, and details regarding their investigation are beyond the scope of the present paper. However, in general, primary amenorrhea is said to be present when menstruation has not yet occurred by the age of 14 years in the absence of secondary sexual characteristics (when it is called delayed puberty) or 16 years in the presence of secondary sexual characteristics. Associated symptoms such as cyclical pelvic pain may suggest the presence of ovulation in association with a Müllerian anomaly or other obstruction that should be appropriately investigated without delay. Most, but certainly not all, ovulatory disorders are suggested by the presence of symptoms of AUB, ranging from complete absence (amenorrhea) to infrequent or irregular onset of menstrual blood flow. Secondary amenorrhea is generally defined as the cessation of menstruation for 6 months consecutively after at least one previous spontaneous menstrual bleed. Using data from extensive epidemiological studies, FIGO has previously determined that for those aged 18–45 years, and using the 5%–95% percentiles from large‐scale population studies, the normal frequency of menses is 24–38 days. Those with a cycle length of fewer than 24 days are deemed “frequent” while those whose cycle length is more than 38 days “infrequent,” a term designed to replace oligomenorrhea. , , , , Even in this category, regularity varies by age; for those aged either 18–25 or 42–45 years, the difference between the shortest and longest cycle should be 9 days or less, while for those aged 26–41 years, it is 7 days or less. Regardless, those with infrequent or irregular menstrual bleeding should be considered to have an ovulatory disorder. Diagnosing the presence of an ovulatory disorder at the extremes of reproductive age can be challenging, depending on the perception of what is normal. For postmenarcheal girls aged under 18 years, infrequent menstrual bleeding or irregular menstrual cycles suggesting ovulatory dysfunction are common, with available evidence suggesting that the individual's “normal” cycle length may not be established until the sixth year after menarche. , , During this pubertal transition, ovulatory dysfunction impacts about 50% of adolescent girls in the first year after menarche with a cycle length that is typically in the range of 21–45 days , but sometimes is as short as 20 days or may even exceed 60 days. In the years after menarche, these variations change such that 6 years later, the range is similar to those of adults. These issues can be explored in detail elsewhere. , However, it should be remembered that while common, and even “normal,” the individual's experience with this transition can be disruptive at a vulnerable time in their social, psychological, and physical development. A somewhat similar experience exists at the opposite end of the reproductive age spectrum, beyond the age of 45 years, as women enter what has been called the menopausal transition, where cycle length typically becomes more infrequent or irregular before culminating in amenorrhea as ovarian secretion of estradiol declines and ultimately ceases. However, this experience is perhaps even less orderly than that of the post‐menarcheal period, as there may be highly variable endocrine changes resulting in unpredictable impacts on menstrual function . Again, what is common, and often portrayed as “normal”, can be highly disruptive, particularly when coupled with other symptoms. Women who present with infertility may have accompanying menstrual symptoms typical of ovulatory disorders. However, women with cyclically normal onset of menstrual bleeding may not be ovulating, or at least not ovulating regularly, as the frequency of single‐cycle anovulation in the context of normal regular cycles is in the range of 3.7%–26.7%. , , Consequently, further evaluation beyond a detailed history will be necessary to identify those with ovulatory disorders. The optimal way to assess for ovulation and, by extension, confirm ovulatory disorders may vary according to the clinical circumstance. The menstrual history of regular, predictable cycles between 24 and 38 days remains a helpful tool, and reflects the overall experience better than evaluation of endocrine or imaging parameters from a single cycle does. While patients and clinicians have traditionally used measurement of basal body temperature, interpretation can be difficult, so this approach should be used with caution. , If available, ovulation predictor kits that measure the levels of luteinizing hormone in urine samples generally accurately reflect levels of serum luteinizing hormone and are a valuable tool for detecting ovulation in a given cycle. Simply measuring progesterone in the predicted luteal phase may provide satisfactory evidence supporting ovulatory function, particularly when the first day of the next menstrual period is known. Such an approach may be helpful in circumstances such as hirsutism, where the incidence of anovulation in women with cyclically predictable menstrual cycles is higher. There are other, less common ovulatory disorders that may require more complex evaluation to determine if they are present in a given individual. For example, identifying LUF cycles, somewhat common in infertile women, requires both confirmation of the LH surge and the performance of serial ultrasound to demonstrate failed rupture of the dominant follicle. It should be remembered that scrutiny of a single cycle may not reflect the overall experience for a given individual. 6.1.2 | Categorization in the FIGO Ovulatory Disorders Classification System The new system recognizes three basic strata once an ovulatory disorder has been diagnosed. The first level is categorization by one of the four primary categories as follows: Type I: Hypothalamus; Type II: Pituitary; Type III: Ovary; and Type IV: PCOS. The second level requires assignment to the known or suspected anatomically based abnormality as directed by the GAIN‐FIT‐PIE acronym. The third or tertiary level identifies a specific entity causing or contributing to the ovulatory disorder. Categorizing into these levels requires that the clinician perform whatever investigations deemed appropriate to localize the site and the presumed underlying mechanism contributing to ovulatory dysfunction. For example, the individual with infrequent and irregular menses, galactorrhea, elevated prolactin, and a magnetic resonance image demonstrating a pituitary tumor would categorize as a type 2 – N (pituitary neoplasm). The same might be said about an individual with irregular and infrequent menstruation, mild hirsutism, and sonographic evidence of at least one symmetrically enlarged ovary (≥10 ml) or an ovary with more than 20 follicles without a dominant follicle or corpus luteum, a circumstance that dictates a type 4 – PCOS classification. Use of the 20‐follicle threshold is utilized only when the patient is examined with an endovaginal ultrasound transducer with a high frequency bandwidth of at least 8 MHz. , It is recognized that the precision in determining the anatomic location and the mechanism of pathogenesis is somewhat aspirational and will vary to a degree by the disorder and the resources available to the clinician. Further discussion of the detection, characterization, and management of ovulatory disorders is beyond the spectrum of the present study, which is designed to provide a structure for clinical care, investigation, and education. The new system is designed for clinicians, educators, and investigators, including those involved in basic, translational, clinical, and epidemiological research. Depending on the audience, educators may focus only on the four primary categories or add the detail afforded by the second GAIN‐FIT‐PIE stratification. To be categorized by the system, the individual or patient must be identified as having an ovulatory disorder. Several potential clinical “entry points” are based on suspicion or knowledge about the presence of an ovulatory disorder that range from delayed menarche to infrequent or irregular menstruation through to presentation with primary or secondary infertility or hirsutism or other features or findings associated with PCOS. The term “ovulatory disorder” is not synonymous with the term “anovulation.” Instead, ovulatory disorders are considered to exist on a spectrum ranging from episodic to chronic (Figure ). Individuals may present with a chronic problem or may experience a singular episode where an anovulatory “cycle” manifests with delayed onset of HMB. Especially in the late reproductive years, women may experience regular, predictable cycles of normal length but experience HMB as the development of follicles in the luteal phase contribute to high premenstrual estradiol levels, a process known as a LOOP cycle. 9 Individuals with primary amenorrhea deserve special attention, and details regarding their investigation are beyond the scope of the present paper. However, in general, primary amenorrhea is said to be present when menstruation has not yet occurred by the age of 14 years in the absence of secondary sexual characteristics (when it is called delayed puberty) or 16 years in the presence of secondary sexual characteristics. Associated symptoms such as cyclical pelvic pain may suggest the presence of ovulation in association with a Müllerian anomaly or other obstruction that should be appropriately investigated without delay. Most, but certainly not all, ovulatory disorders are suggested by the presence of symptoms of AUB, ranging from complete absence (amenorrhea) to infrequent or irregular onset of menstrual blood flow. Secondary amenorrhea is generally defined as the cessation of menstruation for 6 months consecutively after at least one previous spontaneous menstrual bleed. Using data from extensive epidemiological studies, FIGO has previously determined that for those aged 18–45 years, and using the 5%–95% percentiles from large‐scale population studies, the normal frequency of menses is 24–38 days. Those with a cycle length of fewer than 24 days are deemed “frequent” while those whose cycle length is more than 38 days “infrequent,” a term designed to replace oligomenorrhea. , , , , Even in this category, regularity varies by age; for those aged either 18–25 or 42–45 years, the difference between the shortest and longest cycle should be 9 days or less, while for those aged 26–41 years, it is 7 days or less. Regardless, those with infrequent or irregular menstrual bleeding should be considered to have an ovulatory disorder. Diagnosing the presence of an ovulatory disorder at the extremes of reproductive age can be challenging, depending on the perception of what is normal. For postmenarcheal girls aged under 18 years, infrequent menstrual bleeding or irregular menstrual cycles suggesting ovulatory dysfunction are common, with available evidence suggesting that the individual's “normal” cycle length may not be established until the sixth year after menarche. , , During this pubertal transition, ovulatory dysfunction impacts about 50% of adolescent girls in the first year after menarche with a cycle length that is typically in the range of 21–45 days , but sometimes is as short as 20 days or may even exceed 60 days. In the years after menarche, these variations change such that 6 years later, the range is similar to those of adults. These issues can be explored in detail elsewhere. , However, it should be remembered that while common, and even “normal,” the individual's experience with this transition can be disruptive at a vulnerable time in their social, psychological, and physical development. A somewhat similar experience exists at the opposite end of the reproductive age spectrum, beyond the age of 45 years, as women enter what has been called the menopausal transition, where cycle length typically becomes more infrequent or irregular before culminating in amenorrhea as ovarian secretion of estradiol declines and ultimately ceases. However, this experience is perhaps even less orderly than that of the post‐menarcheal period, as there may be highly variable endocrine changes resulting in unpredictable impacts on menstrual function . Again, what is common, and often portrayed as “normal”, can be highly disruptive, particularly when coupled with other symptoms. Women who present with infertility may have accompanying menstrual symptoms typical of ovulatory disorders. However, women with cyclically normal onset of menstrual bleeding may not be ovulating, or at least not ovulating regularly, as the frequency of single‐cycle anovulation in the context of normal regular cycles is in the range of 3.7%–26.7%. , , Consequently, further evaluation beyond a detailed history will be necessary to identify those with ovulatory disorders. The optimal way to assess for ovulation and, by extension, confirm ovulatory disorders may vary according to the clinical circumstance. The menstrual history of regular, predictable cycles between 24 and 38 days remains a helpful tool, and reflects the overall experience better than evaluation of endocrine or imaging parameters from a single cycle does. While patients and clinicians have traditionally used measurement of basal body temperature, interpretation can be difficult, so this approach should be used with caution. , If available, ovulation predictor kits that measure the levels of luteinizing hormone in urine samples generally accurately reflect levels of serum luteinizing hormone and are a valuable tool for detecting ovulation in a given cycle. Simply measuring progesterone in the predicted luteal phase may provide satisfactory evidence supporting ovulatory function, particularly when the first day of the next menstrual period is known. Such an approach may be helpful in circumstances such as hirsutism, where the incidence of anovulation in women with cyclically predictable menstrual cycles is higher. There are other, less common ovulatory disorders that may require more complex evaluation to determine if they are present in a given individual. For example, identifying LUF cycles, somewhat common in infertile women, requires both confirmation of the LH surge and the performance of serial ultrasound to demonstrate failed rupture of the dominant follicle. It should be remembered that scrutiny of a single cycle may not reflect the overall experience for a given individual. FIGO Ovulatory Disorders Classification System The new system recognizes three basic strata once an ovulatory disorder has been diagnosed. The first level is categorization by one of the four primary categories as follows: Type I: Hypothalamus; Type II: Pituitary; Type III: Ovary; and Type IV: PCOS. The second level requires assignment to the known or suspected anatomically based abnormality as directed by the GAIN‐FIT‐PIE acronym. The third or tertiary level identifies a specific entity causing or contributing to the ovulatory disorder. Categorizing into these levels requires that the clinician perform whatever investigations deemed appropriate to localize the site and the presumed underlying mechanism contributing to ovulatory dysfunction. For example, the individual with infrequent and irregular menses, galactorrhea, elevated prolactin, and a magnetic resonance image demonstrating a pituitary tumor would categorize as a type 2 – N (pituitary neoplasm). The same might be said about an individual with irregular and infrequent menstruation, mild hirsutism, and sonographic evidence of at least one symmetrically enlarged ovary (≥10 ml) or an ovary with more than 20 follicles without a dominant follicle or corpus luteum, a circumstance that dictates a type 4 – PCOS classification. Use of the 20‐follicle threshold is utilized only when the patient is examined with an endovaginal ultrasound transducer with a high frequency bandwidth of at least 8 MHz. , It is recognized that the precision in determining the anatomic location and the mechanism of pathogenesis is somewhat aspirational and will vary to a degree by the disorder and the resources available to the clinician. Further discussion of the detection, characterization, and management of ovulatory disorders is beyond the spectrum of the present study, which is designed to provide a structure for clinical care, investigation, and education. DISCUSSION AND CONCLUSION The FIGO HyPO‐P system for the classification of ovulatory disorders is submitted for consideration as a worldwide standard designed to harmonize definitions and categories in a fashion that should inform clinical care, facilitate the education of patients and trainees, and improve the ability of basic, translational, clinical, and epidemiologic research to advance our knowledge of ovulatory disorders, their diagnosis, and their management. The development has the general support of a broad spectrum of national and subspecialty societies, relevant journals, and recognized experts in the realm of ovulatory dysfunction. The lay participants agreed with the need for classification. Their comments helped refine the graphical representation and supported the rationale for a lay‐oriented explanation of ovulatory disorders presented in the context of the new system. Finally, no system should be considered permanent, so review and careful modification and revision should be carried out regularly. MGM: Chair of the Ovulatory Disorders Steering Committee (ODSC); responsible for the concept, design and management of the Delphi system; management of ODSC and stakeholder meetings, compiling and analysis of data, manuscript preparation. AHB: At large member of the ODSC; helped lead design and management of the Delphi process; analysis of data; responsible for converting results into the design of the system; manuscript preparation. SHC: Member of the ODSC; participated in the Delphi design and identification of stakeholders, and manuscript preparation. HODC: Member of the ODSC; participated in the Delphi design and identification of stakeholders, analysis of data, and manuscript preparation. ID: Co‐chair of the ODSC; participated in the Delphi design and identification of stakeholders, assisted with manuscript preparation. RF: Member of the ODSC; participated in the Delphi design and identification of stakeholders and assisted with manuscript preparation. LH: Member of the ODSC; participated in the Delphi design and identification of stakeholders, analysis of data, and manuscript preparation. EM: Member of the ODSC; participated in the Delphi design and identification of stakeholders, and manuscript preparation. ZVDS: Member of the ODSC; participated in the Delphi design and identification of stakeholders, analysis of data, and manuscript preparation. MGM reports grant funding from AbbVie and Pharmacosmos; consulting fees from Abbvie, Myovant, American Regent, Daiichi Sankyo, Hologic Inc and Pharmacosmos as well as royalty payments from UpToDate. He serves a voluntary role as Chair of the SEUD AUB Task Force, the Past Chair of FIGO's committee on Menstrual Disorders and Related Health Impacts, and Founding and Current Chair of the Women's Health Research Collaborative. AHB reports consulting fees from NovoNordisk and is a member of the WHO's Guideline Development on Infertility and a member of the International PCOS Guideline Group. He is a Trustee of the British Fertility Society and is a Director of Balance Reproductive Health Ltd and Balance Health Ltd. HODC is current Chair, FIGO Committee on Menstrual Disorders and Related Health Impacts. She has received clinical research support for laboratory consumables and staff from Bayer AG (paid to institution) and provides consultancy advice (all paid to institution) for Bayer AG, PregLem SA, Gedeon Richter, Vifor Pharma UK Ltd, AbbVie Inc; Myovant Sciences GmbH. HC has received royalties from UpToDate for articles on abnormal uterine bleeding. The rest of the authors have no conflicts of interest. None. Supplementary Table 1. Linking Delphi rounds to HyPO‐P components. Click here for additional data file. Appendix S1: Supporting information Click here for additional data file. |
Identifikation psychosozial belasteter Familien in pädiatrischen Praxen | 1eb6a7ed-ab23-4496-bdef-978ebad67997 | 11615100 | Pediatrics[mh] | In Deutschland lebt ca. ein Fünftel der Familien mit kleinen Kindern unter psychosozial belastenden Bedingungen , die sich schon bei Säuglingen und Kleinkindern negativ auf die Gesundheit und Entwicklung auswirken können . Früh ansetzende präventive Unterstützungsangebote können dazu beitragen, negative Folgen eines Aufwachsens in Belastungslagen abzumildern . Dies ist das Ziel der Frühen Hilfen . Die Frühen Hilfen wurden seit 2012 bundesweit flächendeckend ausgebaut und in kommunalen Netzwerken koordiniert . Sie umfassen Angebote, die allen Schwangeren und Eltern mit kleinen Kindern offenstehen. Mit speziellen Angeboten, wie beispielsweise der längerfristig aufsuchenden Begleitung durch Familienhebammen, richten sie sich aber insbesondere an Familien in Belastungslagen. Gerade Familien mit erhöhtem Unterstützungsbedarf sind mit Präventionsangeboten aber oft nur schwer zu erreichen . Um diesem sogenannten Präventionsdilemma entgegenzuwirken, wurde von Beginn an eine enge Zusammenarbeit zwischen den Frühen Hilfen und der Praxispädiatrie angestrebt . Praxispädiater*innen spielen hierbei eine wichtige Rolle, da die meisten Familien mit kleinen Kindern die kinderärztlichen Früherkennungsuntersuchungen nutzen und Praxispädiater*innen ein hohes Vertrauen entgegengebracht wird. Praxispädiater*innen können das Präventionsdilemma entschärfen, indem sie Familien mit psychosozialen Belastungen identifizieren und bei Bedarf eine Nutzung der Frühen Hilfen empfehlen und anbahnen. In einer im Jahr 2017 bundesweit durchgeführten repräsentativen Befragung von 815 Praxispädiater*innen zeigte sich allerdings, dass die Überleitung von psychosozial belasteten Familien nicht ausreichend gelingt: Praxispädiater*innen vermittelten demnach nur etwa jede 6. belastete Familie in kommunale Angebote der Frühen Hilfen . Um die Versorgung psychosozial belasteter Familien mit kleinen Kindern zu verbessern, entwickelte das Nationale Zentrum Frühe Hilfen (NZFH) in Zusammenarbeit mit der Kassenärztlichen Vereinigung Baden-Württemberg (KVBW) bereits im Jahr 2010 die sogenannte PATH-Intervention („Pediatric Attention To Help“). Diese wurde vom NZFH aus Mitteln des Bundesministeriums für Familie, Senioren, Frauen und Jugend (BMFSFJ) gefördert. Wie in der Infobox dargestellt, umfasst die PATH-Intervention eine spezielle Schulung der Praxispädiater*innen und die Teilnahme an interdisziplinären Qualitätszirkeln Frühe Hilfen (IQZ). Die PATH-Intervention ähnelt anderen in Deutschland erprobten Interventionen, die ebenfalls eine Schulung von Pädiater*innen beinhalten und damit auf ein frühzeitiges Erkennen familialer Belastungen und eine Vermittlung in psychosoziale Unterstützungsangebote abzielen . Eine Besonderheit der PATH-Intervention besteht jedoch darin, dass sie neben einer Schulung auch regelmäßig stattfindende IQZ umfasst, in denen sowohl Fälle belasteter Familien als auch Unterstützungsmöglichkeiten durch regionale Angebote besprochen werden. Dadurch soll nicht nur das handlungsrelevante Wissen gesteigert, sondern auch die Vernetzung zwischen den Teilnehmenden unterschiedlicher Sektoren, insbesondere des Gesundheitswesens und der Kinder- und Jugendhilfe, gestärkt werden. Dabei setzt sich die PATH-Intervention folgende Ziele: Praxispädiater*innen sollen psychosozial belastete Familien identifizieren, ihre Unterstützungsbedarfe mit ihnen besprechen, regionale Unterstützungsmöglichkeiten kennen, Eltern über deren Nutzen informieren und sie zur Inanspruchnahme motivieren. Dies soll im Ergebnis dazu beitragen, dass psychosozial belastete Familien vermehrt passende Angebote der Frühen Hilfen in Anspruch nehmen . Inwieweit diese Ziele mit der PATH-Intervention erreicht werden, wurde bislang noch nicht systematisch überprüft. Diese Lücke wurde mit der aus Mitteln des Innovationsfonds geförderten PATH-Evaluation geschlossen . Diese Evaluation prüft die Wirkung der PATH-Intervention auf die Vermittlung von psychosozial belasteten Familien in Angebote der Frühen Hilfen (primärer Endpunkt) sowie die Wirkung auf die (der Vermittlung vorgelagerten) Schritte der Identifikation psychosozial belasteter Familien und deren Informierung und Motivierung zur Inanspruchnahme von Angeboten Früher Hilfen (sekundäre Endpunkte). Zudem untersucht die Evaluation die Akzeptanz der PATH-Intervention bei allen Beteiligten, das Kosten-Effektivitäts-Verhältnis und die Treatment-Integrität. Im vorliegenden Beitrag wird die Wirkung der PATH-Intervention auf die Identifikation psychosozial belasteter Familien betrachtet. Dieser Aspekt ist von besonderer Relevanz, da das Erkennen einer psychosozialen Belastung nicht nur den ersten Schritt des Vermittlungsprozesses in Angebote der Frühen Hilfen darstellt, sondern darüber hinaus eine wichtige Voraussetzung für jedwedes Eingehen auf psychosoziale Belastungen durch Pädiater*innen ist. Als Hauptfragestellung des vorliegenden Beitrags prüfen wir deshalb, ob die PATH-Intervention die Identifikation psychosozial belasteter Familien durch Praxispädiater*innen verbessert. Unsere Hypothese lautet: Der Anteil identifizierter psychosozial belasteter Familien ist bei Praxispädiater*innen, die an der PATH-Intervention teilgenommen haben (Interventionsgruppe, IG), höher als bei Praxispädiater*innen, die nicht an der PATH-Intervention teilgenommen haben (Kontrollgruppe, KG). Als ergänzende Fragestellung untersuchen wir, ob die Stärke des Effekts der PATH-Intervention von dem Ausmaß der Belastung einer Familie abhängt. Dies ist anzunehmen, da es für Praxispädiater*innen umso einfacher sein sollte, eine Familie als belastet zu identifizieren, je mehr Risikofaktoren für Belastungen diese aufweist. Je mehr solcher Risikofaktoren eine Familie aufweist, desto weniger sollte es einen Unterschied machen, ob ein*e Praxispädiater*in spezielle Kenntnisse oder Kompetenzen für das Erkennen psychosozial belasteter Familien erworben hat. Unsere Hypothese lautet hier: Der Effekt der PATH-Intervention (d. h. der Unterschied zwischen IG und KG im Erkennen psychosozial belasteter Familien) nimmt mit steigender Anzahl von Risikofaktoren einer Familie ab. Zuletzt analysieren wir als explorative Fragestellung, ob die PATH-Intervention einen Einfluss darauf hat, wie Praxispädiater*innen die psychosoziale Belastung von Familien einschätzen, für die mithilfe des familienseitig eingesetzten Messinstruments kein Risikofaktor ermittelt wurde. Studiendesign und Rekrutierung In einer quasiexperimentellen Studie wurden Familien mit Kindern im Alter von bis zu 3 Jahren, die von Praxispädiater*innen betreut wurden, die an der PATH-Intervention teilgenommen hatten (IG), mit Familien von Praxispädiater*innen verglichen, die nicht an der PATH-Intervention teilgenommen hatten (KG). Für die IG rekrutierte das NZFH, unterstützt von der KVBW, Praxispädiater*innen mit Praxissitz in Baden-Württemberg. Einschlusskriterium waren die Teilnahme an der Interventionsschulung der KVBW oder an einer vergleichbaren Schulung zum Thema Frühe Hilfen bzw. Versorgung psychosozial belasteter Familien sowie die Teilnahme an mindestens 2 IQZ Frühe Hilfen in den letzten 2 Jahren. Für die KG wurden in Zusammenarbeit mit dem Berufsverband der Kinder- und Jugendärzte Bayerns Praxispädiater*innen aus Bayern rekrutiert. Um die Akzeptanz der PATH-Intervention beurteilen zu können , wurden mit 10 Praxispädiater*innen der IG und 20 psychosozial belasteten Familien der IG zusätzlich qualitative Interviews geführt. Im vorliegenden Beitrag werden Ergebnisse aus diesen Interviews dargestellt, die Aufschluss über die Identifikation psychosozial belasteter Familien durch Praxispädiater*innen der IG geben und dadurch zur Interpretation der quantitativen Ergebnisse beitragen. Die Rekrutierung der Familien durch die Praxispädiater*innen fand von Ende Juni 2021 bis Mitte April 2022 statt. Alle teilnehmenden Eltern hatten mit ihrem Kind eine U3–U7 Früherkennungsuntersuchung bei Praxispädiater*innen der IG oder KG besucht (Einschlusskriterium). Im Rahmen dieses Arztbesuchs informierten die Kinderarztpraxen die Familien über die Studie und holten ihre schriftliche Einwilligung ein. Von den Familien, die einer optionalen Teilnahme an einem zusätzlichen Interview zugestimmt hatten, wurden 20 Familien qualitativ befragt. Für diese Interviews kamen nur Familien infrage, für die mithilfe des familienseitig eingesetzten Messinstruments mindestens 2 Risikofaktoren festgestellt wurden. Darüber hinaus mussten die Familien auch arztseitig als belastet eingeschätzt oder es musste arztseitig ein Unterstützungsbedarf bei der Familie gesehen worden sein. Quantitative Befragungen Befragungen der Praxispädiater*innen Die teilnehmenden Praxispädiater*innen gaben nach jeder Früherkennungsuntersuchung mit einer teilnehmenden Familie in einem kurzen Papierfragebogen an, ob sie die Familie als psychosozial belastet einschätzen (Ja/Nein). Dieser Fragebogen war mit der Studien-ID der Familie versehen, wodurch eine Zuordnung zum familienseitigen Fragebogen ermöglicht wurde. Zudem füllten die Praxispädiater*innen nach Abschluss der Rekrutierungsphase einmalig einen Papierfragebogen aus, der Angaben zu ihrer Person (z. B. Alter, Geschlecht und Berufserfahrung) und zur Beschreibung ihrer kinderärztlichen Praxis (z. B. Einzel- oder Gemeinschaftspraxis) erhob. Befragungen der Familien Mit Eingang der Einwilligungserklärung beim NZFH erhielten die Eltern per E‑Mail einen individualisierten und mit ihrer Studien-ID versehenen Link zur Online-Befragung, die über das webbasierte Softwaretool REDCap umgesetzt wurde. Die meisten Eltern folgten der Einladung und bearbeiteten den Online-Fragebogen (IG = 86 %, KG = 81 %). Der Zeitraum zwischen der Früherkennungsuntersuchung und der Bearbeitung des Fragebogens durch die Familien betrug im Mittel 13 Tage (Standardabweichung (SD) = 11). Der Online-Fragebogen wurde in 4 Sprachen angeboten (Deutsch, Arabisch, Türkisch und Italienisch). Mit Ausnahme einer Familie füllten alle Familien die deutsche Version des Fragebogens aus. Die psychosoziale Belastung einer Familie wurde in Anlehnung an die NZFH-Studie KiD 0–3 mit dem Psychosozialen Belastungsindex (PSB-Index) erfasst. Dieser Index umfasste in der vorliegenden Studie 23 Indikatoren von Belastungen (z. B. alleinerziehend, beengte Wohnverhältnisse, belastendes Schreiverhalten des Kindes, Zweifel an der erzieherischen Kompetenz; für alle Indikatoren siehe Tabelle Z1 im Online-Material). Wenn die Ausprägungen dieser Indikatoren vorgegebene Grenzwerte überschreiten, werden sie als Risikofaktoren für psychosoziale Belastungen gewertet. Wenn beispielsweise der anhand des PHQ‑4 ermittelte Wert einer Familie ≥ 6 beträgt, stellt dies einen Risikofaktor („yellow flag“) für das Vorliegen einer Depression oder Angststörung und somit einer psychosozialen Belastung dar. Der Gesamtwert des PSB-Index, der PSB-Score, ergibt sich aus der Anzahl der Risikofaktoren einer Familie. Der Anteil fehlender Werte auf Ebene der einzelnen Risikofaktoren und auf Ebene des Gesamtwerts durfte maximal 30 % betragen . Weiterführende Informationen zu den 23 erfassten Risikofaktoren (z. B. deren Items, Berechnung und Grenzwerte) sind im Studienprotokoll ersichtlich. In Bezug auf den PSB-Score ist zu berücksichtigen, dass ein Wert von 0 nicht belegt, dass eine Familie unbelastet ist, da die Familie Risikofaktoren aufweisen kann, die im PSB-Index nicht erfasst werden. Der PSB-Score stellt somit die Untergrenze der tatsächlich vorliegenden Risikofaktoren einer Familie dar. In Einklang mit vorausgegangenen Arbeiten wurden in der vorliegenden Studie Familien, die laut Selbstauskunft mindestens einen Risikofaktor (d. h. PSB-Score ≥ 1) aufweisen, als mindestens geringfügig psychosozial belastet eingestuft. Wurden diese Familien von ihrer*ihrem Praxispädiater*in ebenfalls als psychosozial belastet eingeschätzt, zählten sie als identifiziert. Qualitative Interviews Für den qualitativen Studienstrang wurden problemzentrierte, teilstrukturierte Telefoninterviews durchgeführt. Hierfür wurde für die Praxispädiater*innen und die Familien jeweils ein spezifischer Leitfaden entwickelt. Die Interviews dauerten durchschnittlich 51 min (Praxispädiater*innen) bzw. 25 min (Familien). Die Telefoninterviews wurden zwischen dem 17.03.2022 und dem 16.06.2022 von einem extern beauftragten unabhängigen Forschungsinstitut durchgeführt. Analysen Quantitative Fragebogendaten Die statistischen Auswertungen wurden mit Microsoft Excel (Version 16) und IBM SPSS (Version 29.0.0.0) durchgeführt. Vorab wurde der Anteil fehlender Werte geprüft. Hierzu wurde ermittelt, wie viele Familien den Fragebogen bearbeitet, aber nicht genug Items des PSB-Index beantwortet hatten, um einen Gesamtwert zu ermitteln. Dies traf in der IG und der KG jeweils nur auf einen vergleichbar kleinen Anteil der Familien zu (IG = 4 %, KG = 5 %, Abb. ). Hinsichtlich der ärztlichen Einschätzungen der psychosozialen Belastung der Familien lagen ebenfalls nur sehr wenige fehlende Werte vor (IG = 0 %, KG = 2 %, Abb. ). Aufgrund der sehr kleinen Anteile fehlender Werte ist von keiner Verzerrung der Hypothesenprüfung durch fehlende Werte auszugehen. Da jeweils mehrere Familien von den gleichen Praxispädiater*innen behandelt wurden, wurden für die Analysen Mehrebenenmodelle angewendet, in denen ein zufälliger Effekt („random intercept“) in das Modell einbezogen wurde. Dieser bildete ab, von welchem*welcher Praxispädiater*in eine Familie behandelt wurde. In allen Analysen wurde jeweils der Endpunkt untersucht, ob eine Familie von ihrer*ihrem Praxispädiater*in als psychosozial belastet eingeschätzt wurde. Die Analysen zur Hauptfragestellung und zur ergänzenden Fragestellung wurden mit der Stichprobe der Familien durchgeführt, die gemäß dem familienseitigen Fragebogen als belastet eingestuft worden waren. Wurden diese Familien auch von ihren Praxispädiater*innen als belastet eingeschätzt, bedeutete dies, dass sie identifiziert wurden. Die Hypothesenprüfungen erfolgten anhand logistischer Regressionen, bei denen eine Propensity-Score-Adjustierung vorgenommen wurde, um Einflüsse konfundierender Merkmale statistisch zu kontrollieren. Der Propensity-Score wurde aus 6 arztseitigen Merkmalen (Tab. ), 5 familienseitigen Merkmalen (Tab. ) und der Zeit zwischen der Früherkennungsuntersuchung und der Beantwortung des Online-Fragebogens (in Tagen) gebildet. Die Hauptanalyse umfasste somit die Prädiktoren Gruppe (IG vs. KG) und Propensity-Score. Zur Beurteilung der Robustheit der Ergebnisse der Hauptanalyse wurde zudem eine Sensitivitätsanalyse ohne Propensity-Score-Adjustierung durchgeführt. Zur Prüfung der Hypothese der ergänzenden Fragestellung, in der postuliert wurde, dass die Stärke des Interventionseffekts vom Ausmaß der Belastung einer Familie abhängt, wurden der PSB-Score und der Interaktionsterm Gruppe x PSB-Score als Prädiktoren ergänzt . Die explorative Fragestellung konnte nicht wie geplant ebenfalls mittels einer logistischen Regression analysiert werden, da in der Stichprobe der Familien mit PSB = 0 in der KG keine Familie von den Ärzt*innen als belastet eingeschätzt wurde und das geplante Verfahren bei einer Häufigkeit von 0 nicht angewendet werden kann. Um diese Fragestellung dennoch zu prüfen, wurde Fishers exakter Test angewendet. Qualitative Interviewdaten Die anhand von Audioaufnahmen transkribierten Interviews wurden mit der Methode der inhaltlich strukturierenden qualitativen Inhaltsanalyse ausgewertet . Die Datenanalyse folgte dem systematischen Vorgehen, wie es von Kuckartz vorgeschlagen wird. Die Kategorienbildung erfolgte deduktiv-induktiv. Das anhand der Kategoriensysteme vollständig thematisch strukturierte Material wurde kategorienbasiert, themenspezifisch ausgewertet. Die Ausarbeitung der Kategoriensysteme, die Codierung und die Analyse der Daten erfolgten mithilfe der Software MAXQDA 2022 (Release 22.1.1; ). Für die Interrater-Reliabilität ergab sich sowohl für die Interviews mit den Ärzt*innen als auch für die Interviews mit den Familien eine Übereinstimmung zwischen Erst- und Zweitkodiererin von mehr als 70 % (K ≥ 0,72), was als zufriedenstellend bis gut bewertet werden kann . Registrierung Vor Beginn der Erhebungsphase wurde die Studie im Deutschen Register Klinischer Studien registriert (DRKS00023461, 03.12.2020). Die Universal Trial Number lautet: U1111-260-6575. Das methodische Vorgehen wurde vor Abschluss der Erhebungsphase in einem Studienprotokoll publiziert. In einer quasiexperimentellen Studie wurden Familien mit Kindern im Alter von bis zu 3 Jahren, die von Praxispädiater*innen betreut wurden, die an der PATH-Intervention teilgenommen hatten (IG), mit Familien von Praxispädiater*innen verglichen, die nicht an der PATH-Intervention teilgenommen hatten (KG). Für die IG rekrutierte das NZFH, unterstützt von der KVBW, Praxispädiater*innen mit Praxissitz in Baden-Württemberg. Einschlusskriterium waren die Teilnahme an der Interventionsschulung der KVBW oder an einer vergleichbaren Schulung zum Thema Frühe Hilfen bzw. Versorgung psychosozial belasteter Familien sowie die Teilnahme an mindestens 2 IQZ Frühe Hilfen in den letzten 2 Jahren. Für die KG wurden in Zusammenarbeit mit dem Berufsverband der Kinder- und Jugendärzte Bayerns Praxispädiater*innen aus Bayern rekrutiert. Um die Akzeptanz der PATH-Intervention beurteilen zu können , wurden mit 10 Praxispädiater*innen der IG und 20 psychosozial belasteten Familien der IG zusätzlich qualitative Interviews geführt. Im vorliegenden Beitrag werden Ergebnisse aus diesen Interviews dargestellt, die Aufschluss über die Identifikation psychosozial belasteter Familien durch Praxispädiater*innen der IG geben und dadurch zur Interpretation der quantitativen Ergebnisse beitragen. Die Rekrutierung der Familien durch die Praxispädiater*innen fand von Ende Juni 2021 bis Mitte April 2022 statt. Alle teilnehmenden Eltern hatten mit ihrem Kind eine U3–U7 Früherkennungsuntersuchung bei Praxispädiater*innen der IG oder KG besucht (Einschlusskriterium). Im Rahmen dieses Arztbesuchs informierten die Kinderarztpraxen die Familien über die Studie und holten ihre schriftliche Einwilligung ein. Von den Familien, die einer optionalen Teilnahme an einem zusätzlichen Interview zugestimmt hatten, wurden 20 Familien qualitativ befragt. Für diese Interviews kamen nur Familien infrage, für die mithilfe des familienseitig eingesetzten Messinstruments mindestens 2 Risikofaktoren festgestellt wurden. Darüber hinaus mussten die Familien auch arztseitig als belastet eingeschätzt oder es musste arztseitig ein Unterstützungsbedarf bei der Familie gesehen worden sein. Befragungen der Praxispädiater*innen Die teilnehmenden Praxispädiater*innen gaben nach jeder Früherkennungsuntersuchung mit einer teilnehmenden Familie in einem kurzen Papierfragebogen an, ob sie die Familie als psychosozial belastet einschätzen (Ja/Nein). Dieser Fragebogen war mit der Studien-ID der Familie versehen, wodurch eine Zuordnung zum familienseitigen Fragebogen ermöglicht wurde. Zudem füllten die Praxispädiater*innen nach Abschluss der Rekrutierungsphase einmalig einen Papierfragebogen aus, der Angaben zu ihrer Person (z. B. Alter, Geschlecht und Berufserfahrung) und zur Beschreibung ihrer kinderärztlichen Praxis (z. B. Einzel- oder Gemeinschaftspraxis) erhob. Befragungen der Familien Mit Eingang der Einwilligungserklärung beim NZFH erhielten die Eltern per E‑Mail einen individualisierten und mit ihrer Studien-ID versehenen Link zur Online-Befragung, die über das webbasierte Softwaretool REDCap umgesetzt wurde. Die meisten Eltern folgten der Einladung und bearbeiteten den Online-Fragebogen (IG = 86 %, KG = 81 %). Der Zeitraum zwischen der Früherkennungsuntersuchung und der Bearbeitung des Fragebogens durch die Familien betrug im Mittel 13 Tage (Standardabweichung (SD) = 11). Der Online-Fragebogen wurde in 4 Sprachen angeboten (Deutsch, Arabisch, Türkisch und Italienisch). Mit Ausnahme einer Familie füllten alle Familien die deutsche Version des Fragebogens aus. Die psychosoziale Belastung einer Familie wurde in Anlehnung an die NZFH-Studie KiD 0–3 mit dem Psychosozialen Belastungsindex (PSB-Index) erfasst. Dieser Index umfasste in der vorliegenden Studie 23 Indikatoren von Belastungen (z. B. alleinerziehend, beengte Wohnverhältnisse, belastendes Schreiverhalten des Kindes, Zweifel an der erzieherischen Kompetenz; für alle Indikatoren siehe Tabelle Z1 im Online-Material). Wenn die Ausprägungen dieser Indikatoren vorgegebene Grenzwerte überschreiten, werden sie als Risikofaktoren für psychosoziale Belastungen gewertet. Wenn beispielsweise der anhand des PHQ‑4 ermittelte Wert einer Familie ≥ 6 beträgt, stellt dies einen Risikofaktor („yellow flag“) für das Vorliegen einer Depression oder Angststörung und somit einer psychosozialen Belastung dar. Der Gesamtwert des PSB-Index, der PSB-Score, ergibt sich aus der Anzahl der Risikofaktoren einer Familie. Der Anteil fehlender Werte auf Ebene der einzelnen Risikofaktoren und auf Ebene des Gesamtwerts durfte maximal 30 % betragen . Weiterführende Informationen zu den 23 erfassten Risikofaktoren (z. B. deren Items, Berechnung und Grenzwerte) sind im Studienprotokoll ersichtlich. In Bezug auf den PSB-Score ist zu berücksichtigen, dass ein Wert von 0 nicht belegt, dass eine Familie unbelastet ist, da die Familie Risikofaktoren aufweisen kann, die im PSB-Index nicht erfasst werden. Der PSB-Score stellt somit die Untergrenze der tatsächlich vorliegenden Risikofaktoren einer Familie dar. In Einklang mit vorausgegangenen Arbeiten wurden in der vorliegenden Studie Familien, die laut Selbstauskunft mindestens einen Risikofaktor (d. h. PSB-Score ≥ 1) aufweisen, als mindestens geringfügig psychosozial belastet eingestuft. Wurden diese Familien von ihrer*ihrem Praxispädiater*in ebenfalls als psychosozial belastet eingeschätzt, zählten sie als identifiziert. Die teilnehmenden Praxispädiater*innen gaben nach jeder Früherkennungsuntersuchung mit einer teilnehmenden Familie in einem kurzen Papierfragebogen an, ob sie die Familie als psychosozial belastet einschätzen (Ja/Nein). Dieser Fragebogen war mit der Studien-ID der Familie versehen, wodurch eine Zuordnung zum familienseitigen Fragebogen ermöglicht wurde. Zudem füllten die Praxispädiater*innen nach Abschluss der Rekrutierungsphase einmalig einen Papierfragebogen aus, der Angaben zu ihrer Person (z. B. Alter, Geschlecht und Berufserfahrung) und zur Beschreibung ihrer kinderärztlichen Praxis (z. B. Einzel- oder Gemeinschaftspraxis) erhob. Mit Eingang der Einwilligungserklärung beim NZFH erhielten die Eltern per E‑Mail einen individualisierten und mit ihrer Studien-ID versehenen Link zur Online-Befragung, die über das webbasierte Softwaretool REDCap umgesetzt wurde. Die meisten Eltern folgten der Einladung und bearbeiteten den Online-Fragebogen (IG = 86 %, KG = 81 %). Der Zeitraum zwischen der Früherkennungsuntersuchung und der Bearbeitung des Fragebogens durch die Familien betrug im Mittel 13 Tage (Standardabweichung (SD) = 11). Der Online-Fragebogen wurde in 4 Sprachen angeboten (Deutsch, Arabisch, Türkisch und Italienisch). Mit Ausnahme einer Familie füllten alle Familien die deutsche Version des Fragebogens aus. Die psychosoziale Belastung einer Familie wurde in Anlehnung an die NZFH-Studie KiD 0–3 mit dem Psychosozialen Belastungsindex (PSB-Index) erfasst. Dieser Index umfasste in der vorliegenden Studie 23 Indikatoren von Belastungen (z. B. alleinerziehend, beengte Wohnverhältnisse, belastendes Schreiverhalten des Kindes, Zweifel an der erzieherischen Kompetenz; für alle Indikatoren siehe Tabelle Z1 im Online-Material). Wenn die Ausprägungen dieser Indikatoren vorgegebene Grenzwerte überschreiten, werden sie als Risikofaktoren für psychosoziale Belastungen gewertet. Wenn beispielsweise der anhand des PHQ‑4 ermittelte Wert einer Familie ≥ 6 beträgt, stellt dies einen Risikofaktor („yellow flag“) für das Vorliegen einer Depression oder Angststörung und somit einer psychosozialen Belastung dar. Der Gesamtwert des PSB-Index, der PSB-Score, ergibt sich aus der Anzahl der Risikofaktoren einer Familie. Der Anteil fehlender Werte auf Ebene der einzelnen Risikofaktoren und auf Ebene des Gesamtwerts durfte maximal 30 % betragen . Weiterführende Informationen zu den 23 erfassten Risikofaktoren (z. B. deren Items, Berechnung und Grenzwerte) sind im Studienprotokoll ersichtlich. In Bezug auf den PSB-Score ist zu berücksichtigen, dass ein Wert von 0 nicht belegt, dass eine Familie unbelastet ist, da die Familie Risikofaktoren aufweisen kann, die im PSB-Index nicht erfasst werden. Der PSB-Score stellt somit die Untergrenze der tatsächlich vorliegenden Risikofaktoren einer Familie dar. In Einklang mit vorausgegangenen Arbeiten wurden in der vorliegenden Studie Familien, die laut Selbstauskunft mindestens einen Risikofaktor (d. h. PSB-Score ≥ 1) aufweisen, als mindestens geringfügig psychosozial belastet eingestuft. Wurden diese Familien von ihrer*ihrem Praxispädiater*in ebenfalls als psychosozial belastet eingeschätzt, zählten sie als identifiziert. Für den qualitativen Studienstrang wurden problemzentrierte, teilstrukturierte Telefoninterviews durchgeführt. Hierfür wurde für die Praxispädiater*innen und die Familien jeweils ein spezifischer Leitfaden entwickelt. Die Interviews dauerten durchschnittlich 51 min (Praxispädiater*innen) bzw. 25 min (Familien). Die Telefoninterviews wurden zwischen dem 17.03.2022 und dem 16.06.2022 von einem extern beauftragten unabhängigen Forschungsinstitut durchgeführt. Quantitative Fragebogendaten Die statistischen Auswertungen wurden mit Microsoft Excel (Version 16) und IBM SPSS (Version 29.0.0.0) durchgeführt. Vorab wurde der Anteil fehlender Werte geprüft. Hierzu wurde ermittelt, wie viele Familien den Fragebogen bearbeitet, aber nicht genug Items des PSB-Index beantwortet hatten, um einen Gesamtwert zu ermitteln. Dies traf in der IG und der KG jeweils nur auf einen vergleichbar kleinen Anteil der Familien zu (IG = 4 %, KG = 5 %, Abb. ). Hinsichtlich der ärztlichen Einschätzungen der psychosozialen Belastung der Familien lagen ebenfalls nur sehr wenige fehlende Werte vor (IG = 0 %, KG = 2 %, Abb. ). Aufgrund der sehr kleinen Anteile fehlender Werte ist von keiner Verzerrung der Hypothesenprüfung durch fehlende Werte auszugehen. Da jeweils mehrere Familien von den gleichen Praxispädiater*innen behandelt wurden, wurden für die Analysen Mehrebenenmodelle angewendet, in denen ein zufälliger Effekt („random intercept“) in das Modell einbezogen wurde. Dieser bildete ab, von welchem*welcher Praxispädiater*in eine Familie behandelt wurde. In allen Analysen wurde jeweils der Endpunkt untersucht, ob eine Familie von ihrer*ihrem Praxispädiater*in als psychosozial belastet eingeschätzt wurde. Die Analysen zur Hauptfragestellung und zur ergänzenden Fragestellung wurden mit der Stichprobe der Familien durchgeführt, die gemäß dem familienseitigen Fragebogen als belastet eingestuft worden waren. Wurden diese Familien auch von ihren Praxispädiater*innen als belastet eingeschätzt, bedeutete dies, dass sie identifiziert wurden. Die Hypothesenprüfungen erfolgten anhand logistischer Regressionen, bei denen eine Propensity-Score-Adjustierung vorgenommen wurde, um Einflüsse konfundierender Merkmale statistisch zu kontrollieren. Der Propensity-Score wurde aus 6 arztseitigen Merkmalen (Tab. ), 5 familienseitigen Merkmalen (Tab. ) und der Zeit zwischen der Früherkennungsuntersuchung und der Beantwortung des Online-Fragebogens (in Tagen) gebildet. Die Hauptanalyse umfasste somit die Prädiktoren Gruppe (IG vs. KG) und Propensity-Score. Zur Beurteilung der Robustheit der Ergebnisse der Hauptanalyse wurde zudem eine Sensitivitätsanalyse ohne Propensity-Score-Adjustierung durchgeführt. Zur Prüfung der Hypothese der ergänzenden Fragestellung, in der postuliert wurde, dass die Stärke des Interventionseffekts vom Ausmaß der Belastung einer Familie abhängt, wurden der PSB-Score und der Interaktionsterm Gruppe x PSB-Score als Prädiktoren ergänzt . Die explorative Fragestellung konnte nicht wie geplant ebenfalls mittels einer logistischen Regression analysiert werden, da in der Stichprobe der Familien mit PSB = 0 in der KG keine Familie von den Ärzt*innen als belastet eingeschätzt wurde und das geplante Verfahren bei einer Häufigkeit von 0 nicht angewendet werden kann. Um diese Fragestellung dennoch zu prüfen, wurde Fishers exakter Test angewendet. Qualitative Interviewdaten Die anhand von Audioaufnahmen transkribierten Interviews wurden mit der Methode der inhaltlich strukturierenden qualitativen Inhaltsanalyse ausgewertet . Die Datenanalyse folgte dem systematischen Vorgehen, wie es von Kuckartz vorgeschlagen wird. Die Kategorienbildung erfolgte deduktiv-induktiv. Das anhand der Kategoriensysteme vollständig thematisch strukturierte Material wurde kategorienbasiert, themenspezifisch ausgewertet. Die Ausarbeitung der Kategoriensysteme, die Codierung und die Analyse der Daten erfolgten mithilfe der Software MAXQDA 2022 (Release 22.1.1; ). Für die Interrater-Reliabilität ergab sich sowohl für die Interviews mit den Ärzt*innen als auch für die Interviews mit den Familien eine Übereinstimmung zwischen Erst- und Zweitkodiererin von mehr als 70 % (K ≥ 0,72), was als zufriedenstellend bis gut bewertet werden kann . Die statistischen Auswertungen wurden mit Microsoft Excel (Version 16) und IBM SPSS (Version 29.0.0.0) durchgeführt. Vorab wurde der Anteil fehlender Werte geprüft. Hierzu wurde ermittelt, wie viele Familien den Fragebogen bearbeitet, aber nicht genug Items des PSB-Index beantwortet hatten, um einen Gesamtwert zu ermitteln. Dies traf in der IG und der KG jeweils nur auf einen vergleichbar kleinen Anteil der Familien zu (IG = 4 %, KG = 5 %, Abb. ). Hinsichtlich der ärztlichen Einschätzungen der psychosozialen Belastung der Familien lagen ebenfalls nur sehr wenige fehlende Werte vor (IG = 0 %, KG = 2 %, Abb. ). Aufgrund der sehr kleinen Anteile fehlender Werte ist von keiner Verzerrung der Hypothesenprüfung durch fehlende Werte auszugehen. Da jeweils mehrere Familien von den gleichen Praxispädiater*innen behandelt wurden, wurden für die Analysen Mehrebenenmodelle angewendet, in denen ein zufälliger Effekt („random intercept“) in das Modell einbezogen wurde. Dieser bildete ab, von welchem*welcher Praxispädiater*in eine Familie behandelt wurde. In allen Analysen wurde jeweils der Endpunkt untersucht, ob eine Familie von ihrer*ihrem Praxispädiater*in als psychosozial belastet eingeschätzt wurde. Die Analysen zur Hauptfragestellung und zur ergänzenden Fragestellung wurden mit der Stichprobe der Familien durchgeführt, die gemäß dem familienseitigen Fragebogen als belastet eingestuft worden waren. Wurden diese Familien auch von ihren Praxispädiater*innen als belastet eingeschätzt, bedeutete dies, dass sie identifiziert wurden. Die Hypothesenprüfungen erfolgten anhand logistischer Regressionen, bei denen eine Propensity-Score-Adjustierung vorgenommen wurde, um Einflüsse konfundierender Merkmale statistisch zu kontrollieren. Der Propensity-Score wurde aus 6 arztseitigen Merkmalen (Tab. ), 5 familienseitigen Merkmalen (Tab. ) und der Zeit zwischen der Früherkennungsuntersuchung und der Beantwortung des Online-Fragebogens (in Tagen) gebildet. Die Hauptanalyse umfasste somit die Prädiktoren Gruppe (IG vs. KG) und Propensity-Score. Zur Beurteilung der Robustheit der Ergebnisse der Hauptanalyse wurde zudem eine Sensitivitätsanalyse ohne Propensity-Score-Adjustierung durchgeführt. Zur Prüfung der Hypothese der ergänzenden Fragestellung, in der postuliert wurde, dass die Stärke des Interventionseffekts vom Ausmaß der Belastung einer Familie abhängt, wurden der PSB-Score und der Interaktionsterm Gruppe x PSB-Score als Prädiktoren ergänzt . Die explorative Fragestellung konnte nicht wie geplant ebenfalls mittels einer logistischen Regression analysiert werden, da in der Stichprobe der Familien mit PSB = 0 in der KG keine Familie von den Ärzt*innen als belastet eingeschätzt wurde und das geplante Verfahren bei einer Häufigkeit von 0 nicht angewendet werden kann. Um diese Fragestellung dennoch zu prüfen, wurde Fishers exakter Test angewendet. Die anhand von Audioaufnahmen transkribierten Interviews wurden mit der Methode der inhaltlich strukturierenden qualitativen Inhaltsanalyse ausgewertet . Die Datenanalyse folgte dem systematischen Vorgehen, wie es von Kuckartz vorgeschlagen wird. Die Kategorienbildung erfolgte deduktiv-induktiv. Das anhand der Kategoriensysteme vollständig thematisch strukturierte Material wurde kategorienbasiert, themenspezifisch ausgewertet. Die Ausarbeitung der Kategoriensysteme, die Codierung und die Analyse der Daten erfolgten mithilfe der Software MAXQDA 2022 (Release 22.1.1; ). Für die Interrater-Reliabilität ergab sich sowohl für die Interviews mit den Ärzt*innen als auch für die Interviews mit den Familien eine Übereinstimmung zwischen Erst- und Zweitkodiererin von mehr als 70 % (K ≥ 0,72), was als zufriedenstellend bis gut bewertet werden kann . Vor Beginn der Erhebungsphase wurde die Studie im Deutschen Register Klinischer Studien registriert (DRKS00023461, 03.12.2020). Die Universal Trial Number lautet: U1111-260-6575. Das methodische Vorgehen wurde vor Abschluss der Erhebungsphase in einem Studienprotokoll publiziert. Stichprobe Insgesamt nahmen 29 Praxispädiater*innen aus Baden-Württemberg (IG: n = 15) und Bayern (KG: n = 14) an der Studie teil. Wie in Tab. zu sehen ist, wiesen die Pädiater*innen der IG und der KG im Mittel sehr ähnliche soziodemografische und berufsbezogene Merkmale auf. Nur in Bezug auf die Anzahl der Früherkennungsuntersuchungen pro Monat gaben die Praxispädiater*innen der IG im Mittel eine niedrigere Anzahl an als die Ärzt*innen der KG. Insgesamt rekrutierten die teilnehmenden Praxispädiater*innen 549 Familien (IG: n = 251, KG: n = 298). Abb. zeigt den Fluss der Studienteilnehmenden bis zur Stichprobe der psychosozial belasteten Familien, die den Hypothesenprüfungen zugrunde lag. Diese Stichprobe umfasste 293 Familien, für die auf Basis ihrer Selbstauskunft mindestens ein Risikofaktor festgestellt wurde. Tab. zeigt die soziodemografischen Merkmale dieser Familien. Das Alter und das Geschlecht der Kinder, das Alter des antwortenden Elternteils und der Anteil der Familien mit Migrationsstatus fielen in IG und KG ähnlich aus. Allerdings wiesen in der IG mehr Familien einen höheren Schul- und Berufsabschluss und ein höheres Nettohaushaltseinkommen auf als in der KG. Diese Ähnlichkeiten und Unterschiede zwischen der IG und der KG zeigten sich auch in der Stichprobe der Familien, für die kein Risikofaktor ermittelt wurde (siehe Tabelle Z2 im Online-Material). Von den 10 telefonisch interviewten Praxispädiater*innen waren 5 (50 %) weiblich und 5 (50 %) männlich. Im Mittel arbeiten die Befragten seit 14 Jahren in einer niedergelassenen Praxis. Die 20 Interviews mit Familien wurden alle mit den Müttern der Kinder geführt, mit denen die Früherkennungsuntersuchung aufgesucht wurde (d. h. dem „Zielkind“). Die befragten Mütter waren im Mittel 35 Jahre alt und hatten 2 Kinder. Die Kinder, auf die sich die Mütter in den Interviews bezogen, waren im Durchschnitt etwa 1,5 Jahre alt. Hauptanalyse Zur Prüfung der Hypothese zur Hauptfragestellung wurde zunächst deskriptiv ermittelt, wie viele der gemäß familienseitigem Fragebogen psychosozial belasteten Familien auch von den Praxispädiater*innen der IG und KG als belastet eingeschätzt und somit identifiziert wurden. Wie in Abb. a zu sehen ist, traf dies auf 42 % (60 von 144) der Familien der IG zu und auf 23 % (34 von 149) der Familien der KG. Der deskriptive Gruppenunterschied betrug somit 19 Prozentpunkte. Dieser erwartungskonforme Unterschied zwischen der IG und der KG erwies sich in der inferenzstatistischen Prüfung als signifikant (Odds Ratio (OR) = 2,77; p = 0,020) und wird durch das Ergebnis der Sensitivitätsanalyse bekräftigt (OR = 2,76; p = 0,010). Tab. zeigt die Ergebnisse beider Analysen im Detail. Neben der arztseitigen Belastungseinschätzung geben die Ergebnisse der Telefoninterviews zudem Hinweise darauf, dass die Praxispädiater*innen bei der Abschätzung des Unterstützungsbedarfs einer Familie auch die familienseitigen Ressourcen und Lösungsstrategien in den Blick nehmen. „Also wichtig ist ja immer herauszufinden, wenn die Familie belastet ist, aber sie hat Strategien an der Hand, wie sie mit der Belastung umgehen, dann mache ich mir keine Sorgen“ [#P3]. Ergänzende Analyse Zur Prüfung der Hypothese zur ergänzenden Fragestellung wurde ebenfalls zunächst deskriptiv ermittelt, wie viele der psychosozial belasteten Familien in der IG und der KG je nach Anzahl der Risikofaktoren von den Praxispädiater*innen identifiziert wurden (Abb. b). Da nur sehr wenige Familien 5 oder mehr Risikofaktoren aufwiesen, wurden diese Familien zu einer Kategorie (≥ 5) zusammengefasst. Wie in Tab. (rechte Spalte) zu sehen ist, zeigte sich für den Prädiktor Gruppe x PSB ein Odds Ratio < 1, das nicht signifikant war ( p = 0,643), d. h., die Stärke des Interventionseffekts hing nicht signifikant von der Anzahl der Risikofaktoren einer Familie ab. Dieses Ergebnis ist auch in den in Abb. b dargestellten deskriptiven Häufigkeiten erkennbar: Der Gruppenunterschied bei Familien mit 1–4 Risikofaktoren betrug im Mittel 20 Prozentpunkte und fiel unabhängig von der Anzahl der Risikofaktoren ähnlich groß aus. Erst bei Familien mit 5 oder mehr Risikofaktoren fiel der Gruppenunterschied mit 11 Prozentpunkten deutlich kleiner aus. In Einklang mit diesen quantitativen Befunden liefern auch die qualitativen Ergebnisse Hinweise darauf, dass sich die PATH-Intervention unabhängig von der Stärke bzw. der Anzahl und Art der Belastungen der Familien positiv auf das Erkennen von Belastungen auswirken kann. Aus den Interviewergebnissen mit Praxispädiater*innen der IG geht hervor, dass die Teilnahme an den IQZ die Ärzt*innen für psychosoziale Belastungen von Familien zu sensibilisieren scheint. Der Austausch über Fälle (Fallbesprechungen), ein zentrales Element der IQZ, wird als hilfreich für das eigene ärztliche Handeln beschrieben. Dabei ist es insbesondere das Kennenlernen der verschiedenen Blickwinkel und Erfahrungen aus Gesundheitswesen und Kinder- und Jugendhilfe, das als wertvoll wahrgenommen wird. „… das Erste ist natürlich, dass man aus den Erfahrungen von anderen lernt und dass man beispielhaft dann auch für seine eigenen Patienten an den Beispielen lernt, dass man sensibilisiert wird finde ich ganz wichtig, dass man sensibilisiert zu bestimmten Themen …“ [#P3]. Explorative Analyse Zuletzt wurde explorativ überprüft, ob die PATH-Intervention einen Einfluss auf die ärztliche Einschätzung zur psychosozialen Belastung einer Familie hat, für die auf Basis des familienseitigen Fragebogens kein Risikofaktor festgestellt wurde (PSB-Score = 0). Diesbezüglich zeigte sich deskriptiv, dass in der IG 18 % (11 von 60) dieser Familien von ihrer*ihrem Praxispädiater*in als belastet eingeschätzt wurden, wohingegen dies in der KG für keine Familie (0 von 71) zutraf. Im exakten Test nach Fisher (2-seitig) war dieser Unterschied zwischen IG und KG signifikant ( p < 0,001). Eine Ursache dafür, dass in der IG ein höherer Anteil von Familien mit einem PSB-Score = 0 als belastet identifiziert wurde, könnte darin liegen, dass Ärzt*innen der IG weitere Arten von Belastungen wahrnehmen, die vom PSB-Index bislang nicht erfasst werden. Hierfür bieten sowohl die Interviews mit den Praxispädiater*innen der IG als auch die Interviews mit den Familien Anhaltspunkte. So werden Belastungen, die sich aufgrund der Betreuung mehrerer Kinder ergeben können, nicht (wie im PSB-Index) an einem sehr jungen Alter der Kinder festgemacht; im Bereich frühkindlicher Regulationsstörungen werden auch die Bereiche Schlafen und Füttern betrachtet (PSB-Index: nur Schreien); es werden Erkrankungen oder Behinderungen auch von Familienangehörigen oder älteren Geschwisterkindern als belastungsrelevant angesehen (PSB-Index: nur vom Zielkind) und schließlich werden auch Belastungen gut situierter Familien in den Blick genommen (für eine ausführliche Darstellung siehe Tabelle Z3 im Online-Material). „Und dann natürlich frage ich mich ja durch alle Strukturen durch. Wie ist es mit dem Schlafen? Wie ist es mit dem Essen? Wie ist es mit den Geschwistern? Gibt es Geschwisterrivalitäten? Und irgendwann kommen wir schon auf den Punkt, wo wir feststellen, ah, da liegt irgendwas im Argen“ [#P9]. Insgesamt nahmen 29 Praxispädiater*innen aus Baden-Württemberg (IG: n = 15) und Bayern (KG: n = 14) an der Studie teil. Wie in Tab. zu sehen ist, wiesen die Pädiater*innen der IG und der KG im Mittel sehr ähnliche soziodemografische und berufsbezogene Merkmale auf. Nur in Bezug auf die Anzahl der Früherkennungsuntersuchungen pro Monat gaben die Praxispädiater*innen der IG im Mittel eine niedrigere Anzahl an als die Ärzt*innen der KG. Insgesamt rekrutierten die teilnehmenden Praxispädiater*innen 549 Familien (IG: n = 251, KG: n = 298). Abb. zeigt den Fluss der Studienteilnehmenden bis zur Stichprobe der psychosozial belasteten Familien, die den Hypothesenprüfungen zugrunde lag. Diese Stichprobe umfasste 293 Familien, für die auf Basis ihrer Selbstauskunft mindestens ein Risikofaktor festgestellt wurde. Tab. zeigt die soziodemografischen Merkmale dieser Familien. Das Alter und das Geschlecht der Kinder, das Alter des antwortenden Elternteils und der Anteil der Familien mit Migrationsstatus fielen in IG und KG ähnlich aus. Allerdings wiesen in der IG mehr Familien einen höheren Schul- und Berufsabschluss und ein höheres Nettohaushaltseinkommen auf als in der KG. Diese Ähnlichkeiten und Unterschiede zwischen der IG und der KG zeigten sich auch in der Stichprobe der Familien, für die kein Risikofaktor ermittelt wurde (siehe Tabelle Z2 im Online-Material). Von den 10 telefonisch interviewten Praxispädiater*innen waren 5 (50 %) weiblich und 5 (50 %) männlich. Im Mittel arbeiten die Befragten seit 14 Jahren in einer niedergelassenen Praxis. Die 20 Interviews mit Familien wurden alle mit den Müttern der Kinder geführt, mit denen die Früherkennungsuntersuchung aufgesucht wurde (d. h. dem „Zielkind“). Die befragten Mütter waren im Mittel 35 Jahre alt und hatten 2 Kinder. Die Kinder, auf die sich die Mütter in den Interviews bezogen, waren im Durchschnitt etwa 1,5 Jahre alt. Zur Prüfung der Hypothese zur Hauptfragestellung wurde zunächst deskriptiv ermittelt, wie viele der gemäß familienseitigem Fragebogen psychosozial belasteten Familien auch von den Praxispädiater*innen der IG und KG als belastet eingeschätzt und somit identifiziert wurden. Wie in Abb. a zu sehen ist, traf dies auf 42 % (60 von 144) der Familien der IG zu und auf 23 % (34 von 149) der Familien der KG. Der deskriptive Gruppenunterschied betrug somit 19 Prozentpunkte. Dieser erwartungskonforme Unterschied zwischen der IG und der KG erwies sich in der inferenzstatistischen Prüfung als signifikant (Odds Ratio (OR) = 2,77; p = 0,020) und wird durch das Ergebnis der Sensitivitätsanalyse bekräftigt (OR = 2,76; p = 0,010). Tab. zeigt die Ergebnisse beider Analysen im Detail. Neben der arztseitigen Belastungseinschätzung geben die Ergebnisse der Telefoninterviews zudem Hinweise darauf, dass die Praxispädiater*innen bei der Abschätzung des Unterstützungsbedarfs einer Familie auch die familienseitigen Ressourcen und Lösungsstrategien in den Blick nehmen. „Also wichtig ist ja immer herauszufinden, wenn die Familie belastet ist, aber sie hat Strategien an der Hand, wie sie mit der Belastung umgehen, dann mache ich mir keine Sorgen“ [#P3]. Zur Prüfung der Hypothese zur ergänzenden Fragestellung wurde ebenfalls zunächst deskriptiv ermittelt, wie viele der psychosozial belasteten Familien in der IG und der KG je nach Anzahl der Risikofaktoren von den Praxispädiater*innen identifiziert wurden (Abb. b). Da nur sehr wenige Familien 5 oder mehr Risikofaktoren aufwiesen, wurden diese Familien zu einer Kategorie (≥ 5) zusammengefasst. Wie in Tab. (rechte Spalte) zu sehen ist, zeigte sich für den Prädiktor Gruppe x PSB ein Odds Ratio < 1, das nicht signifikant war ( p = 0,643), d. h., die Stärke des Interventionseffekts hing nicht signifikant von der Anzahl der Risikofaktoren einer Familie ab. Dieses Ergebnis ist auch in den in Abb. b dargestellten deskriptiven Häufigkeiten erkennbar: Der Gruppenunterschied bei Familien mit 1–4 Risikofaktoren betrug im Mittel 20 Prozentpunkte und fiel unabhängig von der Anzahl der Risikofaktoren ähnlich groß aus. Erst bei Familien mit 5 oder mehr Risikofaktoren fiel der Gruppenunterschied mit 11 Prozentpunkten deutlich kleiner aus. In Einklang mit diesen quantitativen Befunden liefern auch die qualitativen Ergebnisse Hinweise darauf, dass sich die PATH-Intervention unabhängig von der Stärke bzw. der Anzahl und Art der Belastungen der Familien positiv auf das Erkennen von Belastungen auswirken kann. Aus den Interviewergebnissen mit Praxispädiater*innen der IG geht hervor, dass die Teilnahme an den IQZ die Ärzt*innen für psychosoziale Belastungen von Familien zu sensibilisieren scheint. Der Austausch über Fälle (Fallbesprechungen), ein zentrales Element der IQZ, wird als hilfreich für das eigene ärztliche Handeln beschrieben. Dabei ist es insbesondere das Kennenlernen der verschiedenen Blickwinkel und Erfahrungen aus Gesundheitswesen und Kinder- und Jugendhilfe, das als wertvoll wahrgenommen wird. „… das Erste ist natürlich, dass man aus den Erfahrungen von anderen lernt und dass man beispielhaft dann auch für seine eigenen Patienten an den Beispielen lernt, dass man sensibilisiert wird finde ich ganz wichtig, dass man sensibilisiert zu bestimmten Themen …“ [#P3]. Zuletzt wurde explorativ überprüft, ob die PATH-Intervention einen Einfluss auf die ärztliche Einschätzung zur psychosozialen Belastung einer Familie hat, für die auf Basis des familienseitigen Fragebogens kein Risikofaktor festgestellt wurde (PSB-Score = 0). Diesbezüglich zeigte sich deskriptiv, dass in der IG 18 % (11 von 60) dieser Familien von ihrer*ihrem Praxispädiater*in als belastet eingeschätzt wurden, wohingegen dies in der KG für keine Familie (0 von 71) zutraf. Im exakten Test nach Fisher (2-seitig) war dieser Unterschied zwischen IG und KG signifikant ( p < 0,001). Eine Ursache dafür, dass in der IG ein höherer Anteil von Familien mit einem PSB-Score = 0 als belastet identifiziert wurde, könnte darin liegen, dass Ärzt*innen der IG weitere Arten von Belastungen wahrnehmen, die vom PSB-Index bislang nicht erfasst werden. Hierfür bieten sowohl die Interviews mit den Praxispädiater*innen der IG als auch die Interviews mit den Familien Anhaltspunkte. So werden Belastungen, die sich aufgrund der Betreuung mehrerer Kinder ergeben können, nicht (wie im PSB-Index) an einem sehr jungen Alter der Kinder festgemacht; im Bereich frühkindlicher Regulationsstörungen werden auch die Bereiche Schlafen und Füttern betrachtet (PSB-Index: nur Schreien); es werden Erkrankungen oder Behinderungen auch von Familienangehörigen oder älteren Geschwisterkindern als belastungsrelevant angesehen (PSB-Index: nur vom Zielkind) und schließlich werden auch Belastungen gut situierter Familien in den Blick genommen (für eine ausführliche Darstellung siehe Tabelle Z3 im Online-Material). „Und dann natürlich frage ich mich ja durch alle Strukturen durch. Wie ist es mit dem Schlafen? Wie ist es mit dem Essen? Wie ist es mit den Geschwistern? Gibt es Geschwisterrivalitäten? Und irgendwann kommen wir schon auf den Punkt, wo wir feststellen, ah, da liegt irgendwas im Argen“ [#P9]. Die Ergebnisse zur Hauptfragestellung zeigen, dass die PATH-Intervention zu einer signifikanten Verbesserung der Identifikationsrate psychosozial belasteter Familien mit kleinen Kindern führte. Der Anteil der psychosozial belasteten Familien, die von ihren Praxispädiater*innen identifiziert wurden, war in der IG etwa doppelt so groß wie in der KG. Absolut betrachtet konnte durch die PATH-Intervention zusätzlich etwa eine von 5 psychosozial belasteten Familien von Praxispädiater*innen identifiziert werden. Die vorliegenden Ergebnisse belegen somit die Wirksamkeit der PATH-Intervention in Bezug auf diesen wichtigen Teilaspekt im Vermittlungsprozess von der pädiatrischen Praxis in die Frühen Hilfen. In Deutschland wurden zwar bereits ähnliche Interventionen durchgeführt, die als Teilziel ebenfalls die Verbesserung der Identifikation psychosozial belasteter Familien beinhalteten . Inwieweit dieses Teilziel erreicht wurde, wurde bisher aber noch nicht berichtet, sodass eine Einbettung unserer Befunde in diese Studien ausbleiben muss. Die vorliegenden Ergebnisse stimmen jedoch mit Befunden zu dem amerikanischen Programm SEEK überein. Durch dieses Programm, das halbtägige Schulungen für Pädiater*innen zum Erkennen und Ansprechen von psychosozialen Belastungen umfasst, konnte die Häufigkeit des Einsatzes eines Screening-Fragebogens zur Erfassung psychosozialer Belastungen um 18–29 Prozentpunkte gesteigert werden . Die erzielten Steigerungen waren somit ähnlich hoch wie die Steigerungen der Identifikation psychosozial belasteter Familien durch die PATH-Intervention. Entgegen der Hypothese zur ergänzenden Fragestellung hing die Stärke des Effekts der PATH-Intervention nicht vom Ausmaß der Belastung der Familien ab. Zudem ging aus der Prüfung der explorativen Fragestellung hervor, dass die Praxispädiater*innen der IG einen größeren Anteil der Familien, für die im familienseitigen Fragebogen kein Risikofaktor festgestellt wurde, als psychosozial belastet einschätzten als Pädiater*innen der KG. Dieses Ergebnis kann einerseits als Hinweis interpretiert werden, dass Praxispädiater*innen der IG die psychosoziale Belastung von Familien überschätzen. Andererseits muss ein Wert von 0 im PSB-Index nicht notwendigerweise bedeuten, dass eine Familie unbelastet ist, da die Familie durchaus Belastungen aufweisen kann, die nicht vom PSB-Index erfasst werden. Hierfür sprechen die qualitativen Befunde, die Hinweise auf weitere in der Kinderarztpraxis auffallende Belastungen enthalten, die im PSB-Index nicht berücksichtigt werden. Die qualitativen Ergebnisse weisen zudem darauf hin, dass die PATH-Intervention Praxispädiater*innen für das Erkennen von Belastungen sensibilisieren kann. In jedem Fall führte die PATH-Intervention dazu, dass generell mehr Familien als belastet eingeschätzt wurden, ungeachtet dessen, wie viele Risikofaktoren die Familien aufwiesen. Dieser höhere Anteil identifizierter (auch geringfügig belasteter) Familien ist positiv zu bewerten, da bisher zu wenige belastete Familien von Praxispädiater*innen in Angebote der Frühen Hilfen übermittelt wurden und Praxispädiater*innen nur bei Familien, die sie als belastet einschätzen, weitere Schritte gehen, um die Belastungen der Familien anzusprechen und bei Bedarf eine Vermittlung in Angebote der Frühen Hilfen anzubahnen. Wenn die PATH-Intervention dazu führt, dass Praxispädiater*innen auch einen Anteil von Familien als belastet einschätzen, die möglicherweise nur gering belastet sind, aber aus ärztlicher Sicht dennoch Unterstützungsbedarf haben, muss dies keine negativen Auswirkungen im Sinne einer Fehlversorgung haben, da auch diese Familien von den präventiven Angeboten der Frühen Hilfen profitieren können . Es ist allerdings zu beachten, dass nicht alle psychosozial belasteten Familien externer Unterstützung bedürfen. Ob tatsächlich ein entsprechender Unterstützungsbedarf besteht, hängt auch davon ab, ob die Belastung einer Familie ihre eigenen Ressourcen übersteigt . Außer der Identifikation von Risikofaktoren spielen somit auch die Exploration und Einschätzung der Ressourcen der Familien durch Praxispädiater*innen eine wichtige Rolle für die Ermittlung eines Hilfebedarfs . Die qualitativen Ergebnisse weisen darauf hin, dass Ärzt*innen der IG entsprechend verfahren und prüfen, inwiefern Belastungslagen durch den Einbezug familialer Ressourcen abgemildert werden können. Einschränkungen der vorliegenden Studie betreffen vor allem die Instrumente, die zur familien- und arztseitigen Erfassung der psychosozialen Belastung eingesetzt wurden. Bei dem PSB-Index handelt es sich um ein Messinstrument, das im Kontext der Frühen Hilfen zur Bestimmung der Prävalenz psychosozialer Belastungen bereits in den großen repräsentativen KiD 0–3 Studien des NZFH eingesetzt wurde. Da der PSB-Index nicht alle potenziellen Risikofaktoren abdeckt, kann er zwar psychosozial belastete Familien identifizieren, unbelastete aber nicht. Die Hypothesenprüfung der vorliegenden Studie begrenzt sich deshalb auf die Identifikation von belasteten Familien, d. h. auf eine Überprüfung des Zugewinns an Sensitivität durch die PATH-Intervention. Eine Überprüfung der Identifikation von unbelasteten Familien, d. h. eine Überprüfung des Zugewinns an Spezifität durch die PATH-Intervention, konnte in der vorliegenden Studie nur ansatzweise vorgenommen werden. Den Fokus zunächst auf die Überprüfung des Zugewinns an Sensitivität zu legen, erscheint vor dem Hintergrund angemessen, dass die Identifikation von belasteten Familien für die weitere Versorgung dieser Familien von höherer Bedeutung ist als die Identifikation von unbelasteten Familien. Weiterhin wurde die psychosoziale Belastung familien- und arztseitig unterschiedlich erhoben. Im Gegensatz zur familienseitigen Messung der psychosozialen Belastung anhand verschiedener Indikatoren beurteilten die Praxispädiater*innen die psychosoziale Belastung der Familien global anhand eines einzigen Items (Schätzen Sie die Familie als psychosozial belastet ein? (Nein/Ja)). Die ärztliche Einschätzung war somit nicht auf die Merkmale begrenzt, auf die sich der PSB-Index stützt, sondern bezog sich vermutlich vielmehr auf einen ganzheitlichen Eindruck der Familie. Die ärztliche Einschätzung konnte zwar mit dem (dichotomisierten) Gesamtwert der familienseitig ermittelten psychosozialen Belastung verglichen werden, es war aber nicht möglich zu prüfen, ob der*die Praxispädiater*in die konkrete(n) Belastung(en) erkannt hat, die sich im PSB-Index abbildete(n). Ein solcher Vergleich wäre aufschlussreich, da damit ergründet werden könnte, ob der durch die PATH-Intervention erhöhte Anteil identifizierter psychosozial belasteter Familien auf ein besseres Erkennen bestimmter Belastungen zurückzuführen ist oder darauf, dass Praxispädiater*innen, die an der PATH-Intervention teilgenommen haben, eine höhere Sensibilität für Belastungen entwickelt haben. Beide Annahmen werden durch Befunde der qualitativen Interviews gestützt. In der vorliegenden Studie konnte erstmals nachgewiesen werden, dass die PATH-Intervention die Identifikation psychosozial belasteter Familien durch Praxispädiater*innen verbessert. Die Verbesserung betrug etwa 20 Prozentpunkte und war für psychosozial belastete Familien mit bis zu 4 Risikofaktoren, die den Großteil der untersuchten Stichprobe ausmachten, vergleichbar groß. Die Ergebnisse einer explorativen Analyse deuten darauf hin, dass die Teilnahme an der PATH-Intervention dazu führt, dass Praxispädiater*innen generell mehr Familien als belastet identifizieren – möglicherweise (auch) durch eine Wahrnehmung von Risikofaktoren, die mit dem PSB-Index bislang nicht erfasst werden. Mit der Verbesserung der Identifikation psychosozial belasteter Familien schafft die PATH-Intervention eine wichtige Grundlage, um die Vermittlung der betroffenen Familien aus der kinderärztlichen Praxis in passgenaue Unterstützungsangebote wie die der Frühen Hilfen zu verbessern. Infobox 1 Komponenten der PATH-Intervention Die Intervention besteht aus 2 Hauptbestandteilen: Regelmäßig stattfindende interprofessionelle Qualitätszirkel (IQZ), bei denen Fachkräfte des Gesundheitswesens und Mitarbeitende der Kinder- und Jugendhilfe zusammenkommen, um Fälle von belasteten Familien strukturiert aus einer interprofessionellen Perspektive zu besprechen. Außer dem fachlichen Austausch sollen durch die IQZ die interprofessionelle Kooperation und Vernetzung zwischen den Beteiligten verbessert werden. Eine eintägige Schulung für Praxispädiater*innen zu klinischer Fallfindung, die Anregungen zur Exploration und Identifikation von psychosozialen Belastungen enthält. In dieser Schulung werden die Praxispädiater*innen zudem in der Technik des motivierenden Elterngesprächs fortgebildet. Seit ihrer Einführung im Jahr 2010 wurde die PATH-Intervention in Baden-Württemberg in der Mehrzahl der Städte und Landkreise etabliert (Stand 2023: 9 Stadtkreise und 35 Landkreise). Die Intervention besteht aus 2 Hauptbestandteilen: Regelmäßig stattfindende interprofessionelle Qualitätszirkel (IQZ), bei denen Fachkräfte des Gesundheitswesens und Mitarbeitende der Kinder- und Jugendhilfe zusammenkommen, um Fälle von belasteten Familien strukturiert aus einer interprofessionellen Perspektive zu besprechen. Außer dem fachlichen Austausch sollen durch die IQZ die interprofessionelle Kooperation und Vernetzung zwischen den Beteiligten verbessert werden. Eine eintägige Schulung für Praxispädiater*innen zu klinischer Fallfindung, die Anregungen zur Exploration und Identifikation von psychosozialen Belastungen enthält. In dieser Schulung werden die Praxispädiater*innen zudem in der Technik des motivierenden Elterngesprächs fortgebildet. Seit ihrer Einführung im Jahr 2010 wurde die PATH-Intervention in Baden-Württemberg in der Mehrzahl der Städte und Landkreise etabliert (Stand 2023: 9 Stadtkreise und 35 Landkreise). Indikatoren psychosozialer Belastung und soziodemografische Merkmale der Familien, für die kein Risikofaktor ermittelt wurde |
Biomarker Profiles and Clinicopathological Features in Head and Neck Squamous Cell Carcinoma Patients | f10f53d1-a15a-4061-9586-a3f5f0be17f9 | 11509126 | Anatomy[mh] | Head and neck cancers are predominantly squamous cell carcinomas (SCC). However, tumors affecting different regions within this category can vary significantly in terms of invasiveness, growth rate, and metastatic potential . To better understand the behavior of these tumors, researchers worldwide have been increasingly focusing on biomarker studies over the past few decades. Biomarkers play a critical role in improving the precision and effectiveness of cancer diagnosis, prediction of treatment response, and management . In this study, we systematically examined the expressions of several published prognostic markers (Ki67, p53, EGFR, COX-2, Cx43) and p16 in head and neck squamous cell carcinomas from various anatomical regions using immunohistochemistry. Reports indicate that HPV (human papillomavirus)-positive oropharyngeal SCC represents a distinct clinicopathological entity with a better prognosis compared to HPV-negative oropharyngeal SCC . Multiple studies have shown that increased expression of p16 is an excellent surrogate marker for HPV-positive oropharyngeal SCC . Ki-67 is a protein that is associated with cell proliferation. High Ki-67 levels in head and neck cancer are generally associated with more aggressive tumors and poorer prognosis. Sometimes, Ki-67 is used as part of the grading system for tumors, helping to guide treatment planning . The p53 protein is crucial for DNA stability and cancer prevention. Its role in head and neck cancer is complex and context-dependent. As a tumor suppressor, p53 regulates the cell cycle, apoptosis, and genomic stability. High levels of normal p53 can indicate a positive prognosis by promoting apoptosis and DNA repair, aiding in cancer control and enhancing treatment effectiveness. Conversely, high p53 levels due to TP53 gene mutations result in stable, mutant p53 proteins, leading to uncontrolled growth, resistance to apoptosis, and genomic instability, often associated with more aggressive cancer and worse outcomes, including in HNSCC . Epidermal growth factor receptor (EGFR) is a protein that, when overexpressed or mutated, can play a significant role in the development and progression of various cancers. High EGFR levels in HNSCC indicate poor prognosis, as EGFR promotes tumor growth, invasion, and metastasis. It serves as both a prognostic marker and a therapeutic target. EGFR inhibitors, such as monoclonal antibodies and tyrosine kinase inhibitors, are critical in treatment . Cyclooxygenase-2 (COX-2) is an enzyme involved in the synthesis of prostaglandins, which play a role in inflammation and cell proliferation. COX-2 overexpression is associated with enhanced tumor growth, angiogenesis, and metastasis. Prostaglandins produced by COX-2 can promote these processes, contributing to the aggressiveness of the cancer. It is also considered both a prognostic marker and a potential therapeutic target in head and neck squamous cell carcinoma (HNSCC) . Membrane connexin 43 (Cx43) forms gap junctions, which are channels enabling direct communication between adjacent cells. Experimental data indicate that gap junction proteins play a significant role in tumor growth and progression. These intercellular channels, composed of about 21 connexins, facilitate metabolic cooperation, growth, and development. The absence of gap junctions and intercellular communication can contribute to tumorigenesis by promoting cell migration, invasion, and metastasis . Of the 21 connexins, Cx43 is the most well-known and extensively studied. It is broadly expressed in epithelial cells, hematopoietic cells, neurons, astrocytes, cardiac neural crest cells, and fibroblasts. Cx43 regulates cell proliferation and apoptosis by forming hemichannels that exchange growth and apoptotic factors, while also promoting tumor progression and metastasis. Recent studies suggest that Cx43 plays a more complex role in various stages of tumor progression . Assessing Cx43 expression can help determine the aggressiveness of cancer, guide treatment strategies, and explore targeted therapies that could potentially restore its function or enhance its tumor-suppressive properties. Measuring Cx43 levels in tumor tissues could also have potential diagnostic use . Aim: Our objective was to investigate how regional variations in the expression of potential biomarkers of SCC progression in the head and neck region contribute to understanding the diverse behaviors of these malignancies and to correlate results with clinicopathological parameters. We performed immunohistochemistry on 91 histologically verified HNSCC cases [48 laryngeal (L), 25 extralaryngeal (E), 18 cases affecting multiple regions (E-L)] from 2010 to 2016 selected from the archive of the County Emergency Hospital, Targu Mures. The laryngeal group (L) included HNSCC from supraglottic, glottic, or subglottic areas; the extralaryngeal group (E) covered tumors in the nasopharynx, oropharynx, hypopharynx, oral cavity, and mobile tongue; and the mixed group (E-L) comprised cases with both laryngeal and extralaryngeal spread. We classified patients’ squamous cell carcinomas into three groups: classic (CSCC), variants (CSCV—basaloid, spindle cell, acantholytic, verrucous, lymphoepithelial, papillary, and adenosquamous), and mixed-type (CSCM—tumors with features of two histological types). CSCC included keratinized and non-keratinized tumors. We used the WHO-recommended three-tier grading system to grade tumors: G1 for well-differentiated, G2 for intermediate, and G3 for poorly differentiated tumors. In the Immunohistochemistry Laboratory of the Department of Anatomy and Embryology at the “George Emil Palade” University of Medicine, Pharmacy, Science, and Technology, the 3 μm thick sections obtained from the formalin fixed and paraffin embedded resection tissue specimens were dewaxed and rehydrated followed by endogenous peroxidase blocking. Antigen retrieval was performed by pressurized steam cooking (citrate solution, pH 10 for p53, pH 6 for Cx43) or microwave treatment (pH 10 forKi67, EGFR, pH 6 for p16, COX-2). We used mouse monoclonal antibodies for p53 (clone DO-7, Novocastra Laboratories Ltd., Leica Biosystems, Deer Park, IL, USA) in 1/800, Ki67 (clone MM1, Novocastra Laboratories Ltd., Leica Biosystems, Deer Park, IL, USA) in 1/150, EGFR (clone EGFR.113, Novocastra Laboratories Ltd., Leica Biosystems, Deer Park, IL, USA) in 1/40, COX-2 (clone COX229, Thermo Fisher Scientific, Waltham, MA, USA) in 1/100 for 1 h, as well as Cx-43 (clone CX-1B1, Thermo Fisher Scientific, Waltham, MA, USA) in 1/100 (overnight, 4 °C). For p16, we applied rabbit clonal antibody (clone R19-D, DB Biotech, Inc., Kosice, Slovakia) in 1/200 (1 h). For signal amplification, the secondary antibody EnVision Flex/HRP (Horseradish peroxidase) (Dako, 20 min) was used. Detection of primary antibodies was achieved using 3,3′-Diaminobenzidine (DAB, Dako, Santa Clara, CA, USA). The slides were then counterstained with hematoxylin, dehydrated, and mounted. Negative controls were performed by omitting the primary antibody. Immunohistochemical reactions were read by an independent pathologist. Nuclear expression was observed for p53 and Ki67 (only the infiltrative component on 1000 cells), while COX-2 displayed cytoplasmic expression. The expression was membranous for EGFR and Cx43. We interpreted the results for Ki67, COX-2, and Cx43 according to the following immunoexpression scoring system: S1: 0–10%, S2: 1–25%, S3: 26–50%, S4 > 50%. For p16, we classified the reaction as positive if at least 75% of the infiltrative tumor cells showed intense cytoplasmic and nuclear positivity, considering the remaining reactions as negative. We classified p53 into two categories: mutant and non-mutant (wild-type). Three types of reactions were considered as mutants: negative nuclear reaction (no cells with a positive nuclear reaction), positive nuclear reaction (more than 90% of tumor cells present an intense, uniform nuclear reaction), and cytoplasmic reaction (tumor cells present an aberrant positive cytoplasmic reaction and not nuclear), while all other reactions were classified as wild-type. EGFR was scored based on membrane staining intensity: 0 = no staining or <10% of tumor cells show membrane staining; 1+ = faint membrane staining in >10% of tumor cells; 2+ = moderate membranous staining for at least 10% of tumor cells; 3+ = >10% of tumor cells with strong membranous staining. For statistical analysis, we used GraphPad InStat 3 software, version 3.06 (GraphPad Software Inc., San Diego, CA, USA). A significant association was taken into consideration at a p value of <0.05, with a 95% confidence interval. Ethics. This study was approved by the ethics committee of “George Emil Palade” University of Medicine, Pharmacy, Science, and Technology of Târgu Mureș, Romania (no. 3211/10/06/2024, provided on 10 June 2024). Most of the tumors were seen in males (95.6%) aged between 51 to 60 (41.76%). The median age at diagnosis was 65 years (range: 39–81 years). Among the main histopathological groups observed, 74 cases (81%) were classified as classic squamous cell carcinoma (CSCC), 6 cases (7%) as variants (CSCV), and 11 cases (12%) as mixed-type (CSCM). Most tumors were classified as grade G2 (41 cases, 51.25%) or G3 (37 cases, 46.25%), with only 2.5% being well-differentiated (G1). 3.1. Ki67 Immunoexpression Most CSCC cases showed S1 (40.62%) and S3 (32.81%) immunoexpression. In the variants, elevated Ki67 expression was observed (S3: 40%, S4: 40%), while in the mixed group, most cases were classified in the middle expression grades (S2: 33.33%, S3: 44.44%). Regarding tumor grading, there was minimal Ki67 expression in well-differentiated tumors. Intermediate differentiated grade tumors mostly showed S1 immunoexpression (51.42%), while poorly differentiated tumors predominantly exhibited S3 expression (35.48%). For tumors located in the L and E-L regions, most tumor cells exhibited S1 and S3 expression. In contrast, the majority of extralaryngeal tumors were S3 (54.54%) and S4 (27.27%). In men, most tumors exhibited S1 (37.83%) and S3 (33.78%) expression. On the other hand, in women, half of the cases were classified as S3 for Ki67 staining, with the remaining cases were equally divided between S2 and S4, and no cases were classified as S1. The ratio of cases with high Ki67 immunoexpression (S4) was higher in patients under 70 years old (29%, 22/75), compared to tumors developed in patients over 71 years old, where moderate immunoexpression (S3) was more common (5/10, 50%) ( and ). Ki67 expression correlated significantly with tumor localization ( p = 0.01) and grade of the tumors ( p = 0.048) ( A,B). 3.2. p53 Immunoexpression For mutant p53 expression, we found 97.95% of cases with negative mutant reactions (0% cell staining) and 2.04% with positive mutant reactions (>90% cell staining); there was no cytoplasmic staining. When examining the localization, mutant p53 expression was substantially higher in both laryngeal and extralaryngeal tumors (L: 64.28%, E: 68.18%). While in the mixed group of tumors, the opposite was observed, with wild-type p53 expression being higher (E-L: 61.11%). Tumors from the CSCV group exclusively exhibited mutant p53 expression, which was also more prevalent in the CSCC malignancies (58.82%). Wild-type staining was higher only in the CSCM group (55.5%). Among women, all cases showed mutant p53 expression, while 58.22% of cases in men were classified similarly. The results showed mutant p53 expression across all tumor grades, including well-differentiated, intermediate, and poorly differentiated tumors. The ratio of cases with mutant p53 expression was higher in patients under 60 years old (32/45, 71.11%) compared to patients over 61 years old (17/37, 45.94%). p53 expression demonstrated a statistically significant correlation with the patients age ( A–C). 3.3. p16 Immunoexpression In the analysis of all the factors, low p16 expression was observed in the majority of tumors, and based on the positivity criteria, p16 expression did not reach a positive level in any case. 3.4. EGFR Immunoexpression Among the histological types, most CSCC (33.33%) and CSCV (50%) tumors showed a 2+ reaction, while CSCM tumors exhibited weak (1+: 33.33%) or no (0+: 33.33%) expression in most cases. In poorly differentiated tumors, EGFR expression was most evenly distributed (0+: 35.48%, 1+: 22.58%, 2+: 19.35%, 3+: 22.58%), likewise in the L region (0+: 25.64%, 1+: 28.20%, 2+: 25.64%, 3+: 20.51%). The same pattern was observed when examining cases separately for women and men. In both the youngest and oldest patients, 1+ expression was the most common; whereas, in the 51–60 and 61–70 age groups, 2+ and 3+ expression were more prevalent. EGFR expression does not show statistically significant correlations with localization, histology type, age, or gender ( p > 0.05) ( A,B). 3.5. COX-2 Immunoexpression Studying COX-2 expression across different histological types, we found that S1, S3, and S4 expression occurred at similar rates in all three groups; whereas, S2 expression was observed in fewer tumors (CSCC: 7.35%, CSCV: 0%, CSCM: 0%). Most tumors exhibited S3 expression across all anatomical regions (L: 33.33%, E: 47.61%, E-L: 25.29%). The tumor grading was directly proportional to the immunoexpression score. Women exhibited either weak or increased staining, while in men, the distribution was more balanced (S1: 28%, S2: 8%, S3: 38.66%, S4: 25.33%). The ratio of cases with moderate and high immunoexpression (S3 + S4) was higher in patients under 70 years old (45/67, 67.16%) compared to those over 71 years old, where low immunoexpression (S1) was more prevalent (5/10, 50%) ( C,D). 3.6. Cx-43 Immunoexpression Cx-43 expression in most CSCC (85.29%) and CSCV (100%) tumors was S1. In CSCM tumors, the distribution was more balanced, with S1 at 72.72%, S2 at 18.18%, and S3 at 9.09%, but no cases had more than 51% expression. Regarding tumor grading, the number of tumors with S1 expression decreased as the tumors became less differentiated, while an increase in S2 expression was observed with poorer differentiation levels. Considering all anatomical regions, low-expression tumors were the most common everywhere (L:39.06%, E:95.45%, E-L 68.18%). In the age groups, we observed that while most tumors exhibited low expression, the number of tumors showing S2 and S3 expression was similar across all age groups. In men, 83.75% of tumors had S1 staining, while in women, no tumors exhibited expression greater than S1 immunoexpression (11%) ( C,D). 3.7. Comparative Analysis of Cx43 Expression Levels We performed a comparative analysis of Cx43 expression levels in relation to the expression of Ki67, COX-2, and EGFR, particularly in the context of p53 mutation status. When evaluating Cx43 expression levels with p53 status, it appeared that the majority of cases in both groups exhibit low Cx43 expression (S1). In the case of tumors with mutant p53 along with Cx43 S1 immunoexpression (84.7%), the most frequent were cases with Cx43 S3 immunoexpression (8.7%), while in wild-type tumors, S2 was more common (S1: 81.2%, S2: 12.5%) ( p = 0.63) . Comparing Ki67 immunoexpression with Cx43, we observed that all cases with Ki67 S1 immunoexpression were associated with Cx43 S1 expression ( p = 0.16). The number of cases with partially preserved Cx43 immunoexpression increased with higher Ki67 levels. However, these cases still exhibited relatively high Ki67 expression (S3 and S4), indicating that moderate Cx43 expression might still be associated with higher proliferation rates . The data suggest that low Cx43 expression (S1) is common and is associated with a wide range of COX-2 expression. Cases with COX-2 S1 immunoexpression were more frequently associated with Cx43 S1 immunoexpression (93.1%) compared to those with COX-2 immunoexpression greater than 26% (78%). This pattern indicates that moderate Cx43 expression might be linked to higher levels of COX-2 activity ( p = 0.12) . When evaluating the EGFR expression, we found that is widely distributed across all levels (0 to 3+) within the S1 expression level of Cx43. Tumors with EGFR 2+ and 3+ expression are more frequently associated with Cx43 S1 expression (87.1%) compared to those with EGFR 0 and 1+ expression (81%). As Cx43 expression increases to moderate levels, the number of cases declines, but EGFR expression remains varied ( p = 0.43) . The relationship between Cx43 expression and other markers (Ki67, COX-2, EGFR, p53) varies and is complex. Low Cx43 expression is common across all groups and does not show a strong correlation with the other markers. However, low Cx43 expression may be more likely to coincide with higher Ki67 and COX-2 levels, possibly indicating a link with more aggressive tumor behavior. EGFR expression, however, appears to be independent of Cx43 levels. Most CSCC cases showed S1 (40.62%) and S3 (32.81%) immunoexpression. In the variants, elevated Ki67 expression was observed (S3: 40%, S4: 40%), while in the mixed group, most cases were classified in the middle expression grades (S2: 33.33%, S3: 44.44%). Regarding tumor grading, there was minimal Ki67 expression in well-differentiated tumors. Intermediate differentiated grade tumors mostly showed S1 immunoexpression (51.42%), while poorly differentiated tumors predominantly exhibited S3 expression (35.48%). For tumors located in the L and E-L regions, most tumor cells exhibited S1 and S3 expression. In contrast, the majority of extralaryngeal tumors were S3 (54.54%) and S4 (27.27%). In men, most tumors exhibited S1 (37.83%) and S3 (33.78%) expression. On the other hand, in women, half of the cases were classified as S3 for Ki67 staining, with the remaining cases were equally divided between S2 and S4, and no cases were classified as S1. The ratio of cases with high Ki67 immunoexpression (S4) was higher in patients under 70 years old (29%, 22/75), compared to tumors developed in patients over 71 years old, where moderate immunoexpression (S3) was more common (5/10, 50%) ( and ). Ki67 expression correlated significantly with tumor localization ( p = 0.01) and grade of the tumors ( p = 0.048) ( A,B). For mutant p53 expression, we found 97.95% of cases with negative mutant reactions (0% cell staining) and 2.04% with positive mutant reactions (>90% cell staining); there was no cytoplasmic staining. When examining the localization, mutant p53 expression was substantially higher in both laryngeal and extralaryngeal tumors (L: 64.28%, E: 68.18%). While in the mixed group of tumors, the opposite was observed, with wild-type p53 expression being higher (E-L: 61.11%). Tumors from the CSCV group exclusively exhibited mutant p53 expression, which was also more prevalent in the CSCC malignancies (58.82%). Wild-type staining was higher only in the CSCM group (55.5%). Among women, all cases showed mutant p53 expression, while 58.22% of cases in men were classified similarly. The results showed mutant p53 expression across all tumor grades, including well-differentiated, intermediate, and poorly differentiated tumors. The ratio of cases with mutant p53 expression was higher in patients under 60 years old (32/45, 71.11%) compared to patients over 61 years old (17/37, 45.94%). p53 expression demonstrated a statistically significant correlation with the patients age ( A–C). In the analysis of all the factors, low p16 expression was observed in the majority of tumors, and based on the positivity criteria, p16 expression did not reach a positive level in any case. Among the histological types, most CSCC (33.33%) and CSCV (50%) tumors showed a 2+ reaction, while CSCM tumors exhibited weak (1+: 33.33%) or no (0+: 33.33%) expression in most cases. In poorly differentiated tumors, EGFR expression was most evenly distributed (0+: 35.48%, 1+: 22.58%, 2+: 19.35%, 3+: 22.58%), likewise in the L region (0+: 25.64%, 1+: 28.20%, 2+: 25.64%, 3+: 20.51%). The same pattern was observed when examining cases separately for women and men. In both the youngest and oldest patients, 1+ expression was the most common; whereas, in the 51–60 and 61–70 age groups, 2+ and 3+ expression were more prevalent. EGFR expression does not show statistically significant correlations with localization, histology type, age, or gender ( p > 0.05) ( A,B). Studying COX-2 expression across different histological types, we found that S1, S3, and S4 expression occurred at similar rates in all three groups; whereas, S2 expression was observed in fewer tumors (CSCC: 7.35%, CSCV: 0%, CSCM: 0%). Most tumors exhibited S3 expression across all anatomical regions (L: 33.33%, E: 47.61%, E-L: 25.29%). The tumor grading was directly proportional to the immunoexpression score. Women exhibited either weak or increased staining, while in men, the distribution was more balanced (S1: 28%, S2: 8%, S3: 38.66%, S4: 25.33%). The ratio of cases with moderate and high immunoexpression (S3 + S4) was higher in patients under 70 years old (45/67, 67.16%) compared to those over 71 years old, where low immunoexpression (S1) was more prevalent (5/10, 50%) ( C,D). Cx-43 expression in most CSCC (85.29%) and CSCV (100%) tumors was S1. In CSCM tumors, the distribution was more balanced, with S1 at 72.72%, S2 at 18.18%, and S3 at 9.09%, but no cases had more than 51% expression. Regarding tumor grading, the number of tumors with S1 expression decreased as the tumors became less differentiated, while an increase in S2 expression was observed with poorer differentiation levels. Considering all anatomical regions, low-expression tumors were the most common everywhere (L:39.06%, E:95.45%, E-L 68.18%). In the age groups, we observed that while most tumors exhibited low expression, the number of tumors showing S2 and S3 expression was similar across all age groups. In men, 83.75% of tumors had S1 staining, while in women, no tumors exhibited expression greater than S1 immunoexpression (11%) ( C,D). We performed a comparative analysis of Cx43 expression levels in relation to the expression of Ki67, COX-2, and EGFR, particularly in the context of p53 mutation status. When evaluating Cx43 expression levels with p53 status, it appeared that the majority of cases in both groups exhibit low Cx43 expression (S1). In the case of tumors with mutant p53 along with Cx43 S1 immunoexpression (84.7%), the most frequent were cases with Cx43 S3 immunoexpression (8.7%), while in wild-type tumors, S2 was more common (S1: 81.2%, S2: 12.5%) ( p = 0.63) . Comparing Ki67 immunoexpression with Cx43, we observed that all cases with Ki67 S1 immunoexpression were associated with Cx43 S1 expression ( p = 0.16). The number of cases with partially preserved Cx43 immunoexpression increased with higher Ki67 levels. However, these cases still exhibited relatively high Ki67 expression (S3 and S4), indicating that moderate Cx43 expression might still be associated with higher proliferation rates . The data suggest that low Cx43 expression (S1) is common and is associated with a wide range of COX-2 expression. Cases with COX-2 S1 immunoexpression were more frequently associated with Cx43 S1 immunoexpression (93.1%) compared to those with COX-2 immunoexpression greater than 26% (78%). This pattern indicates that moderate Cx43 expression might be linked to higher levels of COX-2 activity ( p = 0.12) . When evaluating the EGFR expression, we found that is widely distributed across all levels (0 to 3+) within the S1 expression level of Cx43. Tumors with EGFR 2+ and 3+ expression are more frequently associated with Cx43 S1 expression (87.1%) compared to those with EGFR 0 and 1+ expression (81%). As Cx43 expression increases to moderate levels, the number of cases declines, but EGFR expression remains varied ( p = 0.43) . The relationship between Cx43 expression and other markers (Ki67, COX-2, EGFR, p53) varies and is complex. Low Cx43 expression is common across all groups and does not show a strong correlation with the other markers. However, low Cx43 expression may be more likely to coincide with higher Ki67 and COX-2 levels, possibly indicating a link with more aggressive tumor behavior. EGFR expression, however, appears to be independent of Cx43 levels. The incidence of head and neck cancers continues to show an increasing trend worldwide. In our study population, 95% of the patients were male, with most aged between 51 and 60 years; although, the number of female patients is also rising . HNSCCs are a diverse, aggressive, and genetically complex group of malignancies. For decades, the prognosis of tumors and their expected response to different treatments have mainly been predicted using the TNM classification. In recent years, many studies have explored the importance of various immunohistochemical markers and their clinical utility in offering more precise and personalized prognoses. Across the range of markers studied, one of the most promising is EGFR, with its overexpression or aberrant activation potentially leading to uncontrolled cell growth and tumor development . We observed milder EGFR expression in both younger and older patients, while those aged 50–70 exhibited higher expression levels, which may indicate more aggressive tumor growth and a poorer prognosis in this age group, as reflected in the literature . In the analysis of all factors, low p16 expression was observed in most tumors, not reaching the positivity criteria in any case. This largely aligns with the literature, which indicates that p16 positivity is less likely to be expected of tumors that originated from or involved the larynx. Additionally, none of the tumors diagnosed in extralaryngeal regions showed positive p16 expression . Ki67 levels provide insights into cancer proliferative activity and significantly correlate with tumor grade ( p = 0.048). Well-differentiated tumors showed minimal Ki67 expression, intermediate-grade tumors mostly exhibited S1 expression (51.42%), and poorly differentiated tumors predominantly had S3 expression (35.48%). This finding supports existing literature, which indicates that higher Ki67 expression is associated with a greater likelihood of poorly differentiated, poorer-prognosis tumors . We observed higher Ki67 expression in extralaryngeal tumors compared to tumors originating from or involving the larynx, with a significant correlation based on tumor localization ( p = 0.01). Examining p53 expression in relation to various factors, we observed that mutant p53 expression was prevalent in most cases, indicating a worse prognosis. Specifically, younger patients (under 60 years) had a higher ratio of tumors with mutant p53 expression. This correlation with age was statistically significant ( p = 0.02), suggesting that tumors in younger individuals may be more aggressive and associated with a worse prognosis, reflecting a more aggressive tumor phenotype. The gender distribution analysis revealed that all tumors in women exhibited mutant p53 expression, indicating a worse prognosis compared to men and suggesting the need for more complex therapeutic solutions. The majority of cases with mutant p53 expression had negative mutant reactions with 0% cell staining. Currently, the literature does not emphasize whether mutant p53 expression is classified as negative or positive. Wild-type p53 expression was observed only in histopathologically mixed-type tumors, suggesting that tumors with the features of two histological types may be more aggressive . No significant differences in COX-2 expression were observed across different tumor localizations, with most anatomical regions showing S3 expression, which means the level of COX-2 did not vary much depending on where the tumor was located. The tumor grading was directly proportional to the immunoexpression score. Knowing that high COX-2 levels indicate cancer aggressiveness, this suggests that more aggressive or targeted treatment strategies may be necessary to manage these cases effectively. The ratio of cases with moderate and high immunoexpression (S3 + S4) was higher in younger patients (under 70 years old: 67.16%) compared to those over 71 years old (50%), where low immunoexpression (S1) was more prevalent, suggesting the presence of less aggressive tumors with advancing age . Cx43 is generally considered to have a tumor-suppressive function. It often has reduced expression or altered function in cancers such as HNSCC, leading to impaired cell communication and uncontrolled cancer cell growth, thus contributing to cancer progression . Low-expression tumors were most common across all anatomical regions, with extralaryngeal tumors showing especially high rates of low expression, where 95.45% exhibited S1 expression. This may indicate that tumors in the extralaryngeal region have a poorer prognosis in the studied population. In men, 16.25% of the tumors exhibited stronger expression (S2 + S3); whereas in women, all tumors had low expression. According to the literature, this may suggest that women could develop more aggressive head and neck tumors with a poorer prognosis . The present study highlights that HNSCCs are predominant in older males, with the larynx being the most common site. Our study explored the relationship between histological and clinical parameters and biomarker profiles in HNSCC. High levels of EGFR and Ki67 expression are linked with more aggressive tumors, while low p16 expression and COX-2 levels suggest varying prognoses based on tumor localization and age. Additionally, the findings also suggests that women may develop more aggressive tumors with poorer outcomes, and extralaryngeal tumors often have a more challenging prognosis. Despite progress in understanding HNSCC, finding a consistently reliable biomarker profile is still ongoing. A definitive profile could enhance diagnosis, prognosis, and treatment, but more research and validation are needed. |
Using artificial intelligence to improve human performance: efficient retinal disease detection training with synthetic images | e7a4d6a8-3549-4ec0-866b-b4ce6f84fd6d | 11503156 | Ophthalmology[mh] | Artificial intelligence (AI) models can accurately detect retinal disease but still have multiple challenges for real-world implementation. Humans can generalise better but have limited training opportunities. A scalable teaching method was needed to improve human performance. Using AI-generated synthetic images for training rapidly enhances humans' diagnostic accuracy to match state-of-the-art AI models while preserving robustness. This teaching approach could be applied in other medical imaging fields to train experts worldwide, improving patient outcomes efficiently. It also showcases AI’s potential to empower human skills rather than replace them. Ageing populations worldwide are placing unprecedented pressure on healthcare systems, and the shortage of medical experts is emerging as a critical challenge that must be addressed to ensure the sustainability of healthcare delivery. Artificial intelligence (AI) for medical diagnosis holds potential, demonstrating high performance in detecting various diseases. However, challenges such as decreased performance on unseen imaging modalities and the commercially available AI diagnosis, which is difficult for medical insurance systems to cover, need to be addressed to fully harness AI’s potential and ensure its accessibility and effectiveness for patients worldwide. The lack of large, curated imaging repositories often hinders the full human potential in medical image diagnosis. Medical students might only see a few examples per disease, making it difficult to recognise all cases in practice where there is a large variety in disease presentation and patient morphology. AI models, on the other hand, are typically trained on many thousands or even millions of images. Recently, large text-to-image generative foundation models like stable diffusion (SD) have been developed to generate many images faster and with higher quality than generative adversarial networks, the previous state-of-the-art. We hypothesise that training humans on many examples can be an effective way of teaching and allowing them to match or even outperform specialist state-of-the-art AI models. We developed an AI-assisted teaching approach that intensively enhances non-experts' training by focusing on image analysis, similar to AI model training. We leverage generative AI to create synthetic images to show students many examples without concerns about protecting individual medical data privacy. We test our hypothesis and approach to retinal disease detection in ultra-widefield (UWF) (220°) retinal images. We compared the identification abilities of learners who completed the task with those of AI and experienced experts. We also examined the robustness of their diagnostic capability in the face of a novel imaging modality not covered during training, namely a 50° standard field (SF) of view obtained using a completely different imaging device. Study overview We designed a web-based training course that uses synthetic images to teach students how to recognise disease in UWF retinal images. We conduct a test of real images to evaluate the students’ performance. Each student took the test before and after the training course to assess the effectiveness of our teaching method. Notably, the students were not given feedback when taking the test, and at least 1 week passed between taking the test before and after training. We also compared experienced experts and an AI system on the same test. Additionally, we used a second test consisting of SF retinal images to evaluate how well participants generalise to an unseen modality. We considered images belonging to one of six classes, namely standard, healthy retinas (Normal) and five critical retinal diseases: diabetic retinopathy (DR), age-related macular degeneration (AMD), glaucoma (Gla), retinal vein occlusion (RVO) and retinal detachment (RD). Study participants: students and experts We tested our method on 161 trainee orthoptists at different stages of a 4-year university programme. This is the entire student body of this course, and the implementation rate was 100%. The trainee orthoptist’s curriculum includes ophthalmic diseases; their knowledge in this area is expected to deepen with every year of study. Including students at different stages of their education allows us to evaluate whether our approach is practical for varying levels of pre-existing knowledge. Students were instructed by the student leader (FM) not to engage in any additional learning activities until the experiment’s conclusion. To contextualise the students' performance, we also had eight experienced experts take the test, including five certified orthoptists and three retinal specialists who are experienced ophthalmologists in the surgical retinal diseases field, both experts at interpreting retinal images. Certified orthoptists are qualified professionals who specialise in performing ophthalmic examinations based on ophthalmic knowledge and hold a national medical qualification in Japan. Each of the five orthoptists involved in this study had over 10 years of practical experience. The three retinal specialists all held the ophthalmic specialist certification recognised by the Japan Ophthalmological Society and had more than 10 years of experience in retinal and vitreous surgery. An overview of the demographics and level of education of the participants is shown in . All students are affiliated with a 4-year university that trains certified orthoptists (a national ophthalmic examination specialty certification in Japan). The nurses in this study are a group that has not specialised in the ophthalmic field. Retina specialists are ophthalmologists who specifically specialise in retinal diseases. Evaluation tests To evaluate the ability of study participants to recognise disease in UWF images (Optos 200Tx, Nikon, Japan), we used a web-based test consisting of 120 images, 20 per class. The participant could choose one of seven options for each image: the target labels and an ‘I don’t know’ option. Unlike the training course, correct answers were not displayed after a student made a selection. We used a second test with a different type of retinal imaging to evaluate how well students generalise to an unseen image modality. This test is identical in design but uses SF images (optical coherence tomography-assisted SF fundus camera, Triton, Topcon, Japan) instead of UWF images. An example of each type is shown in . Both provide a retinal picture, but there are two key differences. First, SF images only capture a narrow 50° field of view around the posterior pole, the central part of the retina, whereas UWF images capture a 220° field of view. Second, the UWF camera uses two lasers to scan the retina to produce a pseudocolour image, whereas a standard true colour optical camera has SF images. This leads to a difference in appearance and scale and different types of imaging artifacts can occur in either modality. Both SF and UWF images were selected by the ophthalmologist (HT) based on the clarity of the lesion within their respective imaging ranges, making them suitable for diagnosis. Consequently, the patients in the SF images differed from those in the UWF images. The background of each group of patients is presented in . AI-based image generation SD V.1.4, a large pretrained text-to-image foundation model, was fine-tuned on 6285 UWF retinal images captured with an Optos 200Tx UWF camera (Nikon, Japan) using DreamBooth to produce novel, synthetic UWF images. The dataset consisted of 1666 DR images, 215 AMD images, 1316 Gla images, 393 RVO images, 468 RD images and 2227 normal images. (The images used in the evaluation test are not included.) All images were selected based on diagnoses agreed upon by two or more ophthalmologists: RS, HT and TY. We fine-tuned a separate SD model for each class and generated images until 100 suitable SD images were selected. The same ophthalmologists (RS, HT and TY) assessed the generated images. They selected those suitable for teaching based on the clarity of features unique to the target pathology without additional findings suggesting other diseases. The background of each image of a patient is presented in . The number of generated SD images and selection rate per class was 8000 images for DR with a rate of 1.3%, 1500 for AMD (6.7%), 1000 for Gla (10%), 500 for BRVO (20%), 1100 for RD (9.1%) and 3500 for normal (0.29%). We fine-tuned the SD model and generated the images on a Nvidia RTX3090TI 24GB GPU using Python V.3.1.0. and the PyTorch V.1.13.0 and Diffusers V.0.10.2 libraries. We have provided more details in the ‘SD optimisation’ section of the . Artificial synthetic images generated with our fine-tuned SD model are shown in . Web-based training course using synthetic images During the course, the student is shown an image and asked to select the class they think it belongs to . After responding, the correct answer is immediately displayed alongside some image annotations highlighting and explaining the image’s relevant parts . We generated 600 synthetic images with SD, 100 per class, and divided them into five sets that were used for teaching. Each of the five sets consists of 120 synthetic images, with 20 images of each of the six types. They are named from No. 1 to No. 5. Students completed all five sets in sequence from the No. 1 set to the No. 5 set unless their performance exceeded the accuracy of the worst-performing expert, in which case training would end early. Verification to memorise problems by the SD model To check that the SD model did not memorise any real images (verification to ensure that SD is not simply outputting the learning images as they are), we compared all 600 synthetic images with all 6285 real images used for fine-tuning. We used self-supervised copy detection to find the most similar pairs of real and synthetic images, which were manually inspected. None of these pairs were visually identical, suggesting that our SD model did not memorise images. We have provided more details in the ‘Evaluation of Similarity’ section in . State-of-the-art AI model for comparison As a further comparison, we evaluated a recently proposed, state-of-the-art AI model previously published by researchers specialising in machine learning, which has achieved excellent performance on internal and external test sets. This model was trained on a dataset of 5376 patients (8570 eyes, 13 026 images of Optos 200Tx) and achieved an area under the curve (AUC) of 0.9848 (±0.0004) for detecting disease in UWF images on the same evaluation tests provided to the students. The model produces a probabilistic output by default that needs to be mapped to discrete predictions. We used the ‘conservative’ decision threshold for detecting images with disease, as proposed by the original authors. If an image has been classified as showing disease, the disease with the highest probability is then taken as the model’s prediction. Statistical analysis Statistical analysis was performed using JMP V.16.2.0 (SAS Institute, Cary, NC, USA). Paired t-tests were used for within-subject significance tests regarding learning effects. We conducted a non-parametric multiple comparison test (Steel-Dwass test) when comparing study times and the time required to answer each question on the evaluation test among groups of students. Ethics and data The retinal images used in this study were acquired during clinical practice at Tsukazaki Hospital, Himeji, Japan. We obtained express written consent from each patient for the research use of their data. Written explanations and consent were received from the students, and prior explanations and consent were obtained from the experts. Our research was conducted in accordance with the Declaration of Helsinki and approved by the ethics committee of Tsukazaki Hospital. We designed a web-based training course that uses synthetic images to teach students how to recognise disease in UWF retinal images. We conduct a test of real images to evaluate the students’ performance. Each student took the test before and after the training course to assess the effectiveness of our teaching method. Notably, the students were not given feedback when taking the test, and at least 1 week passed between taking the test before and after training. We also compared experienced experts and an AI system on the same test. Additionally, we used a second test consisting of SF retinal images to evaluate how well participants generalise to an unseen modality. We considered images belonging to one of six classes, namely standard, healthy retinas (Normal) and five critical retinal diseases: diabetic retinopathy (DR), age-related macular degeneration (AMD), glaucoma (Gla), retinal vein occlusion (RVO) and retinal detachment (RD). We tested our method on 161 trainee orthoptists at different stages of a 4-year university programme. This is the entire student body of this course, and the implementation rate was 100%. The trainee orthoptist’s curriculum includes ophthalmic diseases; their knowledge in this area is expected to deepen with every year of study. Including students at different stages of their education allows us to evaluate whether our approach is practical for varying levels of pre-existing knowledge. Students were instructed by the student leader (FM) not to engage in any additional learning activities until the experiment’s conclusion. To contextualise the students' performance, we also had eight experienced experts take the test, including five certified orthoptists and three retinal specialists who are experienced ophthalmologists in the surgical retinal diseases field, both experts at interpreting retinal images. Certified orthoptists are qualified professionals who specialise in performing ophthalmic examinations based on ophthalmic knowledge and hold a national medical qualification in Japan. Each of the five orthoptists involved in this study had over 10 years of practical experience. The three retinal specialists all held the ophthalmic specialist certification recognised by the Japan Ophthalmological Society and had more than 10 years of experience in retinal and vitreous surgery. An overview of the demographics and level of education of the participants is shown in . All students are affiliated with a 4-year university that trains certified orthoptists (a national ophthalmic examination specialty certification in Japan). The nurses in this study are a group that has not specialised in the ophthalmic field. Retina specialists are ophthalmologists who specifically specialise in retinal diseases. To evaluate the ability of study participants to recognise disease in UWF images (Optos 200Tx, Nikon, Japan), we used a web-based test consisting of 120 images, 20 per class. The participant could choose one of seven options for each image: the target labels and an ‘I don’t know’ option. Unlike the training course, correct answers were not displayed after a student made a selection. We used a second test with a different type of retinal imaging to evaluate how well students generalise to an unseen image modality. This test is identical in design but uses SF images (optical coherence tomography-assisted SF fundus camera, Triton, Topcon, Japan) instead of UWF images. An example of each type is shown in . Both provide a retinal picture, but there are two key differences. First, SF images only capture a narrow 50° field of view around the posterior pole, the central part of the retina, whereas UWF images capture a 220° field of view. Second, the UWF camera uses two lasers to scan the retina to produce a pseudocolour image, whereas a standard true colour optical camera has SF images. This leads to a difference in appearance and scale and different types of imaging artifacts can occur in either modality. Both SF and UWF images were selected by the ophthalmologist (HT) based on the clarity of the lesion within their respective imaging ranges, making them suitable for diagnosis. Consequently, the patients in the SF images differed from those in the UWF images. The background of each group of patients is presented in . SD V.1.4, a large pretrained text-to-image foundation model, was fine-tuned on 6285 UWF retinal images captured with an Optos 200Tx UWF camera (Nikon, Japan) using DreamBooth to produce novel, synthetic UWF images. The dataset consisted of 1666 DR images, 215 AMD images, 1316 Gla images, 393 RVO images, 468 RD images and 2227 normal images. (The images used in the evaluation test are not included.) All images were selected based on diagnoses agreed upon by two or more ophthalmologists: RS, HT and TY. We fine-tuned a separate SD model for each class and generated images until 100 suitable SD images were selected. The same ophthalmologists (RS, HT and TY) assessed the generated images. They selected those suitable for teaching based on the clarity of features unique to the target pathology without additional findings suggesting other diseases. The background of each image of a patient is presented in . The number of generated SD images and selection rate per class was 8000 images for DR with a rate of 1.3%, 1500 for AMD (6.7%), 1000 for Gla (10%), 500 for BRVO (20%), 1100 for RD (9.1%) and 3500 for normal (0.29%). We fine-tuned the SD model and generated the images on a Nvidia RTX3090TI 24GB GPU using Python V.3.1.0. and the PyTorch V.1.13.0 and Diffusers V.0.10.2 libraries. We have provided more details in the ‘SD optimisation’ section of the . Artificial synthetic images generated with our fine-tuned SD model are shown in . During the course, the student is shown an image and asked to select the class they think it belongs to . After responding, the correct answer is immediately displayed alongside some image annotations highlighting and explaining the image’s relevant parts . We generated 600 synthetic images with SD, 100 per class, and divided them into five sets that were used for teaching. Each of the five sets consists of 120 synthetic images, with 20 images of each of the six types. They are named from No. 1 to No. 5. Students completed all five sets in sequence from the No. 1 set to the No. 5 set unless their performance exceeded the accuracy of the worst-performing expert, in which case training would end early. To check that the SD model did not memorise any real images (verification to ensure that SD is not simply outputting the learning images as they are), we compared all 600 synthetic images with all 6285 real images used for fine-tuning. We used self-supervised copy detection to find the most similar pairs of real and synthetic images, which were manually inspected. None of these pairs were visually identical, suggesting that our SD model did not memorise images. We have provided more details in the ‘Evaluation of Similarity’ section in . As a further comparison, we evaluated a recently proposed, state-of-the-art AI model previously published by researchers specialising in machine learning, which has achieved excellent performance on internal and external test sets. This model was trained on a dataset of 5376 patients (8570 eyes, 13 026 images of Optos 200Tx) and achieved an area under the curve (AUC) of 0.9848 (±0.0004) for detecting disease in UWF images on the same evaluation tests provided to the students. The model produces a probabilistic output by default that needs to be mapped to discrete predictions. We used the ‘conservative’ decision threshold for detecting images with disease, as proposed by the original authors. If an image has been classified as showing disease, the disease with the highest probability is then taken as the model’s prediction. Statistical analysis was performed using JMP V.16.2.0 (SAS Institute, Cary, NC, USA). Paired t-tests were used for within-subject significance tests regarding learning effects. We conducted a non-parametric multiple comparison test (Steel-Dwass test) when comparing study times and the time required to answer each question on the evaluation test among groups of students. The retinal images used in this study were acquired during clinical practice at Tsukazaki Hospital, Himeji, Japan. We obtained express written consent from each patient for the research use of their data. Written explanations and consent were received from the students, and prior explanations and consent were obtained from the experts. Our research was conducted in accordance with the Declaration of Helsinki and approved by the ethics committee of Tsukazaki Hospital. The performance of all study participants and the AI model is reported in . The eight expert clinicians had an average accuracy of 91.1% (±4.2%) in UWF images. The AI model achieved an accuracy of 73.3%. The students completed the course in 53 min 1 s (±16 min 0 s) , and their average accuracy improved from 43.6% (±18.8%) to 74.1% (±9.3%) (p<0.0001 by paired t-test), with performance improving significantly across all subgroups . For the trainee orthoptists, prestudying performance increased with a year of study, as expected, but even fourth-year students saw their accuracy rise by over 15%. Interestingly, fourth-year students spent the least time studying, almost half of that of first-year students (59.4 min vs 32.0 min, p<0.0001 by Steel-Dwass test), yet spent more time per image during the UWF images evaluation test (9 s vs 7 s, p=0.0002 by Steel-Dwass test). On the second test with SF images, which was not part of the training course, students’ prestudying performance was similar to that on UWF images, with 42.7% (±18.5%). Accuracy improved to 68.7% (±11.5%) (p<0.0001 by paired t-test) after taking the test, with performance also increasing significantly in all subgroups . However, except for fourth-year students, the poststudying performance of the students was lower than for the UWF images. The experts achieved similar performance as before (92.8% (±6.8%)), which is expected as they encounter both types of imaging regularly in their work. However, the performance of the AI model, which had only been trained on UWF images, dropped dramatically to just 40%. The average similarity scores (ranging from 0 to 1, with 1 being the same image) between the original images and the generated SD images for each of the six types were as follows: RD 0.45 (±0.09), RVO 0.45 (±0.07), Gla 0.48 (±0.07), DR 0.45 (±0.09), AMD 0.49 (±0.07) and normal 0.47 (±0.08). The highest similarity score among all 600 generated images was 0.747 for a normal SD image, followed by 0.739 for another normal SD image, and tied for third place were DR and Gla SD images with a score of 0.763 . (Pairs of images with similarity scores from fifth to 10th place are also presented in ). Students’ performance increased significantly after taking the web-based training course and matched or exceeded the performance of a recent, state-of-the-art AI model. This is remarkable, as students studied for less than an hour on average, and the images used in the training course were entirely synthetic, AI-generated images made with a fine-tuned SD model. Another key finding of our work is that humans are far more robust to changes in the imaging device than AI models. Although students’ performance was 8.2 percentage points lower on average for the SF images, they did not experience the dramatic drop of 33.3 percentage points that the AI model experienced. This suggests that humans generalise their learning much better, highlighting the importance of keeping humans in the loop. The study of training techniques for improving medical imaging interpretation has been extensively examined. Reports indicate enhanced X-ray diagnostic techniques by technicians in medically underserved areas such as South Africa and Australia, as well as specialised diagnostic improvements for CT colonographic data, endometrial tumours and malignant skin tumours. This aligns with the educational benefits we obtained from our methods. Moreover, our educational effects demonstrated easy acquisition of AI diagnostic capabilities, suggesting that past efforts in image interpretation training for humans will remain beneficial in the coming era. Compared with past teaching methods, our current approach also excels in cost-effectiveness. The time required to improve a non-expert’s performance substantially is relatively short. In addition, a web-based, fully automated training course can be easily scaled and made accessible worldwide. While there are concerns that AI technology might take away jobs, there is a significant opportunity to use AI in a new way: to enhance human skills. Furthermore, AI-generated images can be shared more quickly as they do not belong to any particular patient, and thus, students can be exposed to significantly more. Ethical and data privacy issues exist when using actual patient images for medical training courses. Our approach of using synthetic images generated by SD avoids these problems and could be used in many future clinical education areas. Another potential concern is that the SD model could memorise individual images, thus not preserving patient privacy. However, our detailed examination using state-of-the-art methods for detecting memorised images could not find such cases, implying that this approach may preserve privacy. We found that non-experts could be trained to match or exceed the performance of the AI model in about an hour of training. However, experts still performed at least 10% better than students, and the AI model could complement humans. For example, students might sometimes miss cases that the AI model would have identified but also recognise if the model is making a mistake. Preliminary analysis suggests that there are indeed cases that are hard for humans but correctly scored by AI models and vice versa. The correlation between the mean accuracy of the students per image and whether the AI model scored the image correctly is relatively weak (Pearson r=0.1537; p=0.0936) . Future work could explore whether providing the students with the output of the AI model would further improve their performance and help close the performance gap with experienced experts. There are several limitations to this study. First, there is no direct comparison with the current, widely used medical education methods that rely on textbooks with few images and lengthy explanations. In other words, the results of this study do not necessarily negate current education methods. Second, the learners in this study are limited to medical personnel. Further investigation is needed to determine whether this approach works well for learners with no medical background. Third, the study does not examine the persistence of learners’ diagnostic capabilities. While AI’s diagnostic capabilities are fixed, it is easy to imagine from the theory of the forgetting curve that continuous learning would be necessary to maintain learners' diagnostic capabilities. On the other hand, further learning could lead to even more significant improvements in diagnostic capabilities, so future studies are needed to determine the frequency and type of additional learning required to maintain and improve learning abilities. In addition to that limitation, the low adoption rate of 0.29% for standard eye images is due to the need to exclude many borderline cases that are even slightly close to showing signs of disease to classify an image as usual, which lacks distinctive features. This process necessitated extra effort from ophthalmologists, indicating a point for improvement in this method in the future. Future work could further explore the potential for using AI to enhance clinical education and whether non-experts benefit from having a diagnostic AI model available for decision support. We hope that our human-centric AI work will help improve clinical education and the level of care in ophthalmology and other fields of medicine. 10.1136/bjo-2023-324923 online supplemental file 1 |
Medical Error: Using Storytelling and Reflection to Impact Resident Error Response Factors | 57c52ee1-905d-4dc8-bf77-27b86de878de | 11466310 | Family Medicine[mh] | By the end of this activity, learners will be able to: 1. Define medical error and integrate this topic into their understanding of the profession of medicine. 2. Identify ways that physicians cope and thrive after medical error. 3. Describe safety culture and identify ways that colleagues can help with error management and recovery. 4. Describe local policies and practices related to medical error. I am a healer, yet sometimes I do more harm than good. D. Hilfiker , “Facing Our Mistakes” The Kohn et al. landmark study in 2000 reported preventable medical errors in hospitals resulted in approximately 98,000 deaths across 33.6 million hospital admissions. More recent studies estimate that 440,000 people die in the United States each year due to preventable medical error. Negative outcomes related to error are evident: rising health care costs due to adverse events and the subsequent need for repeat tests, readmissions, increased length of stay, rising insurance premiums, and the American public's lack of trust in health care professionals and institutions. Other indirect costs that result from poor safety outcomes in the US include loss of income to patients, and sometimes-preventable intermediate or long-range disabilities or chronic health care challenges. Even though medical error is the third leading cause of death in the US, costing between $73.5 and $98 billion in quality adjusted life years, , and error experiences are common among residents, residents receive little formal training in error management and recovery. Maladaptive coping strategies appear frequently among learners , and practicing physicians, with impact on physician quality of life and patient care. Researchers have identified common physician trajectories after error, including the frequently reported phenomenon of second victimhood—fear of litigation, shame, self-blame, and guilt arising from acute awareness of human fallibility and its impact on patients. These consequences appear cumulative and build across one's tenure. Further, shame and embarrassment create barriers to disclosure, reducing opportunity for analysis and process improvement. Second victimhood recovery follows six stages: chaos and accident response, intrusive reflections, restoring personal integrity, enduring the inquisition, obtaining emotional first aid, and moving on (which can involve surviving in, thriving in, or dropping out of medicine). In contrast to maladaptive responses, speaking up about medical error is an important act that impacts patient safety, quality of patient care, and long-term error reduction as transparency improves. The act of disclosure can be healing for the physician as it provides the opportunity to connect with the patient in a meaningful way where care and compassion is expressed, and patient-centered care is the goal. Researchers have identified several factors associated with desired outcomes: error reporting, – disclosure, coping, , constructive change, and growth after error. More specifically, coping may be aided by professional counseling, discussing the error with trusted peers, and engaging in quality projects linked to the error. Coping is essential to safeguard the emotional well-being of physicians and to prevent burnout. Error management support and guidance from more experienced physicians impacts resident emotions and behaviors. Presumably, supporting and promoting resident exposure to medical error disclosure through engagement with more experienced physicians helps them to emulate constructive responses and to identify maladaptive ones. Previous curricular evaluations have demonstrated the value of simulated clinical scenarios to practice communication-based skills in a safe setting using a team approach. – Our curriculum contributes to the existing literature by exploring the dimension of storytelling. Storytelling creates the framework to engage residents in the discussion and practice of disclosure, and to integrate a challenging concept—personal fallibility—into their professional identity. To summarize, development of an approach to medical error is critical for personal and professional resilience and meaningful participation in quality improvement. , Curricula addressing error management and recovery are desired by learners , and can improve resident knowledge, skills, and abilities in this area. , , Mentor storytelling can be particularly effective as well as integrating error management and recovery into everyday activities. We organized key factors related to effective error management and physician growth after error using a logic model for developing health interventions called the predisposing, reinforcing, and enabling constructs in educational diagnosis and evaluation - policy, regulatory, and organizational constructs in educational and environmental development (PRECEDE-PROCEED) model. Predisposing factors included resident knowledge (awareness of mentor error, local policies and procedures, effective error disclosure steps, and related professional values) and beliefs (error recovery self-efficacy and attitude towards error). We also targeted enabling factors, specifically resident skills (error and cause identification, error disclosure, emotion management, and accessing support), as well as a reinforcing factor (talking among colleagues). The facilitator's guide includes additional details on the medical error response factors. The primary goal of our curriculum was to help residents appreciate the pervasiveness of medical error and practice productive error responses. Preexisting error curricula in our family medicine residency program included: (1) the Institute for Healthcare Improvement's patient safety module completed during resident orientation , ; (2) a longitudinal wellness program, Tending the Flame, with a session on medical error and a personal wellness plan (inclusive of coping strategies) development assignment ; and (3) a didactic session specifically on disclosure training (given once every 3 years). In addition, local rotation sites had policies related to medical error and disclosure although residents were not particularly aware of these before our curricular intervention. We used the AAMC Quality Improvement and Patient Safety Competencies and ACGME family medicine milestones to inform the development of the educational objectives and pedagogical techniques (i.e., facilitating reflection and role-modeling for affective or attitudinal targets like self-efficacy). Specific objectives are included in our facilitator's guide , which also describes the implementation for the three 60-minute resident sessions. Slides for each session are included in – , and the faculty survey and pre-/postmodule surveys are also included. Logistics We recruited family medicine residency faculty ( N = 7) in a medium-sized, urban, midwestern program with an e-mail and survey sent 1 week before session one. During weekly didactics we recruited PGY 1, 2, and 3 family medicine residents ( N = 30). We scheduled the three sessions to occur over a 5 months period between August 2022 and January 2023, in order to allow participants to have time to process the session content in between sessions. Session One We used mentor storytelling to demonstrate physician error, emotions/coping strategies, and the ways in which the physician's environment can impact error response (e.g., workload and error culture). Prior to the session, we identified a faculty volunteer to share a personal story of an error and the aftermath with our program recruitment and faculty survey . Once the faculty member was selected, we gave the following guidance to potential storytellers: (1) we believe the most impactful stories will come from program mentors and leadership; (2) the session is confidential and for learning; (3) the purpose of the story sharing is to reflect on ways that physicians react to medical error, amplifying strategies that result in good care for the patient and physician(s) involved; (4) stories do not have to be long, 5–10 minutes works well. The session began with the faculty story and then the faculty facilitator guided residents in reflection using several prompts . Next, the session facilitator presented a small amount of lecture material, facilitated small- and large-group discussions, and prompted timed self-reflection through writing. Session Two During the next session , we stimulated interest by discussing key professional values related to error and then had a large-group discussion of learner fears and barriers related to error. We introduced the concept of safety culture , and residents had the opportunity to reflect on strengths and weaknesses of the current culture of their practice. The faculty presenter engaged learners throughout the session using small- and large-group discussion, polling, and timed self-reflection through writing. Session Three During the third session , the faculty presenter led a review of local policies and procedures related to medical error, and residents practiced self-awareness, error disclosure, root cause reflection, and coping. The faculty member facilitated the review of sample error cases and also instructed students to write a letter to a colleague or future self in the wake of perceived error. Evaluation We emailed residents the premodule survey immediately prior to the first session. Residents were given time to complete the premodule survey immediately prior to the start of the session, but survey links remained active through the start of session three to allow residents who missed session one or two the opportunity to complete the survey. The postmodule survey was sent via email to residents at the end of session three, and residents were given time to complete this survey just after session three concluded. Anonymous pre- and postmodule survey responses were analyzed and compared. To measure knowledge and confidence we collated participant affirmative responses ( agree / strongly agree ) on a 5-point Likert scale (1 = strongly disagree , 5 = strongly agree ). We measured satisfaction by collating a percentage of residents selecting each rating of the emotional difficulty (0 = not emotionally difficult at all , 10 = extremely emotionally difficult ) and helpfulness (0 = not helpful at all, 10 = extremely helpful ) of the curriculum. Due to small sample size, analyses by PGY were not undertaken. The evaluation procedures were approved by the Wright State University Boonshoft School of Medicine Institutional Review Board (FWA# 00002427, approved August 26, 2022). We recruited family medicine residency faculty ( N = 7) in a medium-sized, urban, midwestern program with an e-mail and survey sent 1 week before session one. During weekly didactics we recruited PGY 1, 2, and 3 family medicine residents ( N = 30). We scheduled the three sessions to occur over a 5 months period between August 2022 and January 2023, in order to allow participants to have time to process the session content in between sessions. We used mentor storytelling to demonstrate physician error, emotions/coping strategies, and the ways in which the physician's environment can impact error response (e.g., workload and error culture). Prior to the session, we identified a faculty volunteer to share a personal story of an error and the aftermath with our program recruitment and faculty survey . Once the faculty member was selected, we gave the following guidance to potential storytellers: (1) we believe the most impactful stories will come from program mentors and leadership; (2) the session is confidential and for learning; (3) the purpose of the story sharing is to reflect on ways that physicians react to medical error, amplifying strategies that result in good care for the patient and physician(s) involved; (4) stories do not have to be long, 5–10 minutes works well. The session began with the faculty story and then the faculty facilitator guided residents in reflection using several prompts . Next, the session facilitator presented a small amount of lecture material, facilitated small- and large-group discussions, and prompted timed self-reflection through writing. During the next session , we stimulated interest by discussing key professional values related to error and then had a large-group discussion of learner fears and barriers related to error. We introduced the concept of safety culture , and residents had the opportunity to reflect on strengths and weaknesses of the current culture of their practice. The faculty presenter engaged learners throughout the session using small- and large-group discussion, polling, and timed self-reflection through writing. During the third session , the faculty presenter led a review of local policies and procedures related to medical error, and residents practiced self-awareness, error disclosure, root cause reflection, and coping. The faculty member facilitated the review of sample error cases and also instructed students to write a letter to a colleague or future self in the wake of perceived error. We emailed residents the premodule survey immediately prior to the first session. Residents were given time to complete the premodule survey immediately prior to the start of the session, but survey links remained active through the start of session three to allow residents who missed session one or two the opportunity to complete the survey. The postmodule survey was sent via email to residents at the end of session three, and residents were given time to complete this survey just after session three concluded. Anonymous pre- and postmodule survey responses were analyzed and compared. To measure knowledge and confidence we collated participant affirmative responses ( agree / strongly agree ) on a 5-point Likert scale (1 = strongly disagree , 5 = strongly agree ). We measured satisfaction by collating a percentage of residents selecting each rating of the emotional difficulty (0 = not emotionally difficult at all , 10 = extremely emotionally difficult ) and helpfulness (0 = not helpful at all, 10 = extremely helpful ) of the curriculum. Due to small sample size, analyses by PGY were not undertaken. The evaluation procedures were approved by the Wright State University Boonshoft School of Medicine Institutional Review Board (FWA# 00002427, approved August 26, 2022). Of the cohort of 30 PGY 1, 2, and 3 participating family medicine residents, 22 (73%) completed the premodule survey, and 15 (50%) completed the postmodule survey. Seven pre- and postmodule surveys were able to be matched using anonymous codes. Of those who completed the postmodule survey, nine completed all three sessions, five completed two sessions, and one completed one session. Additionally, seven out of seven faculty completed the faculty survey. State of Preexisting Environment Premodule responses demonstrated the pervasiveness of error experience among residents. Most residents reported having experienced error (55%, n = 12) and having had a mentor or peer share an error story with them (73%, n = 16). Among those with an error experience, 10 (77%) reported the error was disclosed to the patient. Many residents with error experience reported that their team or organization learned from the error (75%, n = 9), acknowledged the error (92%, n = 11), and debriefed as a team (83%, n = 10). Knowledge and Confidence Our curriculum was associated with an increase in residents who reported several target factors: disclosure confidence/self-efficacy, knowledge of local procedures, accessing support as an error response, faculty and peer story sharing, and acknowledgement that mentors have made errors. Specifically, disclosure self-efficacy ( I can be honest about the errors I make as a doctor. ) increased after the curriculum from 86% ( n = 19) to 93% ( n = 14; ). As expected, faculty reported higher levels of confidence compared to residents with error disclosure and personal relationship recovery . Knowledge of related local procedures ( I know what to do at my institution when faced with a medical error. ) increased from 46% ( n = 10) to 93% ( n = 14; ). The curriculum was also associated with an increase in reaching out to others as an error response from 36% ( n = 8) to 87% ( n = 13). Debriefing with the team remained common, but the rate of residents reporting feel bad about myself as an error response increased from 41% ( n = 9) to 60% ( n = 9). After the curriculum, rates of reported faculty and peer story sharing increased, and resident reported awareness of mentor error increased from 68% ( n = 15) to 87% ( n = 13). Incidentally, all faculty respondents reported I have made errors in my care for patients . Responses also demonstrated an increase in resident reported self-awareness ( I acknowledge when I am at increased risk for making errors ) from 77% ( n = 17) to 93% ( n = 14). Satisfaction Overall residents reported the training was helpful (ranking > 5), and six residents (40%) reported an emotionally difficult rating of 5 or greater for the curriculum . Prior to the module, residents were most interested in further training through personal stories of mentor error (73%, n = 16), and after the curriculum, residents reported most interest in additional training in legal and malpractice risk (73%, n = 11). Premodule responses demonstrated the pervasiveness of error experience among residents. Most residents reported having experienced error (55%, n = 12) and having had a mentor or peer share an error story with them (73%, n = 16). Among those with an error experience, 10 (77%) reported the error was disclosed to the patient. Many residents with error experience reported that their team or organization learned from the error (75%, n = 9), acknowledged the error (92%, n = 11), and debriefed as a team (83%, n = 10). Our curriculum was associated with an increase in residents who reported several target factors: disclosure confidence/self-efficacy, knowledge of local procedures, accessing support as an error response, faculty and peer story sharing, and acknowledgement that mentors have made errors. Specifically, disclosure self-efficacy ( I can be honest about the errors I make as a doctor. ) increased after the curriculum from 86% ( n = 19) to 93% ( n = 14; ). As expected, faculty reported higher levels of confidence compared to residents with error disclosure and personal relationship recovery . Knowledge of related local procedures ( I know what to do at my institution when faced with a medical error. ) increased from 46% ( n = 10) to 93% ( n = 14; ). The curriculum was also associated with an increase in reaching out to others as an error response from 36% ( n = 8) to 87% ( n = 13). Debriefing with the team remained common, but the rate of residents reporting feel bad about myself as an error response increased from 41% ( n = 9) to 60% ( n = 9). After the curriculum, rates of reported faculty and peer story sharing increased, and resident reported awareness of mentor error increased from 68% ( n = 15) to 87% ( n = 13). Incidentally, all faculty respondents reported I have made errors in my care for patients . Responses also demonstrated an increase in resident reported self-awareness ( I acknowledge when I am at increased risk for making errors ) from 77% ( n = 17) to 93% ( n = 14). Overall residents reported the training was helpful (ranking > 5), and six residents (40%) reported an emotionally difficult rating of 5 or greater for the curriculum . Prior to the module, residents were most interested in further training through personal stories of mentor error (73%, n = 16), and after the curriculum, residents reported most interest in additional training in legal and malpractice risk (73%, n = 11). To address the lack of a graduate medical education curriculum related to medical error response, we developed three sessions for family medicine residents. We found the use of a model for factor organization and assessment of the curriculum as a necessary initial step. From this evaluation we confirmed that most residents already had prior experience with team or personal error, our curricular intervention positively impacted specific targets, and residents found the curriculum helpful. However, sample size and the format and timing of our postmodule survey limited our ability to fully assess our curriculum. We plan to refine the curriculum based on initial results by adding several new evaluation techniques and offering the module more broadly to other graduate medical education specialty programs in our community. Because of the complexity of physician error response, we found the process of organizing predisposing, enabling, and reinforcing factors from the literature helpful for identifying potential curricular targets. Similarly, our exploration of the preexisting curriculum allowed us to select targets not already being fully addressed. Our residency, like many others, has a formal and informal curriculum (i.e., unspoken norms and lessons that residents learn) for medical error, and identifying this curriculum helped us partner with current faculty champions to develop a synergistic curriculum. Our curriculum evaluation suggests that most residents had prior experience with team or personal error, confirming the importance of the topic, and suggests that a brief curriculum can be effective at impacting important error response factors. Residents reported significant interest in error training, and many found our curriculum helpful. The interest in further training shifting from personal stories of mentor error (premodule) to legal and malpractice concerns (postmodule) could indicate a deficiency in the curriculum or a natural change in focus. From an experiential standpoint, the storytelling by faculty prompted residents to use empathic language to explore what they would have felt and done in the storyteller's shoes and voice the respect they had for the physician doing the right thing. After initial rumination regarding the error itself, residents shared surprising insights into error management and recovery, including a reframing of the error and a thankfulness for physician colleagues that can be called upon to help deal with complications of error. Based on our experience in this curriculum, residents can be remarkably open, especially when openness is modeled by mentors. Surprisingly, the rate of residents who report feeling bad about oneself after error increased from 41% ( n = 9) to 60% ( n = 9) after the curriculum. Perhaps the act of addressing this subject as a group was emotionally charging. The fact that 40% of residents reported an emotionally difficult rating of 5 or greater highlights the importance of including resident support resources during the sessions themselves. We are unsure if this finding will persist in larger sample sizes. Time was a barrier to accomplishing delivery of all components of the curriculum, and we hope to refine or remove less important components. For example, all residents (pre- and postmodule) and faculty reported that good doctors should be honest about errors they make, suggesting this belief may be a less important target for intervention. The time spent on discussion of values and underlying medical ethics could be used to give more time for other components. The literature has shown a disconnect between belief in the appropriateness of disclosure, intent to disclose, and actual disclosure rates. Therefore, resident and patient-oriented outcomes are an important next step in curriculum assessment (e.g., resident milestones data, error reporting rates at primary rotation sites, and institutional patient safety survey information). Limitations of our evaluation stemmed from an ambitious desire to develop a comprehensive curriculum, limited sample size and time, lack of direct skills and knowledge assessments, and missing team and patient-oriented outcomes. As such, our curriculum and evaluation do not allow for specific module content to be associated with changes in resident responses. Similarly, the complexity of error response and the scope and time for our curriculum evaluation limited our ability to associate our curriculum with specific changes in rates of error acknowledgement, disclosure, coping, and growth, or patient-oriented outcomes like satisfaction with error disclosure or care relationship after error. Our small sample size limits generalizability, and we hope to offer a refined curriculum to our entire graduate medical education community, which will help with sample size limitations. Our timeframe allowed only for short-term reassessment of survey responses, which could differ from long-term impacts. Future studies could use resident milestone data to assess measurable outcomes, and a 6-month postsurvey could be undertaken to see if changes in knowledge, skills, and beliefs persist. Organizational data like error reporting rates and patient safety survey data could be tracked by year to evaluate potential impacts organization-wide. The project also uncovered a need for faculty development in medical error response, and sessions may be adapted for this purpose. Our brief curriculum was associated with an increase in related resident reported knowledge, confidence, and story sharing. The medical community will benefit from further refining the model for error management and growth behaviors among residents. It will also be important to move from short-term, self-reported resident knowledge and beliefs to long-term resident and patient-oriented outcomes like error reporting rate, disclosure skills assessment by faculty and patient, and overall error rates. We intend to improve our postmodule assessment by adding direct skills assessment for error disclosure and incorporating resident and patient-oriented outcomes like reporting rates and patient safety survey data. Increasing our sample size by offering the curriculum to our local graduate medical education community including numerous specialty programs will allow us to better compare pre- and postmodule results. Ultimately, development of interdisciplinary and interprofessional error response training will best prepare learners to manage their future errors and their personal recovery. Facilitators Guide.docx Error Session 1.pptx Error Session 1 Handout.pdf Error Session 2.pptx Error Session 3.pptx Error Session 3 Handout - Error Cases.docx Faculty Survey.docx Premodule Resident Survey.docx Postmodule Resident Survey.docx All appendices are peer reviewed as integral parts of the Original Publication. |
Evaluation of Whatman FTA cards for the preservation of yellow fever virus RNA for use in molecular diagnostics | 8ad28606-b8d0-44b4-8734-ae71f69d1cc7 | 9200311 | Pathology[mh] | Yellow fever virus (YFV) is a flavivirus endemic to 47 countries in sub-Saharan Africa, Central America, and South America. The virus is responsible for periodic outbreaks of yellow fever (YF) disease, a hemorrhagic disease with a case fatality rate of 30–60% in severe cases . Each year there is an estimated 200,000 cases of YF and 30,000 YF-associated deaths . As the majority of YF cases are asymptomatic, the prevalence of YFV is considered to be much higher than annual incidence rates suggest. Thus, the WHO classifies a single confirmed case of YF as an outbreak, requiring immediate ring vaccination with the live-attenuated vaccine strain, 17D . Although vaccination against YF is part of routine vaccination programs in many YF endemic countries, there has been an emergence of YFV in urban areas with large, unvaccinated populations. The increased risk of YF outbreaks and limited YF vaccine supplies underscores the need for accurate YFV diagnostics. False negatives could lead to larger outbreaks and false positives could lead to unplanned vaccine usage. Diagnosing YF can be difficult as patients present with non-specific, flu-like symptoms that can be confused with leptospirosis, malaria, and other viral hemorrhagic fevers during differential diagnosis. As such, laboratory diagnosis is required to confirm YF. Both serological methods (YFV IgM or IgG assays followed by plaque-reduction neutralization tests) and/or detection of YFV-specific RNA are included in the WHO algorithm for laboratory diagnosis of YF . Although these assays are successfully used to accurately detect YF, there are important caveats that must be addressed. Serological diagnosis of YF is complicated by the well characterized flavivirus cross-reactive antibody response . Infection with other flaviviruses such as dengue (1–4), Zika, and West Nile viruses must be ruled out by additional serological testing and confirmed at regional reference laboratories. Additionally, IgM generated in response to a wild-type (WT) YFV infection is indistinguishable from IgM generated after vaccination, including in rare cases of vaccine-associated viscerotropic disease, making serological diagnosis during mass vaccination campaigns difficult . Although YF molecular diagnostic assays can differentiate between WT and rare adverse events associated with 17D vaccination , samples are required to be transported on wet ice if arriving within one day and at -20°C if arriving after more than one day. This cold-chain is required to prevent the degradation of YFV RNA but is often not realistic in YF-endemic regions where transmission season is accompanied by hot and humid weather . Furthermore, viremia during YFV infection is transient and highly dependent on when the sample is collected, resulting in some YF clinical samples having very low levels of viral RNA. The use of serological or molecular methods for YF diagnosis is determined by the timing of specimen collection (i.e., acute vs. non-acute) and laboratory capacity of the region. South American countries routinely use the ‘YFall” primers for molecular diagnosis while African countries prefer plaque reduction neutralization tests (PRNTs) and the CDC Arboviral MAC ELISA for serological testing. With the recent WHO evaluation of a qRT-PCR kit for the molecular diagnosis of YFV , it is likely that molecular diagnosis will become more routine in Africa, underscoring the need for RNA stabilization as a critical factor for the implementation of YF molecular diagnostics in new regions. Whatman FTA cards have been shown to be effective in the molecular diagnosis of multiple viral and bacterial diseases . The current WHO diagnostic algorithm calls for the molecular diagnosis of YF from serum samples, which are collected at clinics and diagnosis is completed at regional reference labs . Herein, we show how incorporating FTA cards into YFV molecular diagnostics has the potential to improve YFV surveillance in remote, endemic regions. Viruses The vaccine sub-strain 17D-204 [titer: 7 log 10 plaque forming units (pfu)/ml] and WT strain Asibi (titer: 6 log 10 pfu/mL) used in this study were obtained from the CDC Arboviral Disease Branch, Arbovirus Reference Collection. Experiments involving 17D-204 virus were performed according to biosafety level 2 requirements and experiments involving Asibi virus were performed according to biosafety level 3 safety requirements. Serial dilution of virus was performed in PBS (Gibco, Massachusetts, USA) or flavivirus-negative human serum (EMD Millipore, Massachusetts, USA). FTA card protocol, RNA extraction and genome amplification Contrived specimens were spotted onto FTA micro cards (WHAWB120205, Sigma-Aldrich, St. Louis, USA), allowed to dry and RNA extracted from discs punched from the card according to the manufacturers protocol (i.e., 140 μL of sample distributed evenly over the sample area). In order to minimize cross-contamination, cards were also pre-punched with a sterilized, 6 mm hole punch prior to inoculation. Punches were inoculated with 10 μL of contrived specimen and allowed to dry for one hour at room temperature in the biosafety cabinet. Individual punches were then placed into 1.5 mL tubes and stored as described until RNA extraction was performed using the QIAmp viral RNA extraction kit (Qiagen, California, USA). To extract RNA, individual FTA punches were placed directly into 140 μL of PBS, thoroughly vortexed, added to 560 μL buffer AVL and allowed to incubate for 10 minutes at room temperature. When pooling RNA from two punches, individual punches were placed into 70 μL of PBS and vortexed. The extracts in PBS were then combined (i.e., 140 μL total) and added to 560 μL of AVL. The remaining steps of the QIAmp protocol were then followed as previously described . Final elution was performed using 60 μL of Buffer AVE. RNA from the same preparation of YFV spiked serum was simultaneously extracted according to the manufacturer’s protocol (i.e., sample placed directly into buffer AVL). Viral RNA was detected using the Quantitect probe RT-PCR kit (Qiagen, California, USA) with YFall primers as previously described in a total reaction volume of 25 μL. All samples were tested in triplicate, using 10 μL of RNA. While using high input elute may result in inclusion of more inhibitors in the reaction, inhibition was not observed in our limited analysis using normal human serum. Inactivation of YFV after FTA inoculation Confirmation of viral inactivation was performed using institutionally approved protocols. The vaccine strain, 17D-204 virus was used as a surrogate for all YFVs in order to reduce biosafety concerns. FTA punches were inoculated with 10 μl of 17D-204 virus (5 log 10 pfu/ml), allowed to dry, and were rehydrated in 100 μL sterile water. The resulting FTA extract was used in two assays to confirm inactivation. First, 100 μL of FTA extract was inoculated onto Vero cells and allowed to incubate for 1 hour at room temperature after which DMEM media was added, and the flask placed at 37°C/5% CO 2 and allowed to incubate for one week. Cells were monitored for cytopathic effect (CPE) which occurs with vaccine strains of YFV. After one week, 100 μL of media from the week one flask was transferred onto fresh Vero cells. This process was repeated for three weeks. Secondly, the FTA extract was used in a viral plaque assay. Ten-fold serial dilutions (1:10–1:1,000,000) of FTA extract were prepared in Bovine albumin-1 media (BA-1) and used to perform a plaque assay as previously described . Role of environmental conditions on RNA stability For stability experiments, 17D-204 was diluted in PBS or human, flavivirus-negative serum and inoculated onto punches at concentrations of 7 log 10 pfu/mL (100,000 pfu/punch), 6 log 10 pfu/mL (10,000 pfu/punch), 5 log 10 pfu/mL (1,000 pfu/punch), 4 log 10 pfu/mL (100 pfu/punch), 3 log 10 pfu/mL (10 pfu/punch) and 2 log 10 pfu/mL (1 pfu/punch). The ability of FTA cards to stabilize YFV RNA at an elevated temperature was measured by holding FTA punches with variable titers of YFV at 37°C for two weeks and comparing RNA positivity to control punches held at room temperature (~25°C) for the same amount of time. Samples were testing in triplicate and experiments were conducted in duplicate, leading to n = 6 for all groups. Punches in the 37°C group were placed directly into heat blocks after inoculation to ensure that any effect of temperature on the drying process was captured in the experiment. In experiments utilizing a single punch, RNA was extracted from one punch daily for seven days and at 14 days post-inoculation. For studies using pooled RNA from two punches, only 10 pfu/punch and 1 pfu/punch were utilized. Once a punch no longer tested positive for YFV RNA, punches from that dilution were removed from collection on subsequent days. Laboratories in YF-endemic areas are often subject to high humidity. To assay the effect of humidity on the utility of the FTA cards, a two-part experiment was conducted using the one punch protocol described above. The first part sought to determine whether exposure of cards to high humidity post inoculation affected their ability to stabilize YF RNA. FTA cards were inoculated with 10 pfu/punch or 1 pfu/punch of 17–204 virus and incubated at high humidity (80–85%). RNA was extracted from one punch daily for seven days. Humidity readings were taken at each time point using a humidity and temperature pen (Fisher Brand) to ensure no large fluctuations in humidity occurred due to opening and closing of the incubator door. The second evaluation sought to determine whether exposure of cards to high humidity prior to inoculation affected their ability to stabilize YF RNA. FTA cards were incubated at 37°C and high humidity (80–85%) for one, two, and three days prior to inoculation cards were removed from the incubator, immediately punched, and inoculated with 10 pfu/mL or 1 pfu/mL of 17–204 virus. Punches were returned immediately to the incubator so that drying would occur at high humidity. Finally, the effect of desiccating cards that were previously exposed to humid conditions was assayed to establish if dry FTA cards were required to stabilize RNA after inoculation. FTA cards were placed in an incubator at 37°C for three days at 75–80% humidity, then were transferred into plastic bags containing one or two silica 1-gram gel desiccation packets (Dry & dry, California, USA) and sealed. A card placed in a bag with no desiccation packets was also included as a control at each timepoint. Bags were returned to the incubator for one, two or seven days prior to inoculation when FTA cards were punched and inoculated with 10 pfu/punch or 1 pfu/punch of 17D-204 virus. The punches were returned immediately to the bag containing the specified number of desiccant packets and placed into the incubator. RNA was extracted from punches and assayed for YFV RNA positivity every day for seven days. The experiment was repeated with a different brand of desiccation packet (Whatman FTA, 1-gram packets) to compare the efficacy of 1-gram desiccation packets. Statistics Limit of detection (LOD) and time to negative result were calculated by linear regression modeling, estimating when the mean C t value reached the cut-off threshold of 37. The goodness of fit for these regressions was calculated using R 2 values. For high titer samples, the mean C t value never reached 37 before the end of the time course, so the time to detection loss was extrapolated based on the linear regression model. Significant differences in LOD and time to negative result were determined by calculating the likelihood that the 95% confidence intervals of C t value 37 for two experiments (95% CI) would overlap as described previously . In cases where the time to negative result was estimated due to limitations in data, a comparison of 95% CIs was not completed as the mean number of days to a C t of 37 was extrapolated. The vaccine sub-strain 17D-204 [titer: 7 log 10 plaque forming units (pfu)/ml] and WT strain Asibi (titer: 6 log 10 pfu/mL) used in this study were obtained from the CDC Arboviral Disease Branch, Arbovirus Reference Collection. Experiments involving 17D-204 virus were performed according to biosafety level 2 requirements and experiments involving Asibi virus were performed according to biosafety level 3 safety requirements. Serial dilution of virus was performed in PBS (Gibco, Massachusetts, USA) or flavivirus-negative human serum (EMD Millipore, Massachusetts, USA). Contrived specimens were spotted onto FTA micro cards (WHAWB120205, Sigma-Aldrich, St. Louis, USA), allowed to dry and RNA extracted from discs punched from the card according to the manufacturers protocol (i.e., 140 μL of sample distributed evenly over the sample area). In order to minimize cross-contamination, cards were also pre-punched with a sterilized, 6 mm hole punch prior to inoculation. Punches were inoculated with 10 μL of contrived specimen and allowed to dry for one hour at room temperature in the biosafety cabinet. Individual punches were then placed into 1.5 mL tubes and stored as described until RNA extraction was performed using the QIAmp viral RNA extraction kit (Qiagen, California, USA). To extract RNA, individual FTA punches were placed directly into 140 μL of PBS, thoroughly vortexed, added to 560 μL buffer AVL and allowed to incubate for 10 minutes at room temperature. When pooling RNA from two punches, individual punches were placed into 70 μL of PBS and vortexed. The extracts in PBS were then combined (i.e., 140 μL total) and added to 560 μL of AVL. The remaining steps of the QIAmp protocol were then followed as previously described . Final elution was performed using 60 μL of Buffer AVE. RNA from the same preparation of YFV spiked serum was simultaneously extracted according to the manufacturer’s protocol (i.e., sample placed directly into buffer AVL). Viral RNA was detected using the Quantitect probe RT-PCR kit (Qiagen, California, USA) with YFall primers as previously described in a total reaction volume of 25 μL. All samples were tested in triplicate, using 10 μL of RNA. While using high input elute may result in inclusion of more inhibitors in the reaction, inhibition was not observed in our limited analysis using normal human serum. Confirmation of viral inactivation was performed using institutionally approved protocols. The vaccine strain, 17D-204 virus was used as a surrogate for all YFVs in order to reduce biosafety concerns. FTA punches were inoculated with 10 μl of 17D-204 virus (5 log 10 pfu/ml), allowed to dry, and were rehydrated in 100 μL sterile water. The resulting FTA extract was used in two assays to confirm inactivation. First, 100 μL of FTA extract was inoculated onto Vero cells and allowed to incubate for 1 hour at room temperature after which DMEM media was added, and the flask placed at 37°C/5% CO 2 and allowed to incubate for one week. Cells were monitored for cytopathic effect (CPE) which occurs with vaccine strains of YFV. After one week, 100 μL of media from the week one flask was transferred onto fresh Vero cells. This process was repeated for three weeks. Secondly, the FTA extract was used in a viral plaque assay. Ten-fold serial dilutions (1:10–1:1,000,000) of FTA extract were prepared in Bovine albumin-1 media (BA-1) and used to perform a plaque assay as previously described . For stability experiments, 17D-204 was diluted in PBS or human, flavivirus-negative serum and inoculated onto punches at concentrations of 7 log 10 pfu/mL (100,000 pfu/punch), 6 log 10 pfu/mL (10,000 pfu/punch), 5 log 10 pfu/mL (1,000 pfu/punch), 4 log 10 pfu/mL (100 pfu/punch), 3 log 10 pfu/mL (10 pfu/punch) and 2 log 10 pfu/mL (1 pfu/punch). The ability of FTA cards to stabilize YFV RNA at an elevated temperature was measured by holding FTA punches with variable titers of YFV at 37°C for two weeks and comparing RNA positivity to control punches held at room temperature (~25°C) for the same amount of time. Samples were testing in triplicate and experiments were conducted in duplicate, leading to n = 6 for all groups. Punches in the 37°C group were placed directly into heat blocks after inoculation to ensure that any effect of temperature on the drying process was captured in the experiment. In experiments utilizing a single punch, RNA was extracted from one punch daily for seven days and at 14 days post-inoculation. For studies using pooled RNA from two punches, only 10 pfu/punch and 1 pfu/punch were utilized. Once a punch no longer tested positive for YFV RNA, punches from that dilution were removed from collection on subsequent days. Laboratories in YF-endemic areas are often subject to high humidity. To assay the effect of humidity on the utility of the FTA cards, a two-part experiment was conducted using the one punch protocol described above. The first part sought to determine whether exposure of cards to high humidity post inoculation affected their ability to stabilize YF RNA. FTA cards were inoculated with 10 pfu/punch or 1 pfu/punch of 17–204 virus and incubated at high humidity (80–85%). RNA was extracted from one punch daily for seven days. Humidity readings were taken at each time point using a humidity and temperature pen (Fisher Brand) to ensure no large fluctuations in humidity occurred due to opening and closing of the incubator door. The second evaluation sought to determine whether exposure of cards to high humidity prior to inoculation affected their ability to stabilize YF RNA. FTA cards were incubated at 37°C and high humidity (80–85%) for one, two, and three days prior to inoculation cards were removed from the incubator, immediately punched, and inoculated with 10 pfu/mL or 1 pfu/mL of 17–204 virus. Punches were returned immediately to the incubator so that drying would occur at high humidity. Finally, the effect of desiccating cards that were previously exposed to humid conditions was assayed to establish if dry FTA cards were required to stabilize RNA after inoculation. FTA cards were placed in an incubator at 37°C for three days at 75–80% humidity, then were transferred into plastic bags containing one or two silica 1-gram gel desiccation packets (Dry & dry, California, USA) and sealed. A card placed in a bag with no desiccation packets was also included as a control at each timepoint. Bags were returned to the incubator for one, two or seven days prior to inoculation when FTA cards were punched and inoculated with 10 pfu/punch or 1 pfu/punch of 17D-204 virus. The punches were returned immediately to the bag containing the specified number of desiccant packets and placed into the incubator. RNA was extracted from punches and assayed for YFV RNA positivity every day for seven days. The experiment was repeated with a different brand of desiccation packet (Whatman FTA, 1-gram packets) to compare the efficacy of 1-gram desiccation packets. Limit of detection (LOD) and time to negative result were calculated by linear regression modeling, estimating when the mean C t value reached the cut-off threshold of 37. The goodness of fit for these regressions was calculated using R 2 values. For high titer samples, the mean C t value never reached 37 before the end of the time course, so the time to detection loss was extrapolated based on the linear regression model. Significant differences in LOD and time to negative result were determined by calculating the likelihood that the 95% confidence intervals of C t value 37 for two experiments (95% CI) would overlap as described previously . In cases where the time to negative result was estimated due to limitations in data, a comparison of 95% CIs was not completed as the mean number of days to a C t of 37 was extrapolated. Inoculation of YFV onto FTA cards results in viral inactivation The vaccine strain of YFV, 17D-204, was confirmed to be inactivated after storage on FTA cards. No CPE was observed in Vero cells inoculated with 17D-204 FTA punch extract after three blind passages of 1 week each. Additionally, no plaques were observed after a plaque assay was performed on 17D-204 FTA punch extract. RNA from vaccine and WT strains of YFV can be detected after inoculation onto FTA cards To determine if vaccine and WT strains of YFV could be detected after inoculation onto FTA cards, the vaccine substrain 17D-204, WT strain Asibi, and PBS as a negative control were inoculated onto FTA cards and assayed for YFV RNA. The viruses were inoculated at the same titer (6log 10 pfu/mL). These C t values were compared to a standard extraction (140 μL of sample directly extracted) of both 17D-204 and Asibi. Although the mean C t values were lower when RNA was extracted from an FTA punch, both strain 17D-204 (mean C t value: 23.2 ± 0.12) and strain Asibi (mean C t value: 23.9 ± 0.1) were readily detected from FTA punches whereas the PBS control was negative for YFV RNA (mean Ct value: 43.3 ± 1.9) . The low volume applied to FTA cards decreases assay sensitivity Typically, when extracting viral RNA from a YFV clinical sample, 140 μL of serum is used as input in the extraction protocol. Although this volume can be applied to the entire FTA card, RNA extraction protocols utilizing FTA cards require the card to be punched prior to RNA extraction, limiting the volume of sample that is extracted. The timing of punching, pre-inoculation or post-inoculation was tested to ensure that the ‘pre-punch protocol’ did not impact RNA yield. There was no significant difference in C t values when sample was applied to the whole card and punched later (“post-inoculation punch”, R 2 = 0.99) and when sample was applied to pre-punched cards (“pre-inoculation punch”, R 2 = 0.98) (pre-inoculation punch vs post-inoculation punch 95% CI: -9.3, 10.97, ). To test if this reduction of input volume significantly impacts the sensitivity of the assay, the limit of detection of 10 μL virus applied to FTA punches was assayed and compared to a direct extraction from 10 μL of virus and the ‘gold standard’ of a direct extraction of 140 μL of virus. The LOD of 140 μL of virus (0.22 pfu, R 2 = 0.98) was shown to be significantly lower (140 μL virus vs one FTA punch 95% CI: 0.09, 1.4) than the LOD of one FTA punch (3.01 pfu, R 2 = 0.97); however, there was no significant difference (one FTA punch vs. 10 μL 95% CI: -0.2, 0.4) in the LOD of RNA extracted from one FTA punch compared to 10 μL of virus extracted directly (2.2 pfu, R 2 = 0.97) . To determine if extracting RNA from multiple punches increased sensitivity, RNA was pooled from two punches inoculated with virus diluted in PBS . Increasing the number of punches utilized, significantly lowered (two FTA punch vs one FTA punch 95% CI: 0.4, 1.0) the LOD (0.6 pfu, R 2 = 0.98) when compared to extracting RNA from one punch. This trend continued when RNA was extracted from three and four punches . Because serum is the most common sample used for molecular diagnoses, the experiment was repeated using human, flavivirus-negative serum (EMD Millipore, Burlington, MA) as the diluent. As was observed in the experiment utilizing PBS as the diluent, the LOD of 140 μL of virus diluted in human serum (0.5 pfu, R 2 = 0.96) was significantly lower (140 μL vs FTA punch 95% CI: 0.8, 1.5) than the LOD of an FTA punch inoculated with 10 μL virus diluted in human serum (2.7 pfu, R 2 = 0.95), but the LOD of an FTA punch inoculated with 10 μL virus diluted in serum was not significantly different (10 μL serum vs 10 μL PBS 95% CI: -0.3, 0.4) than a FTA punch inoculated with 10 μL virus diluted in PBS . Heat decreased the sensitivity of YFV molecular diagnostics of FTA cards At room temperature, 100,000 pfu/punch (R 2 = 0.68), 10,000 pfu/punch (R 2 = 0.80), 1,000 pfu/punch (R 2 = 0.64) and 100 pfu/punch (R 2 = 0.82) could be detected for over two weeks . YFV inoculated at 10 pfu/punch could be detected for 6.2 days (R 2 = 0.88) and 1 pfu/punch for 3.5 days (R 2 = 0.74) . At 37°C, 100,000 pfu/punch (R 2 = 0.78) and 10,000 pfu/punch (R 2 = 0.73) were still detectable for more than 14 days; however, the 1,000 pfu/punch sample (R 2 = 0.64) was detectable for 13.7 days . Similarly, the duration of detectability for the 100 pfu/punch (10.1 days, R 2 = 0.70), 10 pfu/punch (4.1 days, R 2 = 0.77) and 1 pfu/punch (2.2 days, R 2 = 0.72) samples decreased at 37°C . Using 95% confidence intervals to compare the number of days YFV was detectable by qRT-PCR, 10 pfu/punch (RT vs 37°C 95% CI: 1.52, 2.77) and 1 pfu/punch (RT vs 37°C 95% CI: 0.7, 2.0) became undetectable in a significantly shorter time at 37°C than at RT. To test if RNA extraction from multiple FTA punches improved the sensitivity of YFV molecular detection at high temperatures, RNA was pooled from two punches at 10 pfu/punch and 1 pfu/punch. It was shown that extracting RNA from two punches significantly increased the number of days RNA could be detected . High humidity negatively impacts RNA stability on FTA cards The initial experiments were conducted in a controlled climate (40–55% humidity). To test if these humidity conditions positively influenced RNA stability on the FTA cards, cards were inoculated with 10 pfu/punch and 1 pfu/punch of YFV and placed into a humidity chamber (80–85%) for one week. Control samples were left in the biosafety cabinet at ambient humidity for one week. Cards incubated at high humidity showed no difference in the time limit for detection of positive C t values when compared to dry cards incubated at low humidity . Despite this, the negative effect of humidity on the ability of FTA cards to stabilize RNA has been frequently documented in regions endemic to YFV where high humidity is common . To assay if the reported loss in RNA yield was due to the card having absorbed moisture from the air prior to inoculation, cards were pre-incubated in the humidity chamber for one, two or three days before inoculation with YFV (illustrated in ). As the pre-incubation period increased, the LOD decreased significantly ( , 1 pfu/punch low humidity vs 1 pfu/punch high humidity 3 days 95% CI: -4.78, -2.83)). The difference in LOD between one and two days of preincubation was not significant (one day vs. two days 95% CI: -0.1, 0.7); however, the difference in LOD after three days was significantly different (two days vs. three days 95% CI: 0.1, 0.9). At 10 pfu/punch, there was no difference in the period of detection between one and two days (one day vs. two days 95% CI: -0.6, 1.5) but there was a significant difference between two and three days (two days vs. three days 95% CI: 2.4, 4.7) . There was a significant difference in time of detection at 1 pfu/punch between both one and two (one day vs. two days 95% CI: 0.2, 3.0) and two and three days (two days vs. three days 95% CI: 2.8, 5.1) of preincubation in high humidity. YFV RNA was undetectable at 1 pfu/punch after preincubating cards for three days in high humidity . Desiccating FTA cards prior to inoculation improves YFV RNA stability In an attempt to reduce the loss of RNA yield at high humidity, pre-humidified cards were incubated with desiccation packets for one, two and seven days prior to inoculation with low titer YFV (illustrated in ). The addition of desiccation packets resulted in an extended detection time compared to the same concentration/punch incubated with no desiccation . For punches inoculated with 10 pfu/punch, two desiccation packets significantly extended the length of detection after one day of desiccation (95% CI: 0.1, 1.1) compared to one packet. For punches inoculated with 1 pfu/punch, there was not a significant change in the length of detection between cards pre-incubated for one or two days with one or two desiccant packets. For cards incubated for 7 days prior to inoculation with 10 pfu/punch (one desiccation packet vs two desiccation packets 95% CI: 0.2, 2.5) and 1 pfu/punch (one desiccation packet vs two desiccation packets 95% CI: 1.4, 3.4) of YFV, the addition of two desiccation packets significantly extended RNA detection when compared to the addition of one packet . When compared to FTA punches incubated at low humidity , one or two days of desiccation prior to inoculation resulted in similar RNA stability . Two brands of desiccant packets were tested and after two days of desiccation, there was no significant difference in the length of detection for either titer of YFV tested . The vaccine strain of YFV, 17D-204, was confirmed to be inactivated after storage on FTA cards. No CPE was observed in Vero cells inoculated with 17D-204 FTA punch extract after three blind passages of 1 week each. Additionally, no plaques were observed after a plaque assay was performed on 17D-204 FTA punch extract. To determine if vaccine and WT strains of YFV could be detected after inoculation onto FTA cards, the vaccine substrain 17D-204, WT strain Asibi, and PBS as a negative control were inoculated onto FTA cards and assayed for YFV RNA. The viruses were inoculated at the same titer (6log 10 pfu/mL). These C t values were compared to a standard extraction (140 μL of sample directly extracted) of both 17D-204 and Asibi. Although the mean C t values were lower when RNA was extracted from an FTA punch, both strain 17D-204 (mean C t value: 23.2 ± 0.12) and strain Asibi (mean C t value: 23.9 ± 0.1) were readily detected from FTA punches whereas the PBS control was negative for YFV RNA (mean Ct value: 43.3 ± 1.9) . Typically, when extracting viral RNA from a YFV clinical sample, 140 μL of serum is used as input in the extraction protocol. Although this volume can be applied to the entire FTA card, RNA extraction protocols utilizing FTA cards require the card to be punched prior to RNA extraction, limiting the volume of sample that is extracted. The timing of punching, pre-inoculation or post-inoculation was tested to ensure that the ‘pre-punch protocol’ did not impact RNA yield. There was no significant difference in C t values when sample was applied to the whole card and punched later (“post-inoculation punch”, R 2 = 0.99) and when sample was applied to pre-punched cards (“pre-inoculation punch”, R 2 = 0.98) (pre-inoculation punch vs post-inoculation punch 95% CI: -9.3, 10.97, ). To test if this reduction of input volume significantly impacts the sensitivity of the assay, the limit of detection of 10 μL virus applied to FTA punches was assayed and compared to a direct extraction from 10 μL of virus and the ‘gold standard’ of a direct extraction of 140 μL of virus. The LOD of 140 μL of virus (0.22 pfu, R 2 = 0.98) was shown to be significantly lower (140 μL virus vs one FTA punch 95% CI: 0.09, 1.4) than the LOD of one FTA punch (3.01 pfu, R 2 = 0.97); however, there was no significant difference (one FTA punch vs. 10 μL 95% CI: -0.2, 0.4) in the LOD of RNA extracted from one FTA punch compared to 10 μL of virus extracted directly (2.2 pfu, R 2 = 0.97) . To determine if extracting RNA from multiple punches increased sensitivity, RNA was pooled from two punches inoculated with virus diluted in PBS . Increasing the number of punches utilized, significantly lowered (two FTA punch vs one FTA punch 95% CI: 0.4, 1.0) the LOD (0.6 pfu, R 2 = 0.98) when compared to extracting RNA from one punch. This trend continued when RNA was extracted from three and four punches . Because serum is the most common sample used for molecular diagnoses, the experiment was repeated using human, flavivirus-negative serum (EMD Millipore, Burlington, MA) as the diluent. As was observed in the experiment utilizing PBS as the diluent, the LOD of 140 μL of virus diluted in human serum (0.5 pfu, R 2 = 0.96) was significantly lower (140 μL vs FTA punch 95% CI: 0.8, 1.5) than the LOD of an FTA punch inoculated with 10 μL virus diluted in human serum (2.7 pfu, R 2 = 0.95), but the LOD of an FTA punch inoculated with 10 μL virus diluted in serum was not significantly different (10 μL serum vs 10 μL PBS 95% CI: -0.3, 0.4) than a FTA punch inoculated with 10 μL virus diluted in PBS . At room temperature, 100,000 pfu/punch (R 2 = 0.68), 10,000 pfu/punch (R 2 = 0.80), 1,000 pfu/punch (R 2 = 0.64) and 100 pfu/punch (R 2 = 0.82) could be detected for over two weeks . YFV inoculated at 10 pfu/punch could be detected for 6.2 days (R 2 = 0.88) and 1 pfu/punch for 3.5 days (R 2 = 0.74) . At 37°C, 100,000 pfu/punch (R 2 = 0.78) and 10,000 pfu/punch (R 2 = 0.73) were still detectable for more than 14 days; however, the 1,000 pfu/punch sample (R 2 = 0.64) was detectable for 13.7 days . Similarly, the duration of detectability for the 100 pfu/punch (10.1 days, R 2 = 0.70), 10 pfu/punch (4.1 days, R 2 = 0.77) and 1 pfu/punch (2.2 days, R 2 = 0.72) samples decreased at 37°C . Using 95% confidence intervals to compare the number of days YFV was detectable by qRT-PCR, 10 pfu/punch (RT vs 37°C 95% CI: 1.52, 2.77) and 1 pfu/punch (RT vs 37°C 95% CI: 0.7, 2.0) became undetectable in a significantly shorter time at 37°C than at RT. To test if RNA extraction from multiple FTA punches improved the sensitivity of YFV molecular detection at high temperatures, RNA was pooled from two punches at 10 pfu/punch and 1 pfu/punch. It was shown that extracting RNA from two punches significantly increased the number of days RNA could be detected . The initial experiments were conducted in a controlled climate (40–55% humidity). To test if these humidity conditions positively influenced RNA stability on the FTA cards, cards were inoculated with 10 pfu/punch and 1 pfu/punch of YFV and placed into a humidity chamber (80–85%) for one week. Control samples were left in the biosafety cabinet at ambient humidity for one week. Cards incubated at high humidity showed no difference in the time limit for detection of positive C t values when compared to dry cards incubated at low humidity . Despite this, the negative effect of humidity on the ability of FTA cards to stabilize RNA has been frequently documented in regions endemic to YFV where high humidity is common . To assay if the reported loss in RNA yield was due to the card having absorbed moisture from the air prior to inoculation, cards were pre-incubated in the humidity chamber for one, two or three days before inoculation with YFV (illustrated in ). As the pre-incubation period increased, the LOD decreased significantly ( , 1 pfu/punch low humidity vs 1 pfu/punch high humidity 3 days 95% CI: -4.78, -2.83)). The difference in LOD between one and two days of preincubation was not significant (one day vs. two days 95% CI: -0.1, 0.7); however, the difference in LOD after three days was significantly different (two days vs. three days 95% CI: 0.1, 0.9). At 10 pfu/punch, there was no difference in the period of detection between one and two days (one day vs. two days 95% CI: -0.6, 1.5) but there was a significant difference between two and three days (two days vs. three days 95% CI: 2.4, 4.7) . There was a significant difference in time of detection at 1 pfu/punch between both one and two (one day vs. two days 95% CI: 0.2, 3.0) and two and three days (two days vs. three days 95% CI: 2.8, 5.1) of preincubation in high humidity. YFV RNA was undetectable at 1 pfu/punch after preincubating cards for three days in high humidity . In an attempt to reduce the loss of RNA yield at high humidity, pre-humidified cards were incubated with desiccation packets for one, two and seven days prior to inoculation with low titer YFV (illustrated in ). The addition of desiccation packets resulted in an extended detection time compared to the same concentration/punch incubated with no desiccation . For punches inoculated with 10 pfu/punch, two desiccation packets significantly extended the length of detection after one day of desiccation (95% CI: 0.1, 1.1) compared to one packet. For punches inoculated with 1 pfu/punch, there was not a significant change in the length of detection between cards pre-incubated for one or two days with one or two desiccant packets. For cards incubated for 7 days prior to inoculation with 10 pfu/punch (one desiccation packet vs two desiccation packets 95% CI: 0.2, 2.5) and 1 pfu/punch (one desiccation packet vs two desiccation packets 95% CI: 1.4, 3.4) of YFV, the addition of two desiccation packets significantly extended RNA detection when compared to the addition of one packet . When compared to FTA punches incubated at low humidity , one or two days of desiccation prior to inoculation resulted in similar RNA stability . Two brands of desiccant packets were tested and after two days of desiccation, there was no significant difference in the length of detection for either titer of YFV tested . The molecular diagnosis of YFV can be limited by the rate at which viral RNA degrades, which is accelerated by the hot and humid conditions present in many endemic regions . Because RNA is highly labile, suspected YFV clinical samples must be transported using continuous cold-chain maintenance, with samples arriving within a day of departure being shipped on wet ice, and those in transit for more than one day being shipped at -20°C . Whatman FTA cards have been employed as a solution for other RNA viruses as they rapidly inactivate virus and stabilize genetic material . Protocols utilizing a wide variety of sample types have been optimized using FTA cards including blood, serum, tick and mosquito homogenates, dead bird impressions, throat swabs, oral fluid and epithelial suspensions . In fact, the WHO and CDC approved the use of FTA cards for measles and poliovirus diagnostic protocols, indicating that the cards are effective in a clinical setting . In this study we optimize the use of Whatman FTA cards in conjunction with YFV molecular diagnostics and show that the incorporation of this widely available commercial product could lessen the requirement for cold-chain, which is often unavailable or incomplete. During a YFV outbreak, molecular diagnostics are highly important as IgM generated in response to WT YFV infection and vaccination are indistinguishable via serology. In this study, the YFall qRT-PCR assay was used to show that WT and vaccine strains of YFV were equally detectable after inoculation onto FTA cards . This molecular assay is commonly used in the field and has been shown to be both sensitive and specific for WT and vaccine strains of YFV . We have previously designed an qRT-PCR assay that distinguishes between the two strains using locked nucleic acid (LNA) technology and believe that the use of FTA cards to transport clinical samples during an outbreak could increase the successful use of both molecular assays. Viremia/RNA in sera during WT YFV infection is transient, especially in mild or subclinical cases, and usually begins to wane 3–6 days after infection . Viremia can last the duration of infection in severe cases of YF but is highly variable . In a study of the 2016–2018 Brazilian YFV outbreak viremia at time of death ranged from over 10 5 pfu/mL to less than 1 pfu/mL . In many cases, the sensitivity of molecular assays was shown to decrease with the use of FTA cards . It is hypothesized that this is due to many factors, including the volume of inoculum used, the humidity of endemic regions and incomplete drying of FTA card post-inoculation with sample . Although the LOD of YFV 17D RNA on FTA punches was shown to be less than 10 pfu/mL in both PBS and flavivirus-negative human sera, the LOD of FTA cards was significantly different than the gold standard of direct RNA extraction. Although achieving 100% of RNA recovery in the extraction is unlikely, loss in RNA yield was likely due to the volume of sample used to extract RNA as the FTA punch holds only 10 μL whereas a direct extraction involves 140 μL of clinical specimen. By extracting RNA from two or more punches, the LOD was significantly improved. Up to nine, 6 mm punches can be made from FTA Micro cards, which means RNA from multiple punches can be pooled leaving enough card for confirmatory replicates to be easily achieved. The reduction in sensitivity when using FTA cards compared to direct RNA extraction may be overcome by the use of multiple punches to allow detection of less than 1 pfu/mL. This would allow for detection of the full range of viral loads seen in YF clinical samples. The most common sample type used in molecular diagnosis of YF is serum. Using 17D-204 virus spiked into flavivirus-negative human serum, we showed no significant difference in RNA preservation compared to 17D-204 virus diluted in PBS. Many studies have shown flavivirus RNA may be more readily detected and detected for longer periods of time in whole blood, as the virus binds strongly to red blood cells . Though the effect of whole blood on YFV molecular diagnostics was not tested here, molecular diagnosis has been made from whole blood placed on FTA cards for Anaplasma and Rickettsia species . Taken together, it is likely that blood samples taken from suspected YF patients would be another acceptable sample type for YF molecular diagnosis from FTA cards. The ability to submit finger prick blood draws could increase the ease of YFV surveillance in remote regions where cold transport is unavailable. YFV transmission season coincides with the rainy season in both South American and African regions, which also corresponds to the hottest time of the year. FTA cards have been shown to stabilize RNA at high temperature, although the duration of stabilization does decrease as temperature increases . Here, at 37°C, high titer YFV RNA was preserved on FTA cards for more than two weeks and low titer RNA was preserved for two days at 37°C. Although there was a reduction in the time YFV RNA could be detected at high temperature compared to room temperature, two days is within the WHO’s current timeline for testing of YFV samples. This suggests that low titer samples would be detected after transport in normal conditions; however, during outbreak settings transport can exceed two days. Extracting RNA from two punches improved the time for RNA to be detectable to 3.7 days, further illustrating the utility of the double punch method. Extreme environmental conditions in endemic regions (i.e., higher temperatures) were not tested here, which is a limitation of this study. However, field trials will be done in different countries to analyze their impact on the YF RNA stabilization on FTA cards. The high humidity of transmission season is also proposed to limit RNA stabilization by FTA cards . One study showed that incomplete drying of FTA cards caused the most significant loss of RNA yield . Our data confirms this finding as storing cards for three days at high humidity affected RNA preservation more than three days of high temperature. To discern how to best mitigate this loss in RNA, the impact of humidity before and after inoculation was tested. Cards that were inoculated dry and then placed at high humidity showed no significant loss in yield; however, cards exposed to a highly humid environment for days prior to inoculation , displayed a significant loss in RNA detection. This suggests that high humidity decreases the ability of FTA cards absorb inoculum as water is absorbed from the environment by the FTA card, limiting sample capacity as others have noted . To counteract absorption of environmental humidity, ‘pre-humidified’ FTA cards were desiccated prior to, and after inoculation. The data presented here showed that two days of pre-desiccation significantly improved RNA stabilization. Desiccating FTA cards for longer was shown to be less effective, which is likely due to desiccation packets becoming fully saturated. If desiccation packets had been swapped for new packets, RNA may have been better stabilized. Nevertheless, YFV inoculated onto pre-desiccated cards could be detected days longer than cards that had been stored in humid conditions. This suggests that if cards are dry when samples are applied and stored in sealed containers with desiccation packets, RNA can be successfully stabilized and detected. FTA cards have been successfully incorporated into the molecular diagnostic protocols for many RNA viruses. Here we showed that FTA cards stabilized and inactivated YFV in average environmental conditions. The implementation of FTA cards could aid YFV surveillance in remote regions by simplifying and ensuring success of diagnostically-viable sample transport to laboratories. The use of FTA cards could make the difference between complete loss of detectable RNA and successful detection in many instances. Early detection of YF cases is integral to the EYE strategy for controlling YF outbreaks . The ability to reliably detect YFV RNA would reduce the burden on serological testing which is often complicated in endemic regions and during an outbreak scenario. As the molecular diagnosis of YFV becomes more routine in YF-endemic regions, RNA stabilization provided by FTA cards may become integral to ensuring accurate diagnosis. S1 Table Increasing the number of punches used in RNA extraction, improves detection. (DOCX) Click here for additional data file. S2 Table Pre-desiccated FTA cards perform similarly at low and high humidity. (DOCX) Click here for additional data file. S3 Table Incubating FTA cards at high humidity prior to inoculation decreases LOD of YFV RNA extracted from cards. (DOCX) Click here for additional data file. S4 Table Brand of 100g desiccation packet does not affect length of YFV RNA detection. (DOCX) Click here for additional data file. S1 Fig WT and vaccine strains of YFV can be detected using FTA cards. WT YFV strain Asibi and YFV vaccine strain 17D-204 were spotted onto FTA cards and allowed to dry for one hour. RNA was extracted from the cards and assayed using YFall qRT-PCR primers. RNA was also extracted directly from the Asibi and 17D-204 isolates using 140 μL of sample directly into lysis buffer. PBS was spotted onto FTA cards as a negative control. The dotted line at C t value 37 indicates the cut-off for YFV RNA positivity and is based on cut-offs used in the molecular diagnosis of YFV in the field. (TIF) Click here for additional data file. S2 Fig Punching FTA cards pre- or post-inoculation of sample does not impact RNA yield. YFV 17D-204 virus was inoculated onto FTA cards (130 μL sample over entire card) or onto FTA card punches (10 μL/ punch) and allowed to dry completely. RNA was then extracted from punches made from the whole FTA card or the pre-punched samples. The dotted line at C t value 37 indicates the cut-off for YFV RNA positivity and is based on cut-offs used in the molecular diagnosis of YFV in the field. Dilutions were performed in PBS. Points and error bars represent the average of two experiments and data was fit using a linear regression. (TIF) Click here for additional data file. S3 Fig Experimental design of effects of humidity YFV RNA yield from FTA cards. Made using BioRender. (PNG) Click here for additional data file. |
MLPA and DNA index improve the molecular diagnosis of childhood B-cell acute lymphoblastic leukemia | 5f4210aa-93d6-42d6-a4ea-d98894f219da | 7359332 | Pathology[mh] | As the most common pediatric cancer, acute lymphoblastic leukemia (ALL) accounts for approximately 25% of childhood malignancies. With improved risk-directed treatment and supportive care, the overall 5-year event-free survival rates for this disease now exceed 80% in developed countries – . The two major features of risk-directed therapy are based on the genetic alterations of the leukemic cells at diagnosis and the determination of initial treatment response (measured by minimal residual disease, MRD, after induction therapy). The interpretation of MRD levels depends upon the subtype of ALL , . Although karyotyping has been the most common approach for detection of numerical chromosomal changes, molecular methods may enhance their detection in childhood B-ALL. Multiplex Ligation-dependent Probe Amplification (MLPA) is a sensitive method based upon the multiplex polymerase chain reaction and capillary electrophoresis that detects multiple copies of around 50 different genomic DNA targets. It has the advantage of lower price and quicker turn-around time than DNA arrays for identification of the important genetic alterations and is now widely used for detection of the important copy number changes in ALL , . Gain or loss of whole chromosomes (aneuploidy) and intrachromosomal amplification of chromosome 21 (iAMP21) accounts for almost 30% of childhood B-ALL identified by traditional methods. High hyperdiploidy with greater than 50 chromosomes comprises up to 30% of childhood B-ALL and most commonly involves gains of chromosomes X, 4, 10, 14, 17 and 21 . It is associated with a good outcome, even in patients with induction failure . Hypodiploidy with less than 44 chromosomes is less common (found in approximately 3% of cases) and is associated with an inferior outcome. Hypodiploid B-ALL can be further divided into three subgroups according to chromosome number. The most common are near-haploidy with 24–31 chromosomes and low-hypodiploidy with 32–39 chromosomes. High-hypodiploidy with 40–43 chromosomes is rare. Low-hypodiploid ALL has a high incidence of TP53 germline mutations . DNA index (DI) is a well-established method for detection of high hyperdiploidy. The MLPA telomere kit identifies specific gain or loss of individual chromosomes and is suitable for screening for whole chromosome numerical changes , . Masked low hypodiploidy, manifesting as doubling of the low hypodiploid chromosome number, can be difficult to diagnose . Here we show that MLPA and DI are useful in its detection, as confirmed by single-nucleotide polymorphism (SNP) arrays and short tandem repeats (STR). B-ALL patients with iAMP21-ALL were initially shown to have a high relapse risk on standard chemotherapy , . It was later demonstrated that treatment on intensive therapy regimens significantly reduced their risk of relapse – . In childhood B-ALL, SNP arrays have successfully identified copy number abnormalities (CNA) involving several signaling pathways. For example, deletions of a number of genes within the B-cell differentiation pathway were identified, including PAX5 , EFB1 and IKZF1 , . Clinically, IKZF1 alterations have been associated with a poor outcome, particularly in association with Ph-positive (Philadelphia chromosome/ BCR-ABL1 positive), and Ph-like ALL (Philadelphia chromosome/ BCR-ABL1 negative but the expression profiles were similar to Ph-positive ALL) – . Ph-like and iAMP21-ALL have been proposed as novel subtypes of B-ALL in the recent WHO classification of hematologic malignancies, due to their poor prognostic associations . In this project, we have used MLPA and DI to study CNA in B-ALL. We show that these approaches are complementary to cytogenetics in improving detection of genetic alterations in childhood B-ALL. Patients and protocols Diagnostic bone marrow (BM) or peripheral blood was obtained from 233 children with B-ALL from January 2002 to July 2018 at the National Taiwan University Hospital. A total of 108 patients were treated on the Taiwan Pediatric Oncology Group TPOG-ALL-2002 protocol, while 125 were treated on TPOG-ALL-2013. Diagnosis of B-ALL was based on BM morphology and the immunophenotype of leukemic cells was determined by flow cytometry. Conventional cytogenetic analysis was carried out as part of the routine work-up . The risk-directed TPOG protocols consist of multiple chemotherapeutic agents of different intensities. The treatment protocol was intensified if complete remission was not achieved after initial induction therapy. After 2013, MRD levels were added to risk assignment for therapy. Events were defined as any relapse, death, or secondary malignancy. The Institutional Review Board of National Taiwan University Hospital approved the study and all of the participants or their guardians provided written informed consent in accordance with the Declaration of Helsinki. Details of the protocols and risk group assignment have been published elsewhere , , . We have summarized the risk classification of protocols in the . Genomic DNA extraction Lymphoblasts were purified from bone marrow or peripheral blood specimens using the Ficoll-Paque centrifugation method, according to the manufacturer’s instructions (GE Healthcare, Piscataway, NJ, USA). Genomic DNA was extracted from leukemic cells using standard phenol/chloroform-based methods. Briefly, 1 million cells were lysed in 10 mM Tris–HCl, 10 mM NaCl, 10 mM EDTA, 20 μg proteinase K, and 0.5% SDS by incubating at 37 °C for 16 h. Total RNA was further removed by adding 500 μg PureLink RNase A (Invitrogen, USA) and incubating for 10 min at 37 °C. An equal volume of phenol–chloroform–isopropanol (25:24:1) was added to lysates and mixed by shaking vigorously, followed by centrifugation at 16,100 × g at 4 °C for 5 min. The upper aqueous phase was transferred to a fresh tube; genomic DNA was then precipitated by adding 2× volume − 80 °C 100% ethanol. The DNA pellet was washed with 75% ethanol and rehydrated with Tris–EDTA buffer. The concentration of DNA was determined using a NanoDrop 1,000 spectrophotometer (Thermo Fisher Scientific, Waltham, Massachusetts, USA) . MLPA analysis Genomic DNA was analyzed using the SALSA MLPA kit (MRC-Holland, Amsterdam, the Netherlands), according to manufacturer’s instructions. The PCR fragments were separated by capillary electrophoresis on a Life Technologies 3,500 Genetic Analyzer (Thermo Fisher Scientific, Waltham, MA, USA). MLPA data were analyzed using Coffalyser.Net v.140721.1958 (MRC-Holland, Amsterdam, The Netherlands). Probe ratio between 0.75 and 1.3 were considered to be within the normal range. Probe ratio below 0.75 or above 1.3 indicated deletion or gain, respectively. Probe ratio below 0.25 or above 1.8 indicated biallelic deletion or amplification. SALSA MLPA P335 ALL-IKZF1 probemix was used for detection of alterations of EBF1 , IKZF1 , CDKN2A , CDKN2B , PAX5 , ETV6 , RB1 and BTG1 genes. SALSA MLPA P327 iAMP21-ERG probemix was used for detecting alterations of ERG gene and iAMP21. SALSA MLPA P329 CRLF2-CSF2RA-IL3RA probemix was used for detecting P2RY8-CRLF2 (PAR1 deletion). Analysis of ploidy status Ploidy status was evaluated by SALSA MLPA P036 Subtelomeres Mix 1 probe mix. Whole chromosomal gain or loss was defined when two probes targeting p and q arms of the same chromosome were respectively gained or deleted simultaneously. Chromosome 19p deletions were defined when the probe targeted the p arm of chromosome 19 was deleted while q arm was normal. DNA index measured by flow cytometry Freshly prepared or frozen leukemia samples were used for DNA index analysis. Peripheral blood derived from normal healthy individuals was used as controls for diploidy. Mononuclear cells were isolated by Ficoll-Paque (GE Healthcare, Chicago, IL, USA) according to the manufacturer’s instructions. Three cell suspensions were prepared: tube A was a mixture of leukemia cells and normal PBMCs in equal numbers; tubes B and C contained normal PBMCs or leukemia cells alone. Each cell suspension (2 million cells) was fixed with 70% ethanol overnight at – 20 °C. Fixed cells were washed with 1× PBS and then incubated with propidium iodide (50 μg) and RNase (10 μg) for 1 h on ice. Cells were filtered with 100 μm cell strainer and then analyzed by FACSCalibur (BD, Franklin Lakes, NJ, USA). DNA quantity of an individual cell population was determined and DNA index represents the ratio of leukemia sample/normal PBMCs fluorescence calculated from tube A. Tubes B and tube C were used as reference to distinguish the leukemia from PBMC peaks in tube A. Theoretical DNA index (tDI) was calculated using the formula: tDI = chromosome numbers × 0.0202 + 0.0675 . Statistical analysis Pearson's correlations, the coefficient of determination and p-values were carried out between the results of DI and tDI from MLPA and cytogenetics. Fisher’s exact test was performed to evaluate the enrichment of 19p deletion in TCF3 gene rearranged ALL. The log-rank test compared different survival curves between patients with different major genetic subtypes, patient with or without IKZF1 deletion and patients with or without IKZF1 plus . Overall survival (OS) was defined as diagnosis to death. Patients who did not suffer any adverse events within the follow-up period were censored. Event-free survival (EFS) of patients with no response to chemotherapy (refractory), death, and second relapse in induction was set to 0. Univariate and multivariate Cox regression were performed to evaluate hazard ratios (HR) and 95% confidence intervals (CI) of risk factors. All statistical analyses were performed using the Statistical Product and Services Solutions (SPSS) statistical package, v18.0 (IBM, Armonk, NY, USA). Diagnostic bone marrow (BM) or peripheral blood was obtained from 233 children with B-ALL from January 2002 to July 2018 at the National Taiwan University Hospital. A total of 108 patients were treated on the Taiwan Pediatric Oncology Group TPOG-ALL-2002 protocol, while 125 were treated on TPOG-ALL-2013. Diagnosis of B-ALL was based on BM morphology and the immunophenotype of leukemic cells was determined by flow cytometry. Conventional cytogenetic analysis was carried out as part of the routine work-up . The risk-directed TPOG protocols consist of multiple chemotherapeutic agents of different intensities. The treatment protocol was intensified if complete remission was not achieved after initial induction therapy. After 2013, MRD levels were added to risk assignment for therapy. Events were defined as any relapse, death, or secondary malignancy. The Institutional Review Board of National Taiwan University Hospital approved the study and all of the participants or their guardians provided written informed consent in accordance with the Declaration of Helsinki. Details of the protocols and risk group assignment have been published elsewhere , , . We have summarized the risk classification of protocols in the . Lymphoblasts were purified from bone marrow or peripheral blood specimens using the Ficoll-Paque centrifugation method, according to the manufacturer’s instructions (GE Healthcare, Piscataway, NJ, USA). Genomic DNA was extracted from leukemic cells using standard phenol/chloroform-based methods. Briefly, 1 million cells were lysed in 10 mM Tris–HCl, 10 mM NaCl, 10 mM EDTA, 20 μg proteinase K, and 0.5% SDS by incubating at 37 °C for 16 h. Total RNA was further removed by adding 500 μg PureLink RNase A (Invitrogen, USA) and incubating for 10 min at 37 °C. An equal volume of phenol–chloroform–isopropanol (25:24:1) was added to lysates and mixed by shaking vigorously, followed by centrifugation at 16,100 × g at 4 °C for 5 min. The upper aqueous phase was transferred to a fresh tube; genomic DNA was then precipitated by adding 2× volume − 80 °C 100% ethanol. The DNA pellet was washed with 75% ethanol and rehydrated with Tris–EDTA buffer. The concentration of DNA was determined using a NanoDrop 1,000 spectrophotometer (Thermo Fisher Scientific, Waltham, Massachusetts, USA) . Genomic DNA was analyzed using the SALSA MLPA kit (MRC-Holland, Amsterdam, the Netherlands), according to manufacturer’s instructions. The PCR fragments were separated by capillary electrophoresis on a Life Technologies 3,500 Genetic Analyzer (Thermo Fisher Scientific, Waltham, MA, USA). MLPA data were analyzed using Coffalyser.Net v.140721.1958 (MRC-Holland, Amsterdam, The Netherlands). Probe ratio between 0.75 and 1.3 were considered to be within the normal range. Probe ratio below 0.75 or above 1.3 indicated deletion or gain, respectively. Probe ratio below 0.25 or above 1.8 indicated biallelic deletion or amplification. SALSA MLPA P335 ALL-IKZF1 probemix was used for detection of alterations of EBF1 , IKZF1 , CDKN2A , CDKN2B , PAX5 , ETV6 , RB1 and BTG1 genes. SALSA MLPA P327 iAMP21-ERG probemix was used for detecting alterations of ERG gene and iAMP21. SALSA MLPA P329 CRLF2-CSF2RA-IL3RA probemix was used for detecting P2RY8-CRLF2 (PAR1 deletion). Ploidy status was evaluated by SALSA MLPA P036 Subtelomeres Mix 1 probe mix. Whole chromosomal gain or loss was defined when two probes targeting p and q arms of the same chromosome were respectively gained or deleted simultaneously. Chromosome 19p deletions were defined when the probe targeted the p arm of chromosome 19 was deleted while q arm was normal. Freshly prepared or frozen leukemia samples were used for DNA index analysis. Peripheral blood derived from normal healthy individuals was used as controls for diploidy. Mononuclear cells were isolated by Ficoll-Paque (GE Healthcare, Chicago, IL, USA) according to the manufacturer’s instructions. Three cell suspensions were prepared: tube A was a mixture of leukemia cells and normal PBMCs in equal numbers; tubes B and C contained normal PBMCs or leukemia cells alone. Each cell suspension (2 million cells) was fixed with 70% ethanol overnight at – 20 °C. Fixed cells were washed with 1× PBS and then incubated with propidium iodide (50 μg) and RNase (10 μg) for 1 h on ice. Cells were filtered with 100 μm cell strainer and then analyzed by FACSCalibur (BD, Franklin Lakes, NJ, USA). DNA quantity of an individual cell population was determined and DNA index represents the ratio of leukemia sample/normal PBMCs fluorescence calculated from tube A. Tubes B and tube C were used as reference to distinguish the leukemia from PBMC peaks in tube A. Theoretical DNA index (tDI) was calculated using the formula: tDI = chromosome numbers × 0.0202 + 0.0675 . Pearson's correlations, the coefficient of determination and p-values were carried out between the results of DI and tDI from MLPA and cytogenetics. Fisher’s exact test was performed to evaluate the enrichment of 19p deletion in TCF3 gene rearranged ALL. The log-rank test compared different survival curves between patients with different major genetic subtypes, patient with or without IKZF1 deletion and patients with or without IKZF1 plus . Overall survival (OS) was defined as diagnosis to death. Patients who did not suffer any adverse events within the follow-up period were censored. Event-free survival (EFS) of patients with no response to chemotherapy (refractory), death, and second relapse in induction was set to 0. Univariate and multivariate Cox regression were performed to evaluate hazard ratios (HR) and 95% confidence intervals (CI) of risk factors. All statistical analyses were performed using the Statistical Product and Services Solutions (SPSS) statistical package, v18.0 (IBM, Armonk, NY, USA). Frequency of copy number abnormalities in children with B-ALL The demographic, clinical, and laboratory characteristics of 233 children with B-ALL are shown in Table . The median age of the cohort was 5.4 years (range < 0.1–17.9 years); 95.3% of the patients were over 1 year of age. The molecular tests performed were those standardized for B-ALL diagnosis including: ETV6-RUNX1 , TCF3-PBX1 , BCR-ABL1 , P2RY8-CRLF2 and KMT2A-AFF1 for 220 samples. Detailed flow diagram of analysis used in this study is demonstrated in Supplementary Fig. . From MLPA testing, overall, 65.7% of the patients (153/233) harbored at least one abnormality (either deletion or amplification) involving the following genes— IKZF1 , CDKN2A / 2B , PAX5 , EBF1 , ETV6 , BTG1 , RB1 , ERG or PAR1 region, whereas the remaining 34.3% (80/233) of patients had none of these abnormalities. Simultaneous aberrations in different genes were observed. A heatmap listing these CNA in the entire cohort are given in Fig. . Details of the CNA in each major cytogenetic subtype are shown in Supplementary Table . DNA index identifies ploidy status in ALL In 112 samples DNA index analysis was performed; 41 cases showed aneuploidy, of which 35 were high hyperdiploid, 3 were hypodiploid and in 3 cases masked hypodiploidy was indicted, as described below. However, DI cannot identify individual chromosome gain or loss. MLPA compared to DI and cytogenetics Good quality genomic DNA was available from 204 samples for MLPA analysis using the MLPA P036 kit which identified 57 patients with high hyperdiploidy, 7 with hypodiploidy and 140 with diploidy or near-diploidy. The numerical chromosomal alterations determined by this MLPA P036 kit were compared with the karyotype and DI results. These results showed concordance in number of chromosomes (r = 0.9780, P < 0.0001) for the 111 patients with both MLPA and DI data available (Fig. a). There was statistically significant positive correlation between karyotype and DI (r = 0.3308, P = 0.0005) (Fig. b), yet lower than MLPA against DI, among 188 patients with karyotype and MLPA data available. The statistically significant positive correlation was also seen between karyotype and MLPA (r = 0.4428, P < 0.0001) (Fig. c), but lower than MLPA against DI. We found that 45% (29/64) of patients with high hyperdiploidy or hypodiploidy identified either by DI or MLPA P036 were non-informative. Details of karyotype, DI and MLPA of the cohort are listed in Supplementary Table . High hyperdiploidy Among 57 cases with high hyperdiploidy, the majority (94.5%) had gained between 5 and 13 chromosomes (modal chromosome number, MCN, 51–63, inclusive), and the most frequent MCN was 54 chromosomes (Supplementary Fig. a). Chromosome gains were non-random and 8 chromosomes accounted for 82% of all gains: 4 (72.7% of cases), 6 (80.7%), 10 (84.2%), 14 (93.0%), 17 (80.7%), 18 (86.0%), 21 (100%), and X (78.9%) (Supplementary Fig. b). Gains of chromosomes 5, 8, 9, 11, 12, and 22 represented 15% of the total and were present in between 11 and 35% of cases. Gains of chromosomes 1, 2, 3, 7, 13, 15, 16, 19, and 20 were rare, totaling 3% of chromosomes gained. These patterns of chromosomal gains in these high hyperdiploid cases were similar to previous reports. The MLPA pattern of iAMP21 and differentiation between iAMP21 and high hyperdiploidy From their MLPA plots, we identified four patients with iAMP21, as shown in Supplementary Fig. . A characteristic chromosome 21 copy number profile has been previously described for cases of iAMP21-ALL from microarray studies and next generation sequencing. It is described as copy number changes from centromere to telomere along chromosome 21, with the highest level of amplification proximal to a telomeric deletion – . Tsuchiya et al. reported a case in which RUNX1 was not located within the highest region of amplification of chromosome 21 . In this cohort, RUNX1 was observed within the most highly amplified region of chromosome 21, with the exception of one case (Supplementary Fig. ). In high hyperdiploid cases, the DI is usually greater than 1.16 and associated with frequent gains of chromosomes 4, 6, 10, 14, 18, 21 and X. We compared the pattern of chromosome 21 gain in high hyperdiploid and iAMP21-ALL in our cohort. SNP arrays analysis was carried out on two iAMP21-ALL samples diagnosed by MLPA (Supplementary Fig. ). For cases with suspected iAMP21, in the absence of SNP arrays, DI and MLPA P036 and P327 kits can provide the definitive answer. Hypodiploid cases Five patients with low DI were diagnosed with hypodiploidy. Three of them had two peaks in the DI, indicating the presence of hypodiploid clone undergo a doubling of the chromosomes during metaphase. This manifestation is known as masked hypodiploidy. As the diagnosis of masked hypodiploidy requires demonstration of loss of heterozygosity (LOH), these three samples were analyzed by SNP arrays and LOH was seen, as shown in case 984 (Fig. ). DI showed two peaks: the smaller one (FL2-A value = 202) is the true hypodiploidy and the higher one (FL2-A value = 393) indicates the doubled hypodiploid population. These hypodiploid samples were also tested using MLPA P036 kit. By comparing MLPA with the value of DI, we were able to identify the specific losses and retention of each chromosome number. Thus, we were able to confirm that the masked hypodiploid population originated from doubling of the low hypodiploid one. In Fig. , the chromosome gains detected by MLPA P036 corresponded to the retained chromosomes. In contrast, the “normal” chromosomes, for example chromosomes 3, 4, 5, 7, 8, 9, 13, 15, 16, 17 and 20 were shown to be lost. The actual gain or loss of each chromosome cannot be inferred from the DNA index. Using the MLPA P036 kit, we identified another two cases of hypodiploidy (patients 508 and 753) in which LOH was confirmed by STR (see below). Details of these patients are listed in the Table . A Short Tandem Repeat (STR) is a microsatellite, consisting of a unit of two to thirteen nucleotides repeated hundreds of times on a DNA strand. STR analysis measures the precise number of repeating units. STR is used for confirmation of donor engraftment following stem cell transplantations and this test is available in all medical centers . Samples of germline (if available) and tumor were sent for STR analysis in order to confirm LOH identified on SNP arrays. We show the interpretation of STR for patient 984 in Supplementary Fig. and all three cases of masked hypodiploidy by STR are shown in Supplementary Table . STR provides a simple method to confirm the presence of LOH. Based upon these observations, we have proposed a flowchart for diagnosis of masked hypodiploidy (Supplementary Fig. ). 19p deletion by MLPA is an indicator of TCF3 translocations in childhood ALL We identified 7 of 12 cases of TCF3-PBX1 and two cases of TCF3-HLF with 19p loss. This enrichment differs from other subtypes of B-cell ALL (P < 0.0001) (Table ). TCF3 is an important transcriptional factor with multiple fusion partners in ALL. Samples with 19p deletions without evidence of TCF3-PBX1 or TCF3-HLF fusions may carry TCF3-ZNF384 fusions. TCF3-ZNF384 fusions represent another important subtype of B-cell ALL with a specific immunophenotype showing frequent CD10 loss and CD13 and CD33 expression. From these observations, we observed one sample with 19p loss, loss of CD10, CD13 and CD33 expression in which the TCF3-ZNF384 fusions was identified by RT-PCR (Supplementary Fig. ). In this series, among a total of 15 samples with 19p loss, 10 of them had TCF3 fusions. Novel subtypes of ALL, intragenic amplifications of PAX5 ( PAX5 AMP ), IKZF1-plus and ERG deletions Recently two papers have reported two novel high-risk subtypes of childhood ALL, PAX5 AMP and IKZF1 plus , . There were 23 IKZF1 plus patients and 5 patients with PAX5 AMP in this cohort. Nine patients (9/233 = 3.9%) were identified with ERG deletions. These ERG deletions were associated with different subtypes of ALL (Fig. ). Survival analysis Among patients with the major cytogenetic alterations, two with TCF3-HLF relapsed and died within 5 years from diagnosis. Patients with high-risk subtypes (Ph-positive/-like, hypodiploidy, MEF2D -r, KMT2A -r, TCF3-HLF , iAMP21) had inferior 5-year EFS (P < 0.0001) and OS (P < 0.0001) (Fig. a, b). The overall outcome was slightly inferior compared to previous TPOG reports, likely due to many of them being referred from other hospitals after relapse . All patients with iAMP21 were not detected at diagnosis. There is a trend that patients with IKZF1 plus had inferior 5 year-EFS and OS than patients without IKZF1 plus , but it did not reach statistical significance (Fig. c, d). Patients with IKZF1 deletions had inferior 5-year EFS and 5-year OS, but it also did not reach statistical significance (Fig. e, f). In the Cox multivariate regression model, IKZF1 deletions were not a strong predictor of poor outcome (Supplementary Table ). The demographic, clinical, and laboratory characteristics of 233 children with B-ALL are shown in Table . The median age of the cohort was 5.4 years (range < 0.1–17.9 years); 95.3% of the patients were over 1 year of age. The molecular tests performed were those standardized for B-ALL diagnosis including: ETV6-RUNX1 , TCF3-PBX1 , BCR-ABL1 , P2RY8-CRLF2 and KMT2A-AFF1 for 220 samples. Detailed flow diagram of analysis used in this study is demonstrated in Supplementary Fig. . From MLPA testing, overall, 65.7% of the patients (153/233) harbored at least one abnormality (either deletion or amplification) involving the following genes— IKZF1 , CDKN2A / 2B , PAX5 , EBF1 , ETV6 , BTG1 , RB1 , ERG or PAR1 region, whereas the remaining 34.3% (80/233) of patients had none of these abnormalities. Simultaneous aberrations in different genes were observed. A heatmap listing these CNA in the entire cohort are given in Fig. . Details of the CNA in each major cytogenetic subtype are shown in Supplementary Table . In 112 samples DNA index analysis was performed; 41 cases showed aneuploidy, of which 35 were high hyperdiploid, 3 were hypodiploid and in 3 cases masked hypodiploidy was indicted, as described below. However, DI cannot identify individual chromosome gain or loss. Good quality genomic DNA was available from 204 samples for MLPA analysis using the MLPA P036 kit which identified 57 patients with high hyperdiploidy, 7 with hypodiploidy and 140 with diploidy or near-diploidy. The numerical chromosomal alterations determined by this MLPA P036 kit were compared with the karyotype and DI results. These results showed concordance in number of chromosomes (r = 0.9780, P < 0.0001) for the 111 patients with both MLPA and DI data available (Fig. a). There was statistically significant positive correlation between karyotype and DI (r = 0.3308, P = 0.0005) (Fig. b), yet lower than MLPA against DI, among 188 patients with karyotype and MLPA data available. The statistically significant positive correlation was also seen between karyotype and MLPA (r = 0.4428, P < 0.0001) (Fig. c), but lower than MLPA against DI. We found that 45% (29/64) of patients with high hyperdiploidy or hypodiploidy identified either by DI or MLPA P036 were non-informative. Details of karyotype, DI and MLPA of the cohort are listed in Supplementary Table . Among 57 cases with high hyperdiploidy, the majority (94.5%) had gained between 5 and 13 chromosomes (modal chromosome number, MCN, 51–63, inclusive), and the most frequent MCN was 54 chromosomes (Supplementary Fig. a). Chromosome gains were non-random and 8 chromosomes accounted for 82% of all gains: 4 (72.7% of cases), 6 (80.7%), 10 (84.2%), 14 (93.0%), 17 (80.7%), 18 (86.0%), 21 (100%), and X (78.9%) (Supplementary Fig. b). Gains of chromosomes 5, 8, 9, 11, 12, and 22 represented 15% of the total and were present in between 11 and 35% of cases. Gains of chromosomes 1, 2, 3, 7, 13, 15, 16, 19, and 20 were rare, totaling 3% of chromosomes gained. These patterns of chromosomal gains in these high hyperdiploid cases were similar to previous reports. From their MLPA plots, we identified four patients with iAMP21, as shown in Supplementary Fig. . A characteristic chromosome 21 copy number profile has been previously described for cases of iAMP21-ALL from microarray studies and next generation sequencing. It is described as copy number changes from centromere to telomere along chromosome 21, with the highest level of amplification proximal to a telomeric deletion – . Tsuchiya et al. reported a case in which RUNX1 was not located within the highest region of amplification of chromosome 21 . In this cohort, RUNX1 was observed within the most highly amplified region of chromosome 21, with the exception of one case (Supplementary Fig. ). In high hyperdiploid cases, the DI is usually greater than 1.16 and associated with frequent gains of chromosomes 4, 6, 10, 14, 18, 21 and X. We compared the pattern of chromosome 21 gain in high hyperdiploid and iAMP21-ALL in our cohort. SNP arrays analysis was carried out on two iAMP21-ALL samples diagnosed by MLPA (Supplementary Fig. ). For cases with suspected iAMP21, in the absence of SNP arrays, DI and MLPA P036 and P327 kits can provide the definitive answer. Five patients with low DI were diagnosed with hypodiploidy. Three of them had two peaks in the DI, indicating the presence of hypodiploid clone undergo a doubling of the chromosomes during metaphase. This manifestation is known as masked hypodiploidy. As the diagnosis of masked hypodiploidy requires demonstration of loss of heterozygosity (LOH), these three samples were analyzed by SNP arrays and LOH was seen, as shown in case 984 (Fig. ). DI showed two peaks: the smaller one (FL2-A value = 202) is the true hypodiploidy and the higher one (FL2-A value = 393) indicates the doubled hypodiploid population. These hypodiploid samples were also tested using MLPA P036 kit. By comparing MLPA with the value of DI, we were able to identify the specific losses and retention of each chromosome number. Thus, we were able to confirm that the masked hypodiploid population originated from doubling of the low hypodiploid one. In Fig. , the chromosome gains detected by MLPA P036 corresponded to the retained chromosomes. In contrast, the “normal” chromosomes, for example chromosomes 3, 4, 5, 7, 8, 9, 13, 15, 16, 17 and 20 were shown to be lost. The actual gain or loss of each chromosome cannot be inferred from the DNA index. Using the MLPA P036 kit, we identified another two cases of hypodiploidy (patients 508 and 753) in which LOH was confirmed by STR (see below). Details of these patients are listed in the Table . A Short Tandem Repeat (STR) is a microsatellite, consisting of a unit of two to thirteen nucleotides repeated hundreds of times on a DNA strand. STR analysis measures the precise number of repeating units. STR is used for confirmation of donor engraftment following stem cell transplantations and this test is available in all medical centers . Samples of germline (if available) and tumor were sent for STR analysis in order to confirm LOH identified on SNP arrays. We show the interpretation of STR for patient 984 in Supplementary Fig. and all three cases of masked hypodiploidy by STR are shown in Supplementary Table . STR provides a simple method to confirm the presence of LOH. Based upon these observations, we have proposed a flowchart for diagnosis of masked hypodiploidy (Supplementary Fig. ). TCF3 translocations in childhood ALL We identified 7 of 12 cases of TCF3-PBX1 and two cases of TCF3-HLF with 19p loss. This enrichment differs from other subtypes of B-cell ALL (P < 0.0001) (Table ). TCF3 is an important transcriptional factor with multiple fusion partners in ALL. Samples with 19p deletions without evidence of TCF3-PBX1 or TCF3-HLF fusions may carry TCF3-ZNF384 fusions. TCF3-ZNF384 fusions represent another important subtype of B-cell ALL with a specific immunophenotype showing frequent CD10 loss and CD13 and CD33 expression. From these observations, we observed one sample with 19p loss, loss of CD10, CD13 and CD33 expression in which the TCF3-ZNF384 fusions was identified by RT-PCR (Supplementary Fig. ). In this series, among a total of 15 samples with 19p loss, 10 of them had TCF3 fusions. PAX5 ( PAX5 AMP ), IKZF1-plus and ERG deletions Recently two papers have reported two novel high-risk subtypes of childhood ALL, PAX5 AMP and IKZF1 plus , . There were 23 IKZF1 plus patients and 5 patients with PAX5 AMP in this cohort. Nine patients (9/233 = 3.9%) were identified with ERG deletions. These ERG deletions were associated with different subtypes of ALL (Fig. ). Among patients with the major cytogenetic alterations, two with TCF3-HLF relapsed and died within 5 years from diagnosis. Patients with high-risk subtypes (Ph-positive/-like, hypodiploidy, MEF2D -r, KMT2A -r, TCF3-HLF , iAMP21) had inferior 5-year EFS (P < 0.0001) and OS (P < 0.0001) (Fig. a, b). The overall outcome was slightly inferior compared to previous TPOG reports, likely due to many of them being referred from other hospitals after relapse . All patients with iAMP21 were not detected at diagnosis. There is a trend that patients with IKZF1 plus had inferior 5 year-EFS and OS than patients without IKZF1 plus , but it did not reach statistical significance (Fig. c, d). Patients with IKZF1 deletions had inferior 5-year EFS and 5-year OS, but it also did not reach statistical significance (Fig. e, f). In the Cox multivariate regression model, IKZF1 deletions were not a strong predictor of poor outcome (Supplementary Table ). In this retrospective study, the MLPA P036 subtelomeres probemix kit provided accurate detection of aneuploidy in childhood B-cell ALL and good correlation with the results from DI. MLPA and DI are superior to traditional cytogenetics, due to the shorter turn-around time, irrespective of mitotic index and improved sensitivity. Detections of specific gains or losses of each chromosome assist the differential diagnosis of hyperdiploidy from iAMP21. In addition, DI is helpful for diagnosis of masked hypodiploidy and LOH should be confirmed by SNP arrays. STR provides a simple method, available in most medical centers in Taiwan, to document LOH in these masked hypodiploid cases. Around 1.7% (4/233) of B-ALL patients had iAMP21. We also identified some of the novel ALL subtypes, including PAX5 AMP , and IKZF1 plus , . TCF3 rearrangements were frequently associated with 19p deletions. High hyperdiploidy accounts for around 20 ~ 25 percentage of childhood B-cell ALL . In this cohort, the most frequent modal chromosome number was 54 followed by 55. The most frequent gains included chromosomes 4, 6, 10, 18, 16, 17, 18, 21 and X, in agreement with previous reports , , . This incidence of high hyperdiploidy was lower in Taiwan than Caucasian populations , , . Using DI and the MLPA P036 kit, the incidence was around 27% in this cohort. In this study, 45% of high hyperdiploid patients were not detected by cytogenetics, manifesting as normal karyotype. In previous TPOG ALL 2002 report, hyperdiploidy accounted for 13.6% in B-ALL (n = 1,209). The incidence was much lower than that of this report. The reason for this discrepancy might be the relative smaller case numbers in this study. For cases without metaphases or normal karyotype, DI and MLPA can be successfully used for diagnosis of high hyperdiploidy . iAMP21-ALL is a novel subtype of B-ALL proposed by WHO , , , . The initial gold standard for diagnosis was FISH using probes directed to the RUNX1 gene, but array-CGH or SNP arrays are now the main method for diagnosis . One MLPA kit can successfully identify iAMP21 due to the density of probes along the long arm of chromosome 21. We identified 4 cases with iAMP21 by MLPA. In these cases, the level of gain was variable along the length of chromosome 21 with the ratio being more than 3.0, higher than in cases where chromosome 21 is gained as part of a high hyperdiploidy karyotype in which the probe ratio for every probe in the kit being ~ 1.5–2.0. These data correlated with other gains, especially of chromosomes 4, 6, 10, 18, 16, 17, 18 and X. If gains of chromosomes X, 4, 6, 10, 14, 17 and 18 are detected at the same time as gains of 21, it is most likely that the patient has high hyperdiploidy rather than iAMP21-ALL. Masked hypodiploidy can be difficult to diagnose. Another study used a similar MLPA approach to identify the aneuploidy status of relapsed B-cell ALL . Three patients with high hyperdiploidy had the highest number of chromosomal gains (median 11). Gains of the classical high hyperdiploidy pattern were less frequent, but gains of non-classical chromosomes, especially 1, 5, 11, 19 and 22, accounted for 49% of all gains in these patients. All three patient relapse samples carried TP53 mutations, two of which were present in the germline. In all three cases, no underlying hypodiploid clone was detected by DI or cytogenetic analyses, making diagnosis difficult. A recent report by Carroll et al. demonstrated that a considerable proportion (25% or higher) of hypodiploidy in children with B-ALL may have been overlooked in previous studies due to the presence of only a doubled hypodiploid population, mistakenly interpreted as typical high hyperdiploidy associated with a favorable risk . In this cohort, the chromosome number in high hyperdiploidy was mostly in the range of 52 ~ 59, which could overlap with masked hypodiploidy. For masked hypodiploid cases, the MLPA P036 kit results, alongside DNA index, can detect the specific gain or loss of each chromosome. LOH can also be confirmed by STR. TCF3, located to 19p, is rearranged with several genes in childhood ALL. The most frequent is TCF3-PBX1 and rarely the poor risk TCF3-HLF . We observed 19p loss in all TCF3-PBX1 and TCF3-HLF cases. TCF3 has also been identified to be rearranged with ZNF384, a novel fusion recently identified – . In cases with 19p deletions without TCF3-PBX1 or TCF3-HLF detected by RT-PCR or cytogenetics, 19p deletions may point to other TCF3 fusions. TCF3-ZNF384 fusions are also frequently associated with CD10 loss, with the presence of CD13 and CD33 , , . These two characteristics are useful for its identification by RT-PCR. In our cohort, patients with iAMP21 and KMT2A fusions had an inferior 5-year EFS and OS in comparison to patients with ETV6-RUNX1 or high hyperdiploidy. Patients with hypodiploidy also had an inferior 5-year EFS and OS, although most of them were not identified at the time of diagnosis. The outcome for patients with iAMP21-ALL may be improved if detected at diagnosis, so that they may be treated with more intensive chemotherapy. No events were seen in patients with PAX5 AMP , while patents with IKZF1 plus showed a trend towards inferior EFS and OS, although the P -value was not significant. IKZF1 deletions showed a trend towards poorer clinical outcomes, as observed in a number of other studies , , . Due to the relative small case numbers in this study, larger studies are indicated in Taiwan in order to evaluate the clinical impact of these genetic alterations in Taiwan. In conclusion, MLPA and DNA index together can rapidly provide reliable information for identification of aneuploidy of childhood B-ALL. Using these methods, diagnosis of aneuploidy in Taiwan might be improved particularly among those cases currently classified within unknown subtype of B-cell ALL, and especially those without metaphases or normal karyotype. STR provides a simple method to demonstrate LOH if masked hypodiploidy is suspected. Other important abnormalities such as IKZF1 deletions, IKZF1 plus and ERG deletions can also be identified by MLPA. These tools are helpful for the diagnosis of some important subtype of ALL. Supplementary Information. |
10-year follow-up of interventional electrophysiology: updated German survey during the COVID-19 pandemic | e544b076-2248-4bd5-9ec8-6698c8112bd7 | 9446632 | Physiology[mh] | Over the last decades, cardiac electrophysiology has become a pivotal subspecialty of cardiology with growing numbers of catheter ablations every year . In many patients with supraventricular tachycardias (SVT) or atrial fibrillation (AF), catheter ablation is considered first-line therapy . The gradual increase in the number of yearly performed catheter ablations is, e.g. portrayed in mandatory quality reports based on the German operational and procedural key system (OPS) with currently about 90,000 catheter ablations in Germany each year . To ensure overall quality, safety, and optimal patient care national and international standards as well as trained specialists in the field of cardiac electrophysiology are encouraged to match this development. In order that aspiring physicians in the field of cardiac electrophysiology receive proper training as heart rhythm specialists, national and international cardiology societies have developed training programs and curricula . To provide an overview and assess the current national status of physician training and patient care in cardiac electrophysiology including infrastructure, training conditions, and ablation procedures, we initiated this survey in 2010 and performed a 5-year follow-up in 2015 . This multi-centre observational study provides a second longer follow-up and overview of a decade of electrophysiological patient care and training comparing data to previous surveys from 2010 and 2015. It is of particular interest as it presents data of a time period in which the worldwide COVID-19 pandemic enforced lock-down measures with cancellation of many elective catheter ablations. Consulting the national legally mandatory quality reports of German hospitals, 340 centres were identified currently performing electrophysiological studies with the following reported OPS (operation and procedure code): 8–835.2 (radiofrequency (RF) ablation), 8–835.3 (irrigated RF ablation), 8–835.4 (ablation with other energy sources), 8–835.9 (MESH ablation), 8–835.a (cryo-ablation), and 8–835.8 (ablation with 3-D mapping). ( https://www.dimdi.de/dynamic/de/klassifikationen/ops/anwendung/zweck/index.html ). As more than one OPS code can be reported for a single ablation procedure (e.g., radiofrequency ablation plus 3D mapping-based ablation), the number of OPS given is not equal to the number of procedures performed. Centres coding for less than 30 ablation procedures a year were excluded to prevent the accidental inclusion of centres employing external electrophysiologists or coding OPS for externally performed procedures. Upon identification of the centres, we contacted the cardiology or interventional electrophysiology department by e-mail and/or phone to complete the same questionnaire that was utilized in previous surveys from 2010 and 2015 . Among the included parameters in the questionnaire were: type of hospital; staff numbers and functions in cardiology and electrophysiology, gender aspects, infrastructure, number and types of EP procedures, techniques used, imaging modalities, presence of or distance to cardiac surgery. Furthermore, more detailed information on protection methods of the esophagus during AF ablation was requested. Gathered data were anonymized and consequently analyzed using R-Studio Version 1.4.1106 (R. RStudio, PBC, Boston, MA). Of all the centres, coding more than 30 ablation procedures per year, 192 (56%) answered the survey and were included in this analysis (Fig. ). Responding centres included 34 (18%) university hospitals, 137 (71%) teaching hospitals (non-university hospitals involved in training of medical students), 19 (10%) non-teaching, and 2 (1%) private medical practices performing ablations in adjoining hospitals. The structure of interventional electrophysiology The electrophysiological departments were mainly part of a cardiology clinic (90%) with only 19 EP centres (11%) being independent with their own budget. A total of 106 centres (55%) were certified training centres for cardiac electrophysiological procedures by the German cardiac society (DGK). Heads of cardiological departments of 31 centres (16%) counted invasive electrophysiology as their main area of expertise. In 148 centres (77%), at least one catheter laboratory was exclusively used for invasive electrophysiology over 90% of the time. Thirty-five centres (18%) used two laboratories predominantly for EP procedures. 3-D mapping systems (CARTO ® n = 104; NavX ® n = 106; Rhythmia ® n = 29; CARTO ® and NavX ® n = 47) were available in 110 (57%) centres. 101 centres (53%) used the catheter laboratory also for all electrical device implantations, 12 (6%) centres in more than 50% of cases and 45 (23%) centres in less than 50% of cases. In the remaining centres ( n = 34; 18%), device implantations were exclusively performed in operating rooms. The primary operator implanting these devices was a cardiologist in 147 (77%) centres and a surgeon in 8 (4%). Both cardiologists and surgeons performed these procedures in the remaining 36 (19%) EP centres. Physicians involved in electrophysiology Altogether there were 219 heads (female: n = 9; 4%) of departments with 27 centres (14%) having more than one head of department (including head for interventional cardiology and electrophysiology) (Table ). Furthermore, 1424 consultants (“Oberarzt”, female: n = 338; 24%) and 3441 physicians in training (female: n = 1652; 48%) were employed. A total of 403 EP consultants (female: “Oberärztin” n = 75; 19%) were employed with 36 (19%) centres having only one and 146 centres (76%) having two or more EP consultants in their team. EP Consultants from 139 centres (72%) also performed coronary interventions (Table ). For EP fellows, there were a total of 432 (female: n = 144: 33%) training positions reported. In 46 (24%) centres, only one fellow was trained as a heart rhythm specialist. No less than 2 fellows were employed in 22 (11%) centres and at least 3 or more fellows in 51 (27%) centres. In contrast, 72 (38%) centres had no EP fellows (Fig. ). As primary operator, 549 (female: n = 126; 23%) EP consultants performed catheter ablations with only one EP consultant present in the cardiological team in 7 centres (4%). Of these primary operators, 203 (37%) were less than 40 years old, 214 (39%) between 40 and 50, and 132 (24%) more than 50 years old; 53 (10%) worked part-time. A median number of 377 catheter ablations per centre were performed in 2020 with two or more physicians present throughout most ablation procedures in 134 (70%) centres (Table ). Less than 100 catheter ablations were performed at 33 (17%) centres, and in 108 (56%) centres, at least 200 ablations were performed. At least 50 (75) PVI were documented in 133 (69%) centres ( n = 122; 64%, respectively); 59 centres (31%) performed less than 50 PVI and 25 (13%) centres were not ablating AF at all. Procedural data The reporting 192 centres performed a total of 76.304 EP procedures including 68.407 catheter ablations in 2020. Most of the centres obtained patient consent already before hospital admission: 39 (20%) centres in all cases; 78 (41%) in over 50% of the cases. (Table ). The most frequent arrhythmia treated by catheter ablation was AF ( n = 35.193; 51%) followed by SVT ( n = 14.045; 21%), atrial flutter ( n = 11.428; 17%), and ventricular tachycardias ( n = 7.641; 11%) (Fig. ). Left-sided accessory pathways were ablated by 176 (92%) centres, out of these, 140 centres (80%) primarily used a transseptal and 36 of the centres (20%) a retrograde approach. 149 centres (78%) performed left-sided VT ablations with the majority of these centres ( n = 88; 59%) using a trans-septal and 61 centres (41%) a retrograde approach to reach the left ventricle. Of note, VT were not ablated in 43 (22%) centres. If necessary, 44 centres (23%) reported to perform epicardial ablations (Table ). The energy source predominantly used by the 167 centres (87% of all participating centres) ablating AF was point-by-point radiofrequency current with 64% of all PVI vs. 34% cryo-balloon ablations (Table ). The proportion of cryo-balloon ablation clearly correlated with the centres´ total number of PVI, the larger the volume, the higher the proportion of RF ablation (Fig. ). In persistent AF, the primary ablation strategy reported was PVI in 147 centres (88%) with a minority of the centres performing PVI plus linear ablation ( n = 4; 2%) or substrate modification using, e.g. defragmentation ( n = 17; 10%). In 72 centres (43%), imaging before AF ablation was routinely performed (MRI in 11 (7%), CT in 35 (21%), rotational angiography in 17 (10%); 3-D Echo in 9 (12%)). Consecutive atrial arrhythmias after AF ablation were ablated in 147 (77%) of participating centres. Sedation with propofol was the preferred standard approach (95%). Only a small number of centres ( n = 7; 4%) performed ablations under general anesthesia. Though rare, an atrio-esophageal fistula remains one of the most feared late complications after PVI because of its often lethal outcome . Therefore, the vast majority of centres (88%) reported using strategies for esophageal protection including: prescribing H 2 blockers (78%) after ablation , reducing energy while ablating along the posterior wall (68%) and the use of esophageal temperature probes (54%) . Cardio-surgical back-up was available in-house in 64 (38%) of the centres performing AF ablations. If not available in-house, the distance to the next hospital having a cardio-surgical unit ranged from 1 to 150 km (mean: 35 ± 31 km). Surgical AF ablations were performed in 44 (26%) centres with 10 (6%) centres performing surgical AF ablations as stand-alone operations. Training centre requirements The requirements to be accredited as an EP training centre according to the European Heart Rhythm Association (EHRA) and the German Cardiac Society (DGK) are illustrated in Table . Only a quarter ( n = 48) of the responding centres fulfilled the requirements provided by the EHRA or DGK ( n = 47; 24%; for the requirement of 75 AF ablations/per year n = 36 (19%)). The electrophysiological departments were mainly part of a cardiology clinic (90%) with only 19 EP centres (11%) being independent with their own budget. A total of 106 centres (55%) were certified training centres for cardiac electrophysiological procedures by the German cardiac society (DGK). Heads of cardiological departments of 31 centres (16%) counted invasive electrophysiology as their main area of expertise. In 148 centres (77%), at least one catheter laboratory was exclusively used for invasive electrophysiology over 90% of the time. Thirty-five centres (18%) used two laboratories predominantly for EP procedures. 3-D mapping systems (CARTO ® n = 104; NavX ® n = 106; Rhythmia ® n = 29; CARTO ® and NavX ® n = 47) were available in 110 (57%) centres. 101 centres (53%) used the catheter laboratory also for all electrical device implantations, 12 (6%) centres in more than 50% of cases and 45 (23%) centres in less than 50% of cases. In the remaining centres ( n = 34; 18%), device implantations were exclusively performed in operating rooms. The primary operator implanting these devices was a cardiologist in 147 (77%) centres and a surgeon in 8 (4%). Both cardiologists and surgeons performed these procedures in the remaining 36 (19%) EP centres. Altogether there were 219 heads (female: n = 9; 4%) of departments with 27 centres (14%) having more than one head of department (including head for interventional cardiology and electrophysiology) (Table ). Furthermore, 1424 consultants (“Oberarzt”, female: n = 338; 24%) and 3441 physicians in training (female: n = 1652; 48%) were employed. A total of 403 EP consultants (female: “Oberärztin” n = 75; 19%) were employed with 36 (19%) centres having only one and 146 centres (76%) having two or more EP consultants in their team. EP Consultants from 139 centres (72%) also performed coronary interventions (Table ). For EP fellows, there were a total of 432 (female: n = 144: 33%) training positions reported. In 46 (24%) centres, only one fellow was trained as a heart rhythm specialist. No less than 2 fellows were employed in 22 (11%) centres and at least 3 or more fellows in 51 (27%) centres. In contrast, 72 (38%) centres had no EP fellows (Fig. ). As primary operator, 549 (female: n = 126; 23%) EP consultants performed catheter ablations with only one EP consultant present in the cardiological team in 7 centres (4%). Of these primary operators, 203 (37%) were less than 40 years old, 214 (39%) between 40 and 50, and 132 (24%) more than 50 years old; 53 (10%) worked part-time. A median number of 377 catheter ablations per centre were performed in 2020 with two or more physicians present throughout most ablation procedures in 134 (70%) centres (Table ). Less than 100 catheter ablations were performed at 33 (17%) centres, and in 108 (56%) centres, at least 200 ablations were performed. At least 50 (75) PVI were documented in 133 (69%) centres ( n = 122; 64%, respectively); 59 centres (31%) performed less than 50 PVI and 25 (13%) centres were not ablating AF at all. The reporting 192 centres performed a total of 76.304 EP procedures including 68.407 catheter ablations in 2020. Most of the centres obtained patient consent already before hospital admission: 39 (20%) centres in all cases; 78 (41%) in over 50% of the cases. (Table ). The most frequent arrhythmia treated by catheter ablation was AF ( n = 35.193; 51%) followed by SVT ( n = 14.045; 21%), atrial flutter ( n = 11.428; 17%), and ventricular tachycardias ( n = 7.641; 11%) (Fig. ). Left-sided accessory pathways were ablated by 176 (92%) centres, out of these, 140 centres (80%) primarily used a transseptal and 36 of the centres (20%) a retrograde approach. 149 centres (78%) performed left-sided VT ablations with the majority of these centres ( n = 88; 59%) using a trans-septal and 61 centres (41%) a retrograde approach to reach the left ventricle. Of note, VT were not ablated in 43 (22%) centres. If necessary, 44 centres (23%) reported to perform epicardial ablations (Table ). The energy source predominantly used by the 167 centres (87% of all participating centres) ablating AF was point-by-point radiofrequency current with 64% of all PVI vs. 34% cryo-balloon ablations (Table ). The proportion of cryo-balloon ablation clearly correlated with the centres´ total number of PVI, the larger the volume, the higher the proportion of RF ablation (Fig. ). In persistent AF, the primary ablation strategy reported was PVI in 147 centres (88%) with a minority of the centres performing PVI plus linear ablation ( n = 4; 2%) or substrate modification using, e.g. defragmentation ( n = 17; 10%). In 72 centres (43%), imaging before AF ablation was routinely performed (MRI in 11 (7%), CT in 35 (21%), rotational angiography in 17 (10%); 3-D Echo in 9 (12%)). Consecutive atrial arrhythmias after AF ablation were ablated in 147 (77%) of participating centres. Sedation with propofol was the preferred standard approach (95%). Only a small number of centres ( n = 7; 4%) performed ablations under general anesthesia. Though rare, an atrio-esophageal fistula remains one of the most feared late complications after PVI because of its often lethal outcome . Therefore, the vast majority of centres (88%) reported using strategies for esophageal protection including: prescribing H 2 blockers (78%) after ablation , reducing energy while ablating along the posterior wall (68%) and the use of esophageal temperature probes (54%) . Cardio-surgical back-up was available in-house in 64 (38%) of the centres performing AF ablations. If not available in-house, the distance to the next hospital having a cardio-surgical unit ranged from 1 to 150 km (mean: 35 ± 31 km). Surgical AF ablations were performed in 44 (26%) centres with 10 (6%) centres performing surgical AF ablations as stand-alone operations. The requirements to be accredited as an EP training centre according to the European Heart Rhythm Association (EHRA) and the German Cardiac Society (DGK) are illustrated in Table . Only a quarter ( n = 48) of the responding centres fulfilled the requirements provided by the EHRA or DGK ( n = 47; 24%; for the requirement of 75 AF ablations/per year n = 36 (19%)). Reporting data from German centres performing electrophysiological studies, this multi-centre observational study is able to describe clear trends in electrophysiology over the recent decade comparing data from 2010 , 2015 , and 2020. Most contacted clinics responded with a complete questionnaire. Collectively, there were 68.407 catheter ablations reported by the responding centres in 2020 illustrating a 39 and 105% increase in yearly performed ablations compared to survey data from 2015 and 2010 , respectively. This is in line with an increase in the number of hospitals performing EP studies in Germany and was observed despite the presence of the COVID-19 pandemic with many weeks of lock-down and cancellation of elective EP procedures in most centres. As training requirements differ not only in Europe but also in the U.S. it is difficult to determine an exact number of necessary ablation procedures needed to be an experienced EP centre . Reference publications are the curricula published by the German cardiac society (DGK) and the European Heart Rhythm Association (EHRA) as well as the 2017 HRS/EHRA/ECAS/APHRS/SOLAECE expert consensus statement on catheter and surgical ablation of atrial fibrillation . These recommendations are very similar, except the required ablation numbers in Europe being slightly higher. The EHRA (DGK) recommend that an EP centre ought to have a (moderate) quantity of at least 200 (250) EP studies and at least 150 (200) catheter ablations a year which was, however, fulfilled by only 58% (55%) of the responding centres. Besides, the EHRA requires a centre to have a cardio-surgical unit which was present in only 38% of the participating German centres. Altogether, only a quarter of responding centres fulfilled all EHRA or DGK criteria. Of note, only 16% of the centres fulfilled the requirement of the DGK of always having two physicians present during catheter ablation procedures. Analyzing these results and comparing them with data from 2010 and 2015, there is still a relevant need to enhance the quality of EP physician training and for collaboration between centres to provide high-quality electrophysiological patient care. Because many centres do not fulfill requirements set by the EHRA and/or DGK, one can assume there is a scarcity of training opportunities for physicians aspiring a career in EP. However, a centre accreditation by neither institution reflects the capacity of a single operator and is only supposed to show which centre would have met certain requirements agreed upon by a committee of experienced electrophysiologists. Very recently, a survey of members of the “Young DGK” (median age 33 ± 3.3 years) regarding training opportunities for cardiology was published . The majority wished more electrophysiological training opportunities with 50% of cardiological fellows reporting not to receive any EP training . These results directly reflect to our survey with still more than a third (38%) of the responding centers reporting to have no EP fellows at all. This has remained almost unchanged throughout the last decade (2010: 34%; 2015: 33%). Thus, the present situation of German cardiac electrophysiology clearly illustrates (1) an increasing number of catheter ablations in the presence of (2) the necessity of more and better training opportunities. In the presence of increasing ablation numbers with growing complexity and novel ablation technologies, a high degree of sub-specialization is needed to perform these ablations. It is therefore surprising that (1) only 11% of the centres have an independent EP department (with/without its own budget) and (2) the majority of EP consultants also performs PCI on a routine basis. This proportion even increased in comparison with data from 2015 (63 vs. 72%). One may speculate that these aspects as well as the above-mentioned limited training opportunities require more dedicated independent EP centres in the future. Despite an overall increase of female physicians in most cardiological specialties, only less than 10% choose a career in EP . Addressing this disparity, a survey by Abdulsalam et al. determined factors influencing physicians in training and career planning. Of the responding participants having an interest in EP, the vast majority that ultimately chose to train as a heart rhythm specialist were men (84 vs. 16%). As potential reasons women reported, e.g. radiation concerns and a perceived “old boys’ club” culture with discrimination/harassment concerns . This issue is also addressed by a survey of Estner et al. showing a large gap between male and female physicians in training (63 vs. 37%) as well as consultants (86 vs. 14%). This corresponds to results from our national survey showing that the proportion of female fellows as well as female EP consultants remain distinctly low with even a decrease in female EP fellows as compared to 2015 (38%; 2020: 33%) and an almost unchanged number of employed female EP consultants (2015: 17%; 2020: 19%). Addressing this issue and improving the training and work environment (e.g., working part-time for both genders, childcare support) will be pivotal to change this disparity in the future. Besides, implementing certain mentorship programs would be of great interest. As it was seen in 2010 and 2015, PVI remains the most performed catheter ablation procedure even showing an increase in number compared to prior results (2010: 35%; 2015: 47%, 2020: 51%). Considering that during the COVID-19 pandemic more elective PVI were cancelled than urgent ablations such as VT ablations, the true number of scheduled PVI may have been even higher. Nevertheless, the trend of an un-proportional increase in PVI as compared to all other ablation procedures over the last decade is demonstrated by survey comparisons from 2010, over 2015 to 2020 (Fig. ). In contrast to AF, the number of supraventricular tachycardia (SVT) and atrial flutter ablations remained relatively constant over the years with 22% (32%) and 20% (25%) in 2015 (2010) and 21 and 17% in 2020, respectively. Following the trend in AF ablations and the demography of western countries, one would not be surprised if the next decade will result in PVI accounting for 2/3 of all catheter ablations. Of note, no relevant change is seen regarding the proportion of RF versus cryo-ablations. Most ablations were performed with point-by-point RF ablation (2015: 63%; 2020: 64%) as compared to the cryo-balloon technology (2015: 33%; 2020: 34%). Besides, we could clearly show the less experienced a centre is the more the cryo-balloon is used (Fig. ). This is in line with the observations of a relevant and increasing portion of centres not ablating consecutive left atrial arrhythmias after PVI compared to 2015 (19 vs. 23% in 2020) . This also most probably reflects the lack of experienced electrophysiologists able to treat consecutive left-sided atrial arrhythmias and the increased use of the technically less demanding cryo-balloon-based ablation by less experienced centres . The STAR AF II Trial and a recent sub-study by Sanchez-Somonte et al. showed that even patients with complete linear block and/or ablation of fractionated electrograms after PVI did not have a better outcome regarding recurring AF. This correlates to our analysis seeing most centres performing PVI only as their first treatment approach for patients with persistent AF as recommended by current guidelines. As the number of EP procedures increases each year, our observational study is supposed to offer interesting insights into current electrophysiological training and treatment concepts and may help recognizing certain issues that need to be addressed in the future. Besides, further studies setting safety, efficacy, and overall treatment quality in relation to the amount of EP procedures performed per year per centre would give interesting insights and may offer perspectives regarding patient care and physician training. Certainly, as in the previous studies from 2010 and 2015, not all centres performing EP studies responded and as coding data are not continuously reliable probably not all centres were identified. Nevertheless, our study does include most centres as well as ablations (75%) performed in Germany in 2020 and gives the chance to observe trends over a decade of electrophysiological advances. As the responding centres account for most ablations performed in 2020, smaller clinics might not be well represented in this survey, leading to the possibility of a slight over-estimation of median number of ablations per centre. To prevent the over-estimation of small centres where fewer catheter ablations are performed, we excluded centres coding for less than 30 ablations per year. This again might over-estimate the percentage of possible training centres fulfilling all requirements by the DGK and EHRA. Data about complications and specific outcome would have been of interest (e.g., safety of certain procedures corresponding to the amount of performed procedures a year). But as this survey was devised to assess structural conditions in electrophysiological patient care and physician training, these data are not available. The present multi-centre observational study demonstrates a distinct rise in the need for electrophysiological treatment with increasing numbers of EP centres and performed ablation procedures as compared to 2010 and 2015. Only about a quarter of the centres fulfilled requirements of the EHRA and DGK for EP training centres, respectively. Training positions for physicians in electrophysiology have not adapted to this rising demand and have remained constant over the years. Women are still only scarcely represented in the field of interventional electrophysiology. PVI with point-by-point radiofrequency current (RF) as the mainly used ablation strategy remains the most performed ablation. |
Effects of anesthesia on cerebral oxygen saturation and prevention of brain injury during carotid endarterectomy | c549ec6f-207e-454e-96ae-46d6554156b2 | 11829386 | Surgical Procedures, Operative[mh] | Carotid endarterectomy (CEA) is a proven intervention for reducing the risk of ipsilateral ischemic stroke, particularly in symptomatic patients and in select asymptomatic patients with high-risk carotid stenosis features . However, during the procedure, temporary occlusion of the carotid artery may lead to inadequate cerebral perfusion, resulting in a 3% risk of perioperative cerebral infarction . This risk is largely caused by inadequate cerebral perfusion during the temporary carotid artery occlusion . Therefore, optimizing intraoperative brain protection strategies has become a key focus in CEA research in recent years . In CEA surgery, sevoflurane and propofol are commonly used general anesthetics. Propofol, as an intravenous anesthetic, is widely recommended for neurosurgical procedures due to its neuroprotective effects . Studies have shown that propofol may help improve patient outcomes by reducing oxidative stress-induced damage, although these effects have primarily been observed in cases of mild ischemic injury . Meanwhile, sevoflurane, an inhalation anesthetic, at 1.0-1.5 minimum alveolar concentration (MAC), can protect endothelial cells from ischemia/reperfusion injury, maintain cerebral oxygen supply-demand balance, and reduce cerebral metabolic rate (CMRO2), potentially offering brain protection . The complementary mechanisms suggest a potential synergistic effect when the two anesthetics are combined, particularly during the critical periods of carotid artery clamping. Additionally, near-infrared spectroscopy (NIRS) has gained attention as a superior method for monitoring regional cerebral oxygen saturation (rSO 2 ) and assessing the likelihood of cerebral ischemia . NIRS offers advantages such as convenience, rapidity, non-invasiveness, and continuous monitoring, enabling timely assessment of cerebral blood supply, vascular injury, and dynamic changes in cerebral tissue oxygen metabolism. Therefore, NIRS can serve as an “early warning” for cerebral hypoxia, assisting anesthesiologists in making timely and effective decisions . Regional saturation of oxygen (rSO 2 ) reflects the balance of oxygen supply and demand in brain tissue, and is related to factors such as arterial oxygen saturation, hemoglobin, cerebral blood flow (CBF), and cerebral metabolic rate (CMRO2). Therefore, monitoring rSO 2 is a non-invasive and effective method for observing changes in cerebral blood flow during the induction of general anesthesia and predicting the occurrence of cerebral ischemia. The objective of this study is to compare the effects of intravenous anesthesia and combined sevoflurane anesthesia on rSO 2 levels during CEA to evaluate differences in brain protection between these two anesthetic approaches, thereby providing a basis for clinical anesthesia selection. This study hypothesizes that combined sevoflurane and propofol anesthesia will offer superior protection against ischemic brain injury compared to intravenous anesthesia alone, based on the complementary mechanisms of these anesthetics in maintaining cerebral oxygenation and reducing oxidative stress. Study design This prospective randomized controlled trial aims to compare the effects of intravenous anesthesia (Group IVA) and combined sevoflurane anesthesia (Group CIA) on cerebral oxygenation during CEA. The study will focus on comparing the impact on rSO 2 , particularly during carotid artery clamping. The hypothesis is that combined sevoflurane anesthesia will provide better brain protection by improving cerebral oxygenation. The primary comparison will assess the trends of rSO 2 changes at different time points between the two groups. General Information Fifty-four patients (43 males, 11 females; aged 44 to 80 years) undergoing unilateral CEA surgery in the First Affiliated Hospital of Xinjiang Medical University were enrolled as participants. They were randomly assigned, using a random number table method, to two groups, A and B, with 27 patients in each group. The study was approved by the Ethics Committee of the First Affiliated Hospital of Xinjiang Medical University. The final version of the experimental protocol, informed consent form, researcher manual, and Clinical Trial Observation Form (CRF) were developed and revised in accordance with the guidelines of the Ethics Committee. Inclusion Criteria: Patients classified as grade II or III according to the criteria of the American Society of Anesthesiologists (ASA); no restrictions on age or sex; unilateral carotid artery stenosis indicated by transcranial angiography (≥ 70% stenosis in asymptomatic patients or ≥ 50% stenosis in symptomatic patients); capable of autonomous behavior and voluntarily signing informed consent. Exclusion Criteria: Acute phase of cerebrovascular disease; carotid artery occlusion; non-visualization of distal carotid artery stenosis; persistent neurological deficits; long-term use of sedatives or antidepressants; systemic consumptive diseases; severe arrhythmias; myocardial infarction, heart failure, or poorly controlled severe hypertension; severe diseases of the respiratory system. Anesthesia methods Induction of anesthesia Anesthesia induction was performed by intravenous injection of midazolam at 0.05–0.1 mg/kg, etomidate at 0.1–0.3 mg/kg, rocuronium bromide at 0.6 mg/kg, and sufentanil citrate at 1 µg/kg. After successful tracheal intubation, adjustments were made to the oxygen flow rate, respiratory ratio, tidal volume, and ventilation frequency. Maintenance of anesthesia Group IVA: Propofol at 4–6 mg/kg − 1 h − 1 and remifentanil at 0.1–0.3 µg/kg − 1 min − 1 were pumped continuously until the end of the surgery. Group CIA: Continuous inhalation of sevoflurane at 1 MAC and continuous pumping of propofol at 2–4 mg/kg − 1 h − 1 and remifentanil 0.1–0.2 µg/kg − 1 min − 1 were used. Sevoflurane inhalation was stopped after carotid artery exposure and was replaced by continuous infusion of propofol at 4–6 mg/kg − 1 h − 1 and remifentanil at 0.1–0.3 µg/kg − 1 min − 1 , until the end of surgery. Both groups received intermittent intravenous injections of rocuronium bromide at 0.15 mg/kg. The anesthetic drugs and respiratory parameters were adjusted intraoperatively to maintain the end-tidal carbon dioxide at 35–40 mmHg, BIS value at 40–60, and nasopharyngeal temperature at 36–37 °C using automatic warming blankets. Intraoperative fluid replacement followed the 4-2-1 rule to maintain blood volume, with efforts made to maintain Hct around 30%. Intraoperative hemodynamic management Adjustments were made in correspondence with different stages of CEA surgery. (1) From the start of surgery to clamping of the carotid artery (common, internal, and external carotid arteries), the hemodynamic parameters of the patients were maintained within a ± 10% fluctuation range of the baseline values. (2) During temporary clamping of the carotid artery (common, external, and internal carotid arteries) to block blood flow, metaraminol (Aramine, Akorn Pharmaceuticals) was pumped intravenously to maintain the hemodynamic parameters within a fluctuation range of + 20% of the baseline values. (3) After carotid artery exposure, the hemodynamic parameters were maintained within a fluctuation range of 0 to -10% of the baseline values. Cerebral oxygen saturation and hemodynamics record To ensure the accuracy of cerebral oxygen saturation monitoring using the INVOS 5100 C cerebral oximeter (Somanetics, Troy, MI, USA), the sensor needs to be securely fixed on the patient’s forehead. Since the surgical area is located on one side of the neck, signal instability or interference may occur during the procedure. Therefore, it is essential to optimize the placement of the device and properly secure the sensor to prevent detachment or displacement, ensuring stable and accurate monitoring data. On this basis, heart rate (HR), mean arterial pressure (MAP), and regional cerebral oxygen saturation (rSO 2 ) were recorded at various time points to compare trends over time, including 5 min before anesthesia induction (T0), 5 min before carotid artery clamping (T1), immediately after clamping (T2), 5 min after clamping (T3), 10 min after clamping (T4), 15 min after clamping (T5), and 15 min after restoration of carotid artery blood flow (T6). All data were obtained through continuous monitoring, and 1-minute averages were extracted as research data. Additionally, arterial and venous blood samples were collected at T1, T6, and 24 h after surgery (T7) for arterial blood gas analysis and S100-β protein monitoring. Statistical methods SPSS 21.0 statistical software was used for analysis. Measurement data are expressed as mean ± standard deviation. Intra-group comparisons were performed using paired t-tests, and inter-group comparisons were made using repeated measures ANOVA. Count data were compared using chi-square tests. For the repeated measures ANOVA, the assumptions of sphericity and normality were tested, and the variances of the differences between the groups of interest were equal, satisfying the assumption of sphericity, and the variable was normally distributed at all levels. A p-value < 0.05 was considered statistically significant. This prospective randomized controlled trial aims to compare the effects of intravenous anesthesia (Group IVA) and combined sevoflurane anesthesia (Group CIA) on cerebral oxygenation during CEA. The study will focus on comparing the impact on rSO 2 , particularly during carotid artery clamping. The hypothesis is that combined sevoflurane anesthesia will provide better brain protection by improving cerebral oxygenation. The primary comparison will assess the trends of rSO 2 changes at different time points between the two groups. Fifty-four patients (43 males, 11 females; aged 44 to 80 years) undergoing unilateral CEA surgery in the First Affiliated Hospital of Xinjiang Medical University were enrolled as participants. They were randomly assigned, using a random number table method, to two groups, A and B, with 27 patients in each group. The study was approved by the Ethics Committee of the First Affiliated Hospital of Xinjiang Medical University. The final version of the experimental protocol, informed consent form, researcher manual, and Clinical Trial Observation Form (CRF) were developed and revised in accordance with the guidelines of the Ethics Committee. Inclusion Criteria: Patients classified as grade II or III according to the criteria of the American Society of Anesthesiologists (ASA); no restrictions on age or sex; unilateral carotid artery stenosis indicated by transcranial angiography (≥ 70% stenosis in asymptomatic patients or ≥ 50% stenosis in symptomatic patients); capable of autonomous behavior and voluntarily signing informed consent. Exclusion Criteria: Acute phase of cerebrovascular disease; carotid artery occlusion; non-visualization of distal carotid artery stenosis; persistent neurological deficits; long-term use of sedatives or antidepressants; systemic consumptive diseases; severe arrhythmias; myocardial infarction, heart failure, or poorly controlled severe hypertension; severe diseases of the respiratory system. Induction of anesthesia Anesthesia induction was performed by intravenous injection of midazolam at 0.05–0.1 mg/kg, etomidate at 0.1–0.3 mg/kg, rocuronium bromide at 0.6 mg/kg, and sufentanil citrate at 1 µg/kg. After successful tracheal intubation, adjustments were made to the oxygen flow rate, respiratory ratio, tidal volume, and ventilation frequency. Maintenance of anesthesia Group IVA: Propofol at 4–6 mg/kg − 1 h − 1 and remifentanil at 0.1–0.3 µg/kg − 1 min − 1 were pumped continuously until the end of the surgery. Group CIA: Continuous inhalation of sevoflurane at 1 MAC and continuous pumping of propofol at 2–4 mg/kg − 1 h − 1 and remifentanil 0.1–0.2 µg/kg − 1 min − 1 were used. Sevoflurane inhalation was stopped after carotid artery exposure and was replaced by continuous infusion of propofol at 4–6 mg/kg − 1 h − 1 and remifentanil at 0.1–0.3 µg/kg − 1 min − 1 , until the end of surgery. Both groups received intermittent intravenous injections of rocuronium bromide at 0.15 mg/kg. The anesthetic drugs and respiratory parameters were adjusted intraoperatively to maintain the end-tidal carbon dioxide at 35–40 mmHg, BIS value at 40–60, and nasopharyngeal temperature at 36–37 °C using automatic warming blankets. Intraoperative fluid replacement followed the 4-2-1 rule to maintain blood volume, with efforts made to maintain Hct around 30%. Anesthesia induction was performed by intravenous injection of midazolam at 0.05–0.1 mg/kg, etomidate at 0.1–0.3 mg/kg, rocuronium bromide at 0.6 mg/kg, and sufentanil citrate at 1 µg/kg. After successful tracheal intubation, adjustments were made to the oxygen flow rate, respiratory ratio, tidal volume, and ventilation frequency. Group IVA: Propofol at 4–6 mg/kg − 1 h − 1 and remifentanil at 0.1–0.3 µg/kg − 1 min − 1 were pumped continuously until the end of the surgery. Group CIA: Continuous inhalation of sevoflurane at 1 MAC and continuous pumping of propofol at 2–4 mg/kg − 1 h − 1 and remifentanil 0.1–0.2 µg/kg − 1 min − 1 were used. Sevoflurane inhalation was stopped after carotid artery exposure and was replaced by continuous infusion of propofol at 4–6 mg/kg − 1 h − 1 and remifentanil at 0.1–0.3 µg/kg − 1 min − 1 , until the end of surgery. Both groups received intermittent intravenous injections of rocuronium bromide at 0.15 mg/kg. The anesthetic drugs and respiratory parameters were adjusted intraoperatively to maintain the end-tidal carbon dioxide at 35–40 mmHg, BIS value at 40–60, and nasopharyngeal temperature at 36–37 °C using automatic warming blankets. Intraoperative fluid replacement followed the 4-2-1 rule to maintain blood volume, with efforts made to maintain Hct around 30%. Adjustments were made in correspondence with different stages of CEA surgery. (1) From the start of surgery to clamping of the carotid artery (common, internal, and external carotid arteries), the hemodynamic parameters of the patients were maintained within a ± 10% fluctuation range of the baseline values. (2) During temporary clamping of the carotid artery (common, external, and internal carotid arteries) to block blood flow, metaraminol (Aramine, Akorn Pharmaceuticals) was pumped intravenously to maintain the hemodynamic parameters within a fluctuation range of + 20% of the baseline values. (3) After carotid artery exposure, the hemodynamic parameters were maintained within a fluctuation range of 0 to -10% of the baseline values. To ensure the accuracy of cerebral oxygen saturation monitoring using the INVOS 5100 C cerebral oximeter (Somanetics, Troy, MI, USA), the sensor needs to be securely fixed on the patient’s forehead. Since the surgical area is located on one side of the neck, signal instability or interference may occur during the procedure. Therefore, it is essential to optimize the placement of the device and properly secure the sensor to prevent detachment or displacement, ensuring stable and accurate monitoring data. On this basis, heart rate (HR), mean arterial pressure (MAP), and regional cerebral oxygen saturation (rSO 2 ) were recorded at various time points to compare trends over time, including 5 min before anesthesia induction (T0), 5 min before carotid artery clamping (T1), immediately after clamping (T2), 5 min after clamping (T3), 10 min after clamping (T4), 15 min after clamping (T5), and 15 min after restoration of carotid artery blood flow (T6). All data were obtained through continuous monitoring, and 1-minute averages were extracted as research data. Additionally, arterial and venous blood samples were collected at T1, T6, and 24 h after surgery (T7) for arterial blood gas analysis and S100-β protein monitoring. SPSS 21.0 statistical software was used for analysis. Measurement data are expressed as mean ± standard deviation. Intra-group comparisons were performed using paired t-tests, and inter-group comparisons were made using repeated measures ANOVA. Count data were compared using chi-square tests. For the repeated measures ANOVA, the assumptions of sphericity and normality were tested, and the variances of the differences between the groups of interest were equal, satisfying the assumption of sphericity, and the variable was normally distributed at all levels. A p-value < 0.05 was considered statistically significant. No statistically significant differences were observed between the two patient groups in terms of age, sex, comorbidities, arrhythmias, and intake and output volumes (all P > 0.05, Table ). The arrhythmias refer to newly observed or transient events during the study period. No statistically significant differences in rSO 2 on the operated side were found between the two groups at T0 and T6 ( P > 0.05). However, at T1, T2, T3, T4, and T5, the rSO 2 of Group CIA patients was significantly higher than that of Group IVA patients ( P < 0.05). Compared with the T2 values, the rSO 2 values at the T3, T4, and T5 time points were markedly increased ( P < 0.05). Blood pressure (MAP) and heart rate (HR) changes from T0 to T6 were consistent between the two groups, with no significant differences (P > 0.05). Although MAP was maintained within ± 20% of baseline, transient increases were observed during carotid artery clamping, likely due to strategies for maintaining cerebral perfusion. These fluctuations showed no significant differences between groups, indicating effective and consistent blood pressure control in both groups. The trends of changes in rSO 2 , HR, and MAP are shown in Figs. , , and ; Table . Comparison of the differences in rSO 2 (ΔrSO 2 ) at T1, T2, and T5 with the baseline (T0) values showed a significantly greater increase in rSO 2 in Group CIA ( P < 0.05, Table ). There was a difference in S100-β protein levels between the two groups at T6 (t = 2.491, p = 0.016). At T1, a highly significant ( P < 0.01) difference in the pH was observed (t = 7.274, P = 0.009). No significant differences in the related indicators were found at the remaining time periods (Table ; Fig. ). Follow-up after surgery: One patient in Group IVA developed contralateral large-area cerebral infarction after surgery, with no improvement after treatment. The possibility of contralateral carotid artery thrombus shedding to cause cerebral infarction was considered. Another patient in Group IVA experienced mild cerebral infarction on the operated side after surgery, manifested as grade 3 muscle strength on the contralateral limb, with no permanent neurological damage after treatment. No abnormalities were observed in Group CIA, and the prognosis of all patients was good. Although we were unable to fully collect follow-up data at 3, 6, 12, and 24 months post-surgery due to the impact of the COVID-19 pandemic, the available follow-up data showed no specific adverse reactions or recurrence in either group at these time points. The characteristic feature of cerebral blood supply is that even if one or two branches of the nutrient vessels are impaired and the cerebral perfusion pressure varies within a specific range, cerebral autoregulation (CA) can still maintain a constant supply of blood to the brain, preventing hypoperfusion or hyperperfusion of the cerebral tissue . The measurement of cerebral blood flow (CBF) during CEA is crucial . While the velocity of CBF can be measured using intraoperative transcranial Doppler (TCD), it is challenging to do so in neurosurgical operating rooms. Therefore, NIRS monitoring of regional cerebral oxygen saturation is an attractive option. It is a non-invasive method for the evaluation of the balance between cerebral oxygen supply and demand. The near-infrared irradiation shows a high degree of penetration through the scalp and skull to reach deep brain tissue, allowing continuous measurement of the oxygen saturation in arterial blood, venous blood, the anterior cerebral artery, frontal lobe tissue, and other tissue layers. When determining factors (arterial oxygen saturation, hemoglobin) remain constant, fluctuations in rSO 2 are attributed to changes in CBF . Therefore, by continuously monitoring the changes in rSO 2 (ΔrSO 2 ), it is possible to detect and prevent cerebral ischemia . Studies have reported that rSO 2 values showing a > 20% decrease from baseline are associated with a sensitivity of 80% and specificity of 82% for the occurrence of neurological disease . NIRS can identify the presence of cerebral ischemia in CEA patients during carotid artery clamping, with a sensitivity of 80% and specificity of 94%, 6.5 min faster than somatosensory-evoked potentials . In this study, rSO 2 was found to increase rapidly with the onset of anesthesia until temporary occlusion of the carotid artery. The rSO 2 remained higher than baseline levels at the end of surgery. This is due to an increase in FiO 2 after tracheal intubation, leading to increased oxygen supply, changes in MAP during surgery, and decreased CBF caused by the general anesthetics. A previous study of healthy volunteers showed a similar pattern of rSO 2 changes to those observed here . The effects of sevoflurane and propofol on CBF are different. Sevoflurane reduces the CBF at concentrations < 1.5 MAC but increases CBF at concentrations > 1.5 MAC, believed to be due to the vasodilatory effect of sevoflurane and the CA effect. Studies have shown that the middle cerebral artery blood velocity (MCABV) during propofol anesthesia is lower than that during sevoflurane anesthesia, consistent with the known effects of these drugs on CBF. This indicates that sevoflurane has inherent cerebral vasodilation properties, while propofol anesthesia reduces MCABV by 26% compared to the values in conscious patients, with the decrease in CBF matching the decrease in cerebral metabolism caused by propofol . Furthermore, it has been confirmed by positron emission tomography that the reductions in cerebral metabolism induced by propofol and sevoflurane in humans are similar, although propofol was associated with lower CBF and cerebral blood volume than sevoflurane . In the present study, the concentration of sevoflurane in Group CIA was approximately 1 MAC throughout the entire surgery, with an average sevoflurane concentration of 1.36 vol%. The rSO 2 continued to rise relative to the baseline, possibly due to cerebral vasodilation, increased oxygen supply, and decreased cerebral oxygen metabolic rate. Severe stenosis of the internal carotid artery can induce blood-brain barrier damage, impairing the autoregulatory functions of vasodilation or vasoconstriction of vessels on the surgical side. Park et al. reported that after switching from sevoflurane to propofol in CEA patients, the pressure at the clamped end of the surgical carotid artery increased, which they attributed to the vasodilatory effect of sevoflurane and the vasoconstrictive effect of propofol. In the present study, the ΔrSO 2 and maximum increase in Group CIA were significantly higher than those in Group IVA, possibly because the decrease in CBF with sevoflurane at concentrations < 1.5 MAC was greater than that with propofol. Interestingly, there was no difference in ΔrSO 2 between the two groups on the contralateral side during the perioperative period, while a significant difference was observed on the surgical side. This is related to differences in the drug response between the surgical and contralateral sides. We believe that the combined inhalation anesthesia used in Group CIA improved the regulatory capacity of the surgical side region. Kuzkov et al. . concluded that combined inhalation anesthesia was preferable for maintaining anesthesia during CEA, consistent with our results. In addition, although the statistical significance ( P < 0.05) of the 1–3% differences in rSO 2 between Group IVA and Group CIA at various time points indicates a consistent pattern favoring Group CIA, it is essential to assess whether these differences are clinically meaningful in the context of cerebral protection during CEA. rSO 2 values can vary among individuals due to baseline physiological differences, and even small increments in rSO 2 can be crucial in maintaining oxygen levels above ischemic thresholds, reducing the risk of perioperative neurological deficits. Studies have shown that maintaining rSO 2 above certain critical levels reduces the risk of perioperative neurological deficits . Therefore, even a 1–3% improvement in rSO₂ may help maintain safer oxygen levels during ischemia, and repeated small gains could potentially cumulatively reduce the duration and severity of cerebral hypoxia, thereby lowering the risk of brain damage. The formation of oxygen free radicals is believed to be related to neuronal damage caused by ischemia-reperfusion during CEA. Propofol is reported to inhibit the release of excitatory amino acids, reduce intracellular calcium influx, and scavenge oxygen free radicals, thereby reducing the neurotoxicity of excitatory amino acids, protecting cell membranes, and providing preventive and therapeutic effects on brain ischemia-reperfusion injury . One study observed a decreased incidence of neurological dysfunction in animals subjected to cerebral ischemia under sevoflurane anesthesia, with the neuroprotective effect persisting for 8 weeks after ischemia . In the present study , it was observed that S100-β protein levels increased to varying degrees during carotid artery clamping. A significant elevation in S100-β protein levels was closely associated with the occurrence of neurological symptoms, suggesting a strong link between its increase and ischemic injury to brain tissue. In this study, all patients maintained stable rSO 2 levels on the non-operated side, with no significant decreases observed. This may explain why the increase in S100-β protein observed was less pronounced compared to other studies, and it may also limit the applicability of our results in scenarios involving more severe decreases in rSO 2 levels. Additionally, although we took measures to reduce extracerebral contamination, it may still have impacted the monitoring results, particularly in interpreting S100-β protein levels, thus introducing potential confounding factors. Future studies should include a broader patient population to better understand the relationship between rSO 2 levels and increases in S100-β protein. This study found a significant difference in pH at T1 between the two groups ( P = 0.009), possibly due to differences in anesthesia regimens. The CIA group used sevoflurane in the early stage, which may have affected CO₂ clearance and acid-base balance through metabolic processes or interactions with alveolar ventilation. Differences in ventilatory and metabolic regulation between intravenous and inhalation anesthesia may also contribute to pH changes. Although PaCO₂ was not directly monitored, ETCO₂ was maintained at 35–40 mmHg through adjustments to ventilation parameters, a range generally considered sufficient to maintain normal PaCO₂. While pH changes may affect the oxygen dissociation curve, this typically requires larger fluctuations. The limited pH change observed in this study and the lack of significant rSO₂ differences at T1 (see Table ) suggest that the pH change had minimal impact on cerebral oxygenation. This study has several limitations. First, the relatively small sample size may limit the generalizability of the findings, and the lack of formal sample size calculation may affect the accuracy of statistical power, thereby limiting the interpretability of the results. Second, the study was conducted at a single center, and variations in practice patterns across different institutions could affect the applicability of the results. Additionally, the follow-up period was only 24 h, which does not allow for the assessment of the long-term effects of the anesthesia methods. Another limitation is that some potential confounding factors, such as patients’ baseline health status and differences in intraoperative management, were not fully controlled for in this study, which may impact the accuracy of the results. Although BIS values were maintained within 40–60, specific BIS data and fluctuation trends were not recorded or reported, which may limit the assessment of anesthesia depth consistency between groups. Future studies should record and analyze BIS changes to verify the consistency of depth control across different anesthesia regimens. Furthermore, we did not use tools such as Mini-Mental State Examination to assess postoperative delirium, agitation, and cognitive function. Additionally, integrating clinical outcome assessments with rSO 2 monitoring is crucial to further enrich the clinical significance of our study findings. Moreover, the monitoring PaO 2 and glucose levels play an important role in regulating oxidative stress and free radical production . Incorporating the analysis of PaO 2 and glucose values into CEA anesthesia management protocols is essential for optimizing cerebral oxygenation and reducing oxidative stress. We also recognize the importance of PCO2 in influencing cerebral rSO 2 . Although PCO2 was not monitored in this study, we understand its potential impact on the results and will consider including this parameter in future study designs. These additions will help to better understand the mechanisms of anesthesia on brain protection and optimize clinical strategies. In conclusion, this study suggests that both general intravenous anesthesia and combined inhalation anesthesia may help mitigate the decrease in rSO₂ caused by temporary clamping during CEA. Combined inhalation anesthesia showed a tendency toward improved rSO₂ levels compared to intravenous anesthesia, which might be associated with better outcomes. However, further studies are needed to confirm these findings. |
Comprehensive anaesthesia management strategies for orthognathic surgical procedure | 9d077b0c-2f95-4bee-b4e2-1fc7a3f585a3 | 11801685 | Dentistry[mh] | Orthognathic surgery is performed to correct malocclusion and facial asymmetry. Although it usually involves young healthy patients, it is important to bear in mind that facial deformities are sometimes a characteristic of many different syndromes. The aim of this paper is to analyze the main factors involved in the anaesthesia management of these patients. Orthognathic surgery is a surgical procedure that restores a patient's entire facial structure by repositioning the maxillary and mandibular bones to address dentofacial abnormalities. Due to its complexity, precision and planning are essential components of this surgery and are also key factors in the approach to anaesthesia management both overall and at specific stages of the procedure. - Surgical procedure Orthognathic surgery is a surgical intervention that corrects dentofacial deformities by moving the maxillary and mandibular bones to balance the facial characteristics of the patient. The procedure, known as bimaxillary orthognathic surgery, involves both maxillary and mandibular osteotomies that are performed either independently or simultaneously. - Maxillary surgery Maxillary orthognathic surgery is performed to realign the maxilla in order to achieve facial harmony and restore the functionality of the bone, which is essential for chewing, breathing, and speaking. Orthognathic surgery of the maxilla consists of cutting the maxillary bone in a procedure called a Le Fort I Osteotomy, in which the maxillary bone is advanced, retruded, lengthened, shortened, or rotated. - Mandibular surgery The most commonly performed orthognathic surgery on the mandible is mandibular advancement surgery. This procedure is necessary for individuals with a small, receding jaw in relation to the maxilla - a condition known as retrognathia or class II occlusion. In individuals with a protruding jaw, a procedure called mandibular setback is indicated. - Bimaxillary surgery In most orthognathic surgery patients, both the maxilla and mandible must be repositioned to achieve correct occlusion and facial harmony in a procedure known as maxillomandibular surgery, or bimaxillary orthognathic surgery. Once the maxillary and mandibular bones are in the desired position, they are fixed in place with titanium plates. Being a particularly complex procedure, precision and planning are key factors not only in the surgical technique itself, but also in the approach to anaesthesia management both overall and at specific stages of the procedure. We perform a systematic review of the literature regarding anesthesia management in orthognathic surgery. The review process involved a search in PUBMED for relevant articles, using keywords related to "anesthesia management", "orthognathic surgery", "perioperative care" and other related terms. The search included studies published until the current year. Inclusion criteria were defined to focus on key aspects of anesthesia management specifically for orthognathic surgery, including airway management, bleeding control, postoperative nausea and vomiting (PONV) prevention, antibiotic prophylaxis, analgesia, and deep venous thrombosis (DVT) prevention. Exclusion criteria were also established to omit studies that did not directly relate to the perioperative anesthesia management or those involving unrelated surgical procedures. Full texts of the selected articles were then thoroughly reviewed. Data extraction was performed to gather information on the methodologies, findings, and conclusions related to anesthesia management in the context of orthognathic surgery. The analysis aimed to identify common strategies and techniques, assess their effectiveness, and highlight areas where further research is needed. The systematic approach ensured a comprehensive understanding of the current best practices in anesthesia management for patients undergoing orthognathic surgery. After the systematic review of the included articles, we found the following points to be of relevant interest, in order to define orthognathic surgery perioperative management and recommendations 1. Airway 2. Intubation 3. Positioning, care of pressure sores and eye protection 4. Throat pack 5. Fluid management, oedema protection 6. Bleeding 7. Postoperative nausea and vomiting 8. Intra and postoperative analgesia 9. Antibiotic prophylaxis 10. DVT prophylaxis 11. Emergence and extubation - Airway Airway management is straightforward in most patients; however, direct laryngoscopy can be difficult in patients with retrognathia, maxillary protrusion, or limited mouth opening . Videolaryngoscopy (VL) facilitates vocal cord visualization. Intubation is faster with the McGrath videolaryngoscope (Aircraft Medical, Edinburgh. UK) compared with Macintosh or GlideScope (Verathon, Inc, Bothell, WA) laryngoscopy in patients with an anticipated difficult airway , and has also been shown to increase the success rate in patients with a simulated difficult airway . The McGrath VL also reduces the need for Magill forceps, probably because the airway does not need to be as carefully aligned as in conventional laryngoscopy, and VL provides a more direct route from the nasopharynx to the trachea, particularly in patients with an anterior larynx, thereby reducing the need for manipulation of the nasotracheal tube. - Intubation In maxillofacial surgery, intubation is performed through the nasal cavity. The tracheal tube is taped to the forehead to give a clear view of the face and facial features and allow the surgeon to correctly align the mandible and maxilla. Prior to insertion of the tracheal tube, the nasal cavity must be prepared with vasoconstrictors, such as oxymetazoline. Nasal intubation is not suiTable for long procedures, because suctioning and ventilation are difficult through the tube owing to its small calibre, shape and length. - Eye protection and perioperative pressure injuries Failure to close the patient's eyes can result in corneal ulceration, and pressure from surgical instruments or the surgeon’s fingers can cause corneal abrasion. To prevent these adverse events, the eyes must be lubricated and covered with eye pads. Patients undergoing orthognathic surgery are susceptible to pressure injuries due to the length of the procedure, so it is essential to take preventive measures. Even though young, healthy patients with no vascular pathology or diabetes are less likely to develop pressure injuries, it is advisable to protect the skin over bony prominences, especially in patients who are thin or underweight . The most common sites for pressure injury are the sacrum and the heel of the foot. - Throat packs Throat packs are commonly inserted by an anaesthetist or surgeon after induction of anaesthesia for dental, maxillofacial, nasal, or upper airway surgery. The purpose is to absorb blood and other secretions, debris, and tooth fragments in order to keep the airway clear before extubation. All team members must be aware of the positioning of the gauze in the pharynx . Some authors have put forward strategies to ensure that the pack is removed at the end of the surgical procedure. These include tying or suturing the pack to the endotracheal tube (not possible with nasotracheal tubes), leaving a portion of the pack protruding from the mouth, placing reminder labels, using a checklist, etc. The risk of pack retention is particularly high when there is a change of anaesthetist or other team member, so care must be taken in these circumstances. Some authors recommend performing direct laryngoscopy and/or suctioning the pharynx before extubation to confirm that the pack has been removed. In 2018, the Difficult Airway Society (DAS), the British Association of Oral and Maxillofacial Surgery (BAOMS), and the British Association of Otorhinolaryngology, Head and Neck Surgery (ENT-UK) published an evidence-based consensus statement containing a protocol for throat pack use. (Fig. ). - Fluid therapy and oedema management. Intraoperative fluid therapy is an essential part of anaesthesia, but it is important to bear in mind that the fluids used are not only a vehicle for drug administration, but are also drugs themselves. Conventionally, high volumes of crystalloids have been administered to patients undergoing major surgery to compensate for perioperative dehydration and intraoperative losses due to fluid shift into the “third space”. However, there is evidence that a positive fluid balance is associated with complications and even with an increased risk of mortality. Physiological fluid flow does not cause interstitial oedema; however, damage to the vascular endothelial barrier will result in a pathological fluid flow that produces oedema at the surgical site. The endothelial surface layer is made up of endothelial cells and the endothelial glycocalyx layer (composed of membrane-bound glycoproteins proteoglycans and glycosaminoglycans), which preserves transendothelial permeability, regulates inflammation, and prevents platelet aggregation and leucocyte adhesion. All these functions are lost in the event of fluid overload. Evidence has shown that patients are not usually hypovolemic after preoperative overnight fasting, and that even after prolonged preoperative fasting, cardiopulmonary healthy patients remain intravascularly normovolaemic. More liberal administration of crystalloids in healthy patients undergoing moderate-risk surgery leads to a better recovery profile compared with patients who received restricted amounts of the same crystalloids. Perioperative protection of the endothelial glycocalyx is a plausible strategy for the prevention of interstitial oedema . It is interesting to note that 30% of every litre of 0.9 % saline administered remains in the intravascular compartment after equilibration. Colloid administration is context sensitive, but when used in normovolaemic patients, 79% of gelatines and a 84% of hydroxyethyl starch remain in the intravascular compartment. Most fluid therapy studies have been performed in abdominal surgery, but an interesting paper by Nishimura et al . reported that infused fluid often moves from the intravascular to the interstitial space. The authors also observed that increases in infused fluid volume may increase intravascular pressure, leading to more outward fluid movement from the intravascular to the interstitial compartment, suggesting that this could contribute to postoperative oedema , particularly in the presence of anaesthesia and/or surgical stress . Fluid therapy also has an impact on hospital length of stay (LOS). Huamán et al . found that the administration of intravenous fluid was significantly associated with increased LOS, and that colloids, specifically hydroxyethyl starches, were significantly more likely to increase LOS compared with crystalloids. - Oedema. Several methods are more effective than restrictive fluid therapy in reducing postoperative oedema. 1. Cooling is highly effective against oedema and has no side effects. For this reason, intraoperative irrigation fluids are cooled and hialotherapy is used after surgery . 2. Anti-Trendelenburg positioning (at least 45º) could help prevent postoperative oedema and inflammation by reducing head and neck congestion. 3. The administration of corticosteroids is thought to inhibit mast cell production and secretion of cytokine, kinin and histamine. They also reduce the production of thromboxane and bradykinin, thus reducing blood vessel dilatation and permeability . Glucocorticoids, especially dexamethasone (> 0.15 mg/kg) not only shorten LOS, but also reduce postoperative oedema . Dexamethasone also has other advantages: it has an analgesic effect; it is thought to inhibit prostaglandin synthesis ; it protects against postoperative nausea and vomiting; and contributes to nerve healing after injury to the dental alveolar nerve . High-dose dexamethasone (> 0.2 mg/kg) has opioid-sparing effects and also decreased pain scores, and for this reason it is used in orthognathic patients . No side effect have been registered in our patients - Bleeding. The maxillomandibular region is highly vascularized and intraoperative blood loss is often significant, despite controlled hypotension during surgery . Although, according to the literature, 27% to 30% of patients undergoing bimaxillary osteotomy procedures require allogenic blood transfusions, none of our patients have so far required transfusion; however, haemostasis improves the view of the surgical field, and hence significantly reduces operating time. Hypotensive anaesthesia (lowering the patient´s blood pressure during anaesthesia) is effective in reducing surgical bleeding, and has been practiced for several decades . In hypotensive anaesthesia, baseline mean arterial pressure (MAP) is reduced by around 30%, although this percentage will vary depending on the patient’s pathology . Optimal analgesia is the key to blood pressure management, but other pharmacological strategies may also be required. Fentanyl and remifentanil are both used to achieve pain control, and perfusion of dexmedetomidine (an alpha 2 agonist) has recently been added to boost analgesia and lower blood pressure. The two main strategies for achieving hypotensive anaesthesia are: 1. Deep anaesthesia 2. Standard anaesthesia and administration of hypotensive drugs Blood pressure can be reduced with several different drugs. The ideal hypotensive drug should be easy to administer, have a short time of onset, an effect that disappears rapidly when administration is discontinued, rapid elimination, and no side effects. 3- Nitrates, though popular in the past, are now no longer used because of their side effects. 4. Beta-adrenergic receptor antagonists act by reducing cardiac output. Esmolol, which is metabolized by plasmatic esterase, is the drug of choice due to its rapid onset and short duration. 5. The calcium channel blocker clevidipine has favourable pharmacokinetic characteristics and is effective in reducing blood pressure. Tranexamic acid (trans-4-(aminomethyl) cyclohexane carboxylic acid) (TA) is a synthetic derivate of the amino acid lysine that competitively inhibits the activation of plasminogen to plasmin by binding to Kringle domains. Tranexamic acid is also a competitive inhibitor of tissue plasminogen activator. It blocks the lysine binding sites of plasminogen, resulting in inhibition of plasminogen and fibrin binding to plasminogen, and therefore impairment of fibrinolysis. It is distributed throughout the body and has a plasma half-life of 120 min. The best-known trial, Clinical randomisation of antifibrinolytics in significant haemorrhage (CRASH-2), assessed the effects of early administration of tranexamic acid in trauma patients with, or at risk of, substantial bleeding, and showed that administration of TA reduced bleeding-related mortality (14.5% vs 16%; relative risk (RR) 0.91, 95% CI 0.85-0.97; p =0.0035), but did not reduce transfusion requirements . Though the efficacy and safety of the drug have been established, there is no consensus about the ideal time of administration and the dose (perfusion, bolus), especially in orthognathic surgery. However, there is evidence that 10 mg/l is required for 80% inhibition of tissue activator activity, so this is the dosing regimen used in our department. TA is contraindicated in patients with kidney failure, thromboembolic diseases, haematuria, and in pregnant women. Dakir et al . confirmed that preoperative administration of an intravenous 10 mg/kg bolus of tranexamic acid reduced blood loss compared with placebo during surgery . - Postoperative nausea and vomiting. Postoperative nausea and vomiting (PONV) is a common adverse effect of anaesthesia and surgery, and occurs in up of 80% of high-risk patients. Nausea and vomiting are not only highly distressing, but are also associated with a longer stay in the post-anaesthesia care unit (PACU), increased health care costs and hospital re-admission . Patient-specific risk factors for PONV in adults include female sex, a history of PONV and /or motion sickness, non-smoking status, and young age . Previously, patients considered a low risk for PONV were given either no prophylaxis or only 1 prophylactic drug. This approach has now changed considerably because PONV risk scores only provide a rough estimate, and patients identified as low risk may still develop PONV. Furthermore, PONV scores do not take into account factors such as the emetogenic risk of the surgical procedure and inter-individual variability in antiemetic effectiveness . As mentioned above, certain types of surgery, including orthognathic procedures, may be associated with an increased risk of PONV. Several drugs with different mechanisms of action are available: corticosteroids (dexamethasone, methylprednisolone), antihistamines (dimenhydrinate, promethazine) and anticholinergics (scopolamine), neurokinin1 receptors (aprepitant, fosaprepitant, etc.), dopamine- 2 receptors (amisulpiride, droperidol, haloperidol, etc), and 5-hydroxitryptamine3 antagonists (palosetron, dolasetron, granisetron, ondansetron, etc.) . If general anaesthesia is required in orthognathic surgery, the use of total intravenous anaesthesia (TIVA) has been shown to reduce PONV. A meta-analysis of 229 randomized controlled trials in different surgical procedures concluded that TIVA offers a benefit in reducing the incidence of PONV compared with volatile anaesthesia . The use of propofol for anaesthesia (or sedation) is associated with a 3.5-fold reduction in the incidence of PONV in adults and a 5.7-fold reduction in children . Sub-hypnotic doses of propofol (20-40 mg) are also effective as a rescue treatment for PONV. Other strategies that could minimize PONV risk are nitrous oxide-sparing anaesthesia, reversal of neuromuscular blockade with sugammadex, and use of intravenous lidocaine and dexmedetomidine infusion for analgesia (due to the opioid-sparing effect of these drugs) . Opioid-free anaesthesia reduces PONV risk, but this benefit must be weighed up against the risk of inadequate analgesia, hypertension, and bleeding. Intraoperative fluid administration may affect the risk of PONV. For example, 10-30 ml/kg infusion of intraoperative crystalloids reduces the risk of PONV; however, crystalloids should be avoided in orthognathic surgery . The combination of ondansetron and dexamethasone is one of the most widely studied and utilized multimodal PONV prophylaxes. Many new drugs are being added to this combination, with promising results. Palosetron: this second generation 5-HT3 receptor antagonist has a 100-fold affinity to the 5-HT3 receptor and a terminal half-life of 40 h. PONV is significantly reduced for 72 hours after surgery, and it is usually used in high-risk female patients. Monotherapy for PONV prophylaxis is more effective than other 5-HT3 antagonists and dexamethasone, and its efficacy is comparable to aprepitant . Its main drawback, however, is its high cost. Aprepitant: This competitive neurokinin (NK-1) receptor antagonist was initially approved for the treatment of chemotherapy-induced nausea and vomiting, and is administered orally for this indication. However, a prodrug of aprepitant, fosaprepitant, with a half-life of 9-13 h and a duration of action of 40 h, is available for intravenous administration . Fosaprepitant is more effective in preventing PONV than ondansetron, and has the same efficacy as palosetron . Amisulpiride: Dopamine receptor antagonist and antipsychotic approved for prophylactic and rescue therapy of PONV. It is administered intravenously at a dose of 5-10 mg. A wide variety of nonpharmacologic techniques have been used to control emetic symptoms in the postoperative period, such as acupressure, acupuncture and transcutaneous electrical nerve stimulation. Alcohol pads applied under the nose are a highly cost-effective treatment for transient PONV in adults and children. There is no reliable evidence that aromatherapy reduces postoperative nausea and vomiting . Some authors recommend P6 acupoint stimulation for PONV prevention. This has no side effects and significantly reduces nausea, vomiting and the need for rescue antiemetic drugs. Aside from pharmacological strategies, it is important to bear in mind the role of throat packs and stomach aspiration prior to emergency surgery in reducing the risk of PONV. - Analgesia. Many head and neck procedures are not associated with severe postoperative pain, and pain management with non-steroidal anti-inflammatory drugs and paracetamol usually suffices. However, evidence shows that up to 21% of orthognathic patients continue to feel pain 1 year after surgery . Severe, acute, postoperative pain increases the risk of the pain becoming chronic . Multimodal analgesia is essential, and local anaesthetic infiltration by surgeons must be combined with opioids to decrease side effects and improve their adjuvant effect . Some authors recommend pre-emptive analgesia, arguing that postoperative pain can be prevented if certain analgesics are administered before the surgical stimulus; however, both peripheral and central stimuli must be blocked in order to achieve this goal. In 1983, Woolf defined pre-emptive analgesia as the treatment of pain before the surgical stimulus and the maintenance of this treatment during the high intensity harmful stimuli and during the postoperative period. Kissin suggested changing pre-emptive to preventive analgesia in order to limit the technique to the pre-operative period . In our practice, we administer NSAIDs (dexketoprofen 50 mg), acetaminophen 1 g, and dexamethasone 0.15 mg/kg immediately after induction to reduce the hypersensitisation of pain receptors, and then administer a further dose immediately after the surgery or immediately before admission to the PACU. Dexmedetomidine is an α2-adrenoreceptor agonist that primarily inhibits norepinephrine release and attenuates central nervous system excitation. The binding of postsynaptic receptors by α2-agonists leads to inhibition of sympathetic activity, which decreases blood pressure (BP) and heart rate (HR), and results in sedation and pain control. There is evidence that dexmedetomidine contributes to bleeding management. Although there is no evidence of its analgesic effect in orthognathic surgery, there is strong evidence in other procedures. It can be directly applied to the peripheral nervous system, causing a dose-dependent inhibition of C-fibres and α-fibres, and it acts on the locus coeruleus area, inhibiting nociceptive neurotransmission through the posterior horn of the spinal cord . Alpha-2 adrenergic receptors also act on the presynaptic membrane and inhibit the release of norepinephrine, which in turn induces hyperpolarization and inhibits pain signals to the brain. These drugs promote the release of acetylcholine from spinal interneurons, thereby increasing synthesis and releasing nitric oxide that regulates analgesia. Hialotherapy, which is the application of cold compression through a facemask at a regulated temperature of 15 ºC, significantly reduces pain and oedema at 48-72 hours . Postoperative pain management is usually achieved with NSAID, paracetamol, and corticoid combinations; opioids are rarely needed. - Antibiotic prophylaxis. Surgical site infection (SSI) is defined as an infection occurring within 30 days of surgery, or within 1 year in the case of patients receiving implants. The risk of SSI is dependent on factors such as the duration of surgery, the wound class, and the patient´s American Society of Anaesthesiology (ASA) classification. Increased risk of surgical site infection is generally accepted as an indication for antibiotic prophylaxis . Orthognathic surgery can be considered a clean contaminated surgery. These procedures would be expected to have higher infection rates than non-contaminated surgeries. Our review of the literature identified the following recommendations. 1. Preoperative antibiotics reduce the risk of surgical infection (weak recommendation). 2. There is limited evidence supporting postoperative antibiotic dosing. A 3 - day regimen of postoperative antibiotics may reduce the risk of surgical site infection compared to 1 day (weak recommendation). 3. Further research is required in this area . - DVT prophylaxis Although deep venous thrombosis (DVT) increases the risk in surgical procedures, it is uncommon in orthognathic surgery. The risk of DVT is determined by patient characteristics and the clinical setting. Prolonged periods in a head-up position can lead to venous pooling. Further risk factors, as well as the use of hormone therapy (oral contraceptives or hormone replacement therapy) and obesity, are found in orthognathic patients. Biochemical abnormalities (deficiencies of antithrombin, protein C or S, and activated protein resistance [factor Leiden V mutation, etc]) can also predispose patients to DVT. There are few published data on the incidence of DVT after oral and maxillofacial surgery. One report, based on the recollections of 103 consultants, estimated the incidence at 0.00035%. Van de Perre et al . reported 3 episodes of deep vein thrombosis in 2049 patients undergoing orthognathic surgeries, one of which resulted in pulmonary embolism ; however, these authors only measured parameters during the first 48 postoperative hours . Even though orthognathic patients usually have few comorbidities, it is mandatory to minimise the risk of DVT, so we believe that mechanical thromboprophylaxis (compression stockings) should be used, as recommended in NICE guidelines . - Emergence and extubation. After completion of surgery and removal of the throat pack, the oropharynx should be suctioned to remove traces of blood, clots, and debris in order to avoid laryngospasm. Positive pressure after extubation is also recommended. Patients must be transferred to a PACU for the first few postoperative hours. Optimizing the quality of care in orthognathic surgery will improve outcomes and speed up recovery. Careful anaesthesia management is directly involved in the result. We must know the procedure and the associated risks and protocolize the care and anticipate the potential problems. This review emphasizes the importance of a multidisciplinary approach to anesthesia management in orthognathic surgery. Effective airway management is crucial, especially in patients with anatomical challenges like syndromic patients. The use of videolaryngoscopy, particularly the McGrath videolaryngoscope, is highlighted for improving intubation success in difficult cases. Fluid management is also critical in order to prevent postoperative edema. The use of intraoperative cooling and corticosteroids like dexamethasone is effective in reducing complications and improving recovery. Intraoperative bleeding control, often managed through hypotensive anesthesia and tranexamic acid, is essential to facilitate the surgical procedure Finally, the prevention of postoperative nausea and vomiting (PONV) is crucial for patient recovery. The review supports the use of total intravenous anesthesia (TIVA) and dexamethasone as effective strategies in reducing PONV. In summary, successful anesthesia management in orthognathic surgery requires meticulous planning, advanced airway techniques, careful control of fluids and bleeding, and proactive prevention of postoperative complications. We could conclude in several items. 1.- Anaesthesia management in orthognathic surgery is critical due to the procedure´s complexity and the potential presence of syndromic features. 2.- Nasal intubation is performed to allow a clear view of the face and facial features during surgery; we might bear in mind the videolaryngoscopy to facilitate vocal cordal visualization in patients with difficulties in direct laryngoscopy. 3.-Critical aspects such as prevention of ocular injuries, the use of throat pack, deep venous prophylaxis, fluid therapy, edema management and infection prevention are addressed and are crucial for optimal recovery 4.-Multimodal analgesia is considered essential, combining local infiltration with opioid and non-opioids analgesia. Hialotherapy contributes to reducing postoperative pain and edema Collaboration between anesthesiologists, surgeons, and other healthcare professionals is essential to optimize the quality of care in orthognathic surgery. Protocols and anticipation of potential issues are key aspects in achieving excellent outcomes. In summary, comprehensive and careful anesthetic management, along with a multidisciplinary approach and the implementation of preventive protocols, is essential to improve outcomes and expedite recovery in orthognathic surgery. |
Pyoluteorin-deficient | 620b51f0-06eb-44a6-ab48-23b588f2c661 | 11634903 | Microbiology[mh] | The quality of plant rhizosphere microbiomes is critical in plant-disease suppression. Cooperative interactions between the beneficial microbes can efficiently suppress the invasion of plant pathogens, enhance efficient colonization, and positively affect competition for niches in the plant rhizosphere – . A diverse and positively interacting microbial community structure is a precondition in promoting plant growth, and the effective control of plant diseases. Abundant evidence suggests that disease control can be more efficient by the collective actions of the microbial community rather than the individual contributions of specific bacterial or fungal species . In order to elevate the disease-controlling effects of the natural microbiome, synthetic consortia, consisting of two or more microbial strains, able to control plant pathogens (multi-strain biocontrol agents, MSBCAs), have been employed to improve the growth and harvest yield of crops . In the reductionist synthetic community approach, only a few well-characterized members of the natural microbiome are assembled to form a defined synthetic community (SynCom) . However, in using this approach, the interactions between different bacteria, which can directly determine the type of microbial community structure, need to be carefully considered , . Fluorescent pseudomonads and plant-associated bacilli occupy a significant position in the microbial community structure of plant rhizosphere , and represent excellent models for beneficial bacteria with biocontrol function – , and might be promising candidates for applying together within a SynCom. Fluorescent pseudomonads, the most abundant bacteria in the plant rhizosphere, have a dominant influence on plant growth and development and play a vital role in plant-disease control , . Pseudomonas putida strain IsoF effectively eliminates both, plant pathogenic bacteria, and bacterial competitors, by injecting toxic effectors into neighboring bacterial cells utilizing a type IVB secretion system (T4BSS) . Among the group of fluorescent pseudomonads, Pseudomonas protegens has obtained considerable attention in biocontrol research due to its extraordinary antimicrobial properties, which are mainly due to their secondary metabolites, such as pyoluteorin , orfamide A , 2,4-diacetylphloroglucinol (DAPG) , pyrrolnitrin , and pyoverdine . Also, volatile organic compounds (VOCs) enhance the competitiveness of P. protegens within plant rhizosphere . Extensive documentation supports the notion that a gene cluster involved in inositol degradation bestows Pseudomonas with exceptional colonization abilities . Plant-associated bacilli are widely applied in the biocontrol of plant diseases. Bacillus strains are advantageous, due to their ability to form resistant endospores, and are well-suitable for large-scale fermentation . The plant-associated B. velezensis , previously B. amyloliquefaciens subsp. plantarum , a member of the B. subtilis species complex, is widely applied as a powerful BCA, and known to produce a diverse array of antagonistic metabolites . A representative of this taxonomic group, B. velezensis DMW1, isolated from potato inner tissues, is reported to synthesize non-ribosomal the lipopeptides iturin, surfactin, fengycin, the polyketides difficidin, bacillaene, macrolactin, and the dipeptide bacilysin. DMW1 possesses a remarkable capacity to effectively suppress the growth of pathogenic bacteria, fungi, and oomycetes , . We assumed that combining selected representatives of the two powerful groups of gram-negative and gram-positive BCAs, P. protegens Pf-5 and B . velezensis DMW1, within a two-strain SynCom might result in an efficient BCA surpassing the efficiency of the corresponding single strain BCAs. However, no mutualistic effect was registered. By contrast, in this combination, P. protegens Pf-5 did strongly suppress B. velezensis DMW1 leading to annulment of its biocontrol action. The objective of this study was to elucidate the molecular interactions between both strains and to identify the reason for suppressing DMW1 by Pf-5 with the goal of constructing more efficient consortia from both bacteria. The coexistence in their natural environments between plant-associated bacilli and pseudomonads was recently reviewed by Lyng and Kovács . However, in many cases the interactions observed in dual cultures were negative. Amensalism (suppressive effects) against Bacillus was often observed . Secondary metabolites such as lipopeptides produced by several pseudomonads can mediate inhibition of B. velezensis during colonizing tomato roots . Moreover, the pseudomonad type VI secretion system (T6SS) and antibiotic 2,4-diacetylphloroglucinol can impact biofilm formation in B. subtilis . Recently, it was reported that antagonism between bacilli and pseudomonads is shaped by competition for iron – . In order to design more compatible consortia, we have identified in this study the key secondary metabolite responsible for targeted inhibition of B. velezensis , and were able to design a two-strain-SynCom, more efficient than the single strains. By contrast to the wild type, the Pf-5 mutant lacking the hybrid polyketide pyoluteorin can smoothly coexist with B. velezensis . The two-strain-SynCom enhanced biofilm formation, metabolite production, and tomato root colonization. This cooperative interaction improved the efficacy of tomato bacterial wilt disease control and reshaped the microbial community structure in the tomato rhizosphere. Our study demonstrated that by targeted switching off of a functional metabolite produced by a member of the synthetic consortia, their efficiency in disease control can be markedly enhanced. P. protegens Pf-5 overcomes Bacillus strains including B. velezensis DMW1 P. protegens Pf-5, the model strain of biocontrol Pseudomonas , was selected as one member of a two-member consortium consisting of representatives of gram-negative and gram-positive biocontrol bacteria. Several Bacillus strains, representing different species of the B. subtilis species complex, were chosen as the second member of the synthetic consortium. It was observed that P. protegens Pf-5 suppressed the growth of B. velezensis DMW1, the model strain B. velezensis FZB42, B. subtilis SYST2, B. paralicheniformis NMSW12, B. safensis GBSW22, B. pumilus NMSW10, and B. halotolerans DGL6 (Fig. ). This indicated that P. protegens Pf-5 contained an active factor or metabolite, which could suppress the growth of all the Bacillus strains used in this study. To explore the active factor(s), a bipartite consortium, consisting of P. protegens Pf-5, and B. velezensis DMW1, chosen due to its excellent biocontrol properties , was established. Pf-5 and DMW1 were differently labeled with either red fluorescence or green fluorescence proteins. The competitive test on solid LB agar showed that Pf-5 entered gradually into the growth region of DMW1, thus inhibiting the growth of DMW1 during the time (Fig. ). The cell number of DMW1 in the contact region with Pf-5 (near Pf-5, Pf-5(N)) was found significantly reduced compared to the DMW1 region not contacted by Pf-5 (far Pf-5, Pf-5(F)). Only less than 3‰ of the cell number, detected in the unaffected control region, were registered after 48 h growth (Fig. b, c). By contrast, the cell number of Pf-5 in the DMW1 contact region (near DMW1, DMW1(N)) was not different compared to the Pf-5 region without contact with DMW1 (far DMW1, DMW1(F)) (Fig. ). B. velezensis DMW1 and its different metabolites (crude extract, surfactin, iturin, and fengycin) did not much impair the growth of P. protegens Pf-5 (Fig. e, f) suggesting that P. protegens Pf-5 strongly suppressed DMW1, but is less inhibited by B. velezensis and its metabolites. The factor responsible for the superiority of P. protegens Pf-5 over B. velezensis DMW1 is regulated by the GacS/GacA two-component system To find out which factor is responsible for Pf-5’s superiority over DMW1 within the two-member consortium, the GacS/GacA two-component system, which globally regulates the synthesis of many secondary metabolites, was considered. The mutant Δ gacA , impaired in expression of the GacS/GacA two-component system, was constructed, and labeled by red fluorescence. The competition experiments showed that Δ gacA lost the ability to overcome DMW1 and displayed no antagonistic activity against DMW1 in the criss-cross experiment (Fig. ). The cell number of DMW1 in the region, adjacent to the Δ gacA mutant growth region (near Δ gacA , Δ gacA (N)), was not significantly affected (Fig. ). As expected, the cell number of the Δ gacA mutant growing in the regions contacted (near DMW1, DMW1(N)) or not contacted (far DMW1, DMW1(F)) by DMW1 was not significantly different (Fig. ). After complementing the mutant gacA strain with the gacA + wild-type gene, resulting in strain Δ gacA-C , the competitive superiority of Δ gacA-C was completely restored (Fig. d, e). DMW1 had no apparent effect on the cell number of Δ gacA-C (Fig. ). Agar-diffusion tests performed with the culture, and the crude extract containing secondary metabolites produced by the Pf-5 wild-type, the Δ gacA mutant, and the complement Δ gacA-C strain corroborated that the GacA/GacS system governs a metabolic factor responsible for the antagonistic activity of Pf-5 exerted against DMW1 (Fig. ). The gacA mutant lost the ability of the wild strain to suppress the growth of DMW1 completely, but this ability was restored by complementing the mutant strain with the gacA + wild-type gene. The secondary metabolite pyoluteorin is responsible for the growth superiority of P. protegens Pf-5 in the two-member community HPLC analysis of the crude extracts obtained from Pf-5 wild-type, Δ gacA , and Δ gacA-C , corroborated earlier findings that synthesis of secondary metabolites, such as pyoluterin, orfamide A, and 2,4-diacetylphloroglucinol (DAPG) is controlled by the GacS/GacA system . The gacA deficient mutant strain, impaired in the synthesis of the GacA/GacS system, produced none of these metabolites (Fig. and Supplement Fig. ). In order to identify the compound(s) responsible for antagonistic activity of Pf-5 against DMW1, the genes responsible for the synthesis of antimicrobial metabolites were knocked out by triparental mating: ofa A (orfamide A), plt B (pyoluteorin), phl A (2,4-diacetylphloroglucinol), prn A (pyrrolnitrin), rzx B (rhizoxin), hcn ABC (hydrogen cyanide), and pFL4656 (non-ribosomal peptide). Only the Δ pltB mutant, unable to synthesize pyoluteorin, lost its ability to antagonize DMW1 (Fig. ). In vitro experiments performed with purified pyoluteorin confirmed that pyoluterin efficiently inhibits the growth of DMW1 (Fig. ). Criss-cross experiments demonstrated that, similar as in the Δ gacA mutant, Δ pltB lost the growth advantage against DMW1 (Fig. d, e). Similar to the Δ gacA mutant strain, the cell number in the Δ pltB mutant was not affected by the presence of DMW1 (Fig. ). Taken together, these results demonstrated that the polyketide pyoluteorin is the critical factor, responsible for the growth advantage of P. protegens Pf-5 over B. velezensis DMW1. In the pyoluteorin-deficient P. protegens Pf-5 mutant, thirty-one percent of the differentially expressed genes were significantly upregulated, including those associated with cell motility and metabolite production To further explore the changes of DMW1 after treatment with Pf-5 or Δ pltB , the transcriptome was analyzed. Gene Ontology (GO) term enrichment among differentially expressed genes (DEGs) was performed to comprehensively reveal the biological functions encoded by these genes. This GO enrichment analysis not only highlighted the active roles of DEGs across various biological dimensions but also provided deeper insights into how these genes participate in and influence key functional mechanisms within the bacterium (Fig. , Table ). The results of the KEGG pathway enrichment analysis revealed significant enrichments in multiple key metabolic and biosynthetic pathways, including Carbon metabolism, Biosynthesis of amino acids, Quorum sensing, Carbon fixation pathways in prokaryotes, Biosynthesis of secondary metabolites, Non-ribosomal synthesized-peptides, Propanoate metabolism, RNA degradation, and Valine, leucine, and isoleucine biosynthesis (Fig. ). Based on the KEGG pathway analysis, we focused on the differentially expressed genes (DEGs) enriched under the two major categories of Cellular Processes and Metabolism. In-depth analysis revealed that compared to Pf-5 treatment, Δ pltB treatment significantly enriched DEGs closely related to processes such as flagellin synthesis, amino acid metabolism, and non-ribosomal peptide synthesis. These findings not only highlighted the unique impact of Δ pltB treatment on cell motility, fundamental metabolic activities, and non-ribosomal synthesis of antimicrobial lipopeptides and polyketides but also provided important clues for further exploring its potential biological effects and mechanisms (Table ). We analyzed the expression patterns of DEGs related to flagellin synthesis, and non-ribosomal-peptide synthetases after co-culturing Bacillus with the wild-type Pf-5 and the mutant Δ pltB , respectively (Table ). 21 out of 22 genes involved in flagellin synthesis were upregulated after co-culturing with the mutant Δ pltB . Furthermore, in screening for genes related to non-ribosomal peptide synthetase, 7 out of 11 genes were found to be upregulated (Table ). The data described above demonstrated that B. velezensis DMW1 growing together with P. protegens Pf-5 was suppressed by the pyoluteorin production of Pf-5. In order to evaluate whether the synthesis of antimicrobial secondary metabolites produced by B. velezensis is differently affected by Pf-5, and the Δ pltB mutant strain, unable to secrete pyoluteorin, we analyzed the expression of their transcripts by RT-qPCR. The results showed that the relative expression level of the genes, responsible for the non-ribosomal peptide synthetase of iturin, fengycin, surfactin, difficidin, macrolactin, and bacillaene, and the genes involved in flagellum synthesis in B. velezensis DMW1, were enhanced by 4.24–13.10 times when co-cultivated with the Δ pltB mutant strain compared to the co-cultivation with the Pf-5 wild-type strain (Fig. ). The pyoluteorin-deficient P. protegens Pf-5 mutant promotes the production of secondary metabolites in DMW1 HPLC analysis corroborated that synthesis of the main secondary metabolites of DMW1 was significantly increased when co-cultivated with Δ pltB (Fig. ). Compared with the DMW1 treatment alone, the relative abundance of fengycin, difficidin, and bacillaene in Pf-5 treatment decreased by 10.6%, 10.1%, and 37.3% respectively. Compared with Pf-5 or DMW1 treatment, the relative abundance of the metabolites in Δ pltB treatment significantly increased. Among the main metabolites, the lipopeptides iturin, fengycin, and surfactin involved in plant-induced systemic resistance and antifungal action increased by 1.37, 1.32, and 2.02 times compared to DMW1, while the antibacterial polyketides macrolactin, difficidin, and bacillaene increased by 1.84, 1.21, and 1.39 times compared to DMW1 alone (Fig. ). It has been shown previously that these secondary metabolites are important for the biocontrol exerted by DMW1 against fungal and bacterial plant pathogens . In conclusion, by using a Δ pltB mutant strain, unable to synthesize pyoluteorin, together with the efficient DMW1 biocontrol strain in a bipartite synthetic consortium, it is possible to enhance the efficiency of the consortium without negatively affecting the growth of the gram-positive partner strain. Removal of pyoluteorin synthesis enables the P. protegens Pf-5/ B. velezensis DMW1 consortium to coexist harmoniously, and to be efficient in biofilm formation, root colonization, and plant-disease control Efficient biofilm formation is a precondition for root colonization by plant-associated bacteria. DMW1 formed dense wrinkled biofilms when cultivated solely. However, in the presence of Pf-5, this phenotype disappeared, and the hybrid biofilm appeared thin and sparse, and was characterized by an overwhelming number of Pf-5 cells. By contrast, the replacement of the Pf-5 wild-type by the mutant Δ pltB , restored the dense, wrinkled phenotype, which resembled the biofilm phenotype produced in the absence of Pf-5. Fluorescence microscopy revealed that the number of red-labeled Pseudomonas and green-labeled Bacillus cells was nearly equal (Fig. ). The expression of the biofilm-related genes epsB , and matrix components-related gene blsA , and the flagellar protein-related gene fliC in DMW1 was, compared to the Pf-5 wild type, significantly upregulated in coculture with the Δ pltB mutant (Fig. ). This means that pyoluteorin produced by Pf-5 plays the key role in inhibiting biofilm formation and suppressing of DMW1. Adding purified pyoluteorin (10 µg/mL) had the same inhibiting effect on biofilm formation as adding the Pf-5 wild-type strain (Supplementary Fig. ). If applied solely, the B. velezensis DMW1 efficiently colonized tomato roots. Thereby, the meristem, and the elongation zone of the main root, especially the junction between the epidermal cells, and the root hairs, were preferentially colonized. But, when the P. protegens Pf-5 strain was co-applied with DMW1, both bacteria colonized different sites of the tomato roots. By contrast, when the Pf-5 wild-type was replaced by the Δ pltB mutant strain, B. velezensis and P. protegens (Δ pltB ) did nicely coexist within the same sites (Fig. and Supplementary Fig. ). This result is in accordance with our previous finding that DMW1 and Δ pltB can form mixed biofilms consisting of the members of both species (Fig. ). The degree of colonization was quantified by qPCR followed by absolute quantification of the number of gene copies , , which correspond to the colony forming units (CFU) colonizing the root tissues according to the standard curves (Supplementary Fig. ). The degree of colonization by DMW1 was significantly declined when co-applied with Pf-5 but was not significantly affected when co-applied with Δ pltB (Fig. ). As expected DMW1 had no effect on the colonization rate of Pf-5 (Fig. ). Our results demonstrated that the inhibitory effect of Pf-5 on the root colonizing efficiency of DMW1 can be overcome when expression of pyoluteorin was turned off. There was no statistical difference between the single DMW1 treatment and the treatment of supplementary DMW1 with Δ pltB . Both, P. protegens and B. velezensis could control when applied solely, tomato bacterial wilt disease caused by R. solanacearum , . Here, we found that DMW1, Pf-5, and its mutant Δ pltB efficiently controlled the tomato bacterial wilt disease by 63.46%, 73.08%, and 72.12% (Fig. f, g). Notably, when Pf-5 or Δ pltB were co-applied with DMW1, the control effect was significantly enhanced. The best result was obtained with the bipartite consortium consisting of DMW1 and the Δ pltB mutant strain yielding 91.35% and there was no significant difference from those consisting of DMW1 and Pf-5. These results indicated that in the absence of pyoluteorin production, the bipartite consortium consisting of the P. protegens Pf-5 mutant strain and B. velezensis DMW1 controls tomato bacterial wilt disease in a very efficient manner. P. protegens Pf-5 lacking pyoluteorin enhances the microbial diversity and abundance of tomato rhizosphere As demonstrated by the chord diagram, representatives of the gamma-proteobacteria, flavobacteria, alpha-proteobacteria, and beta-proteobacteria were the predominant components of the tomato rhizosphere. The addition of P. protegens Pf-5 increased the diversity of the rhizosphere microbial community. In the presence of Pf-5 or Δ pltB , an increase of the members of the Bacteroidetes phylum, and of the Proteobacteria compared to the water control treatment was registered (Supplementary Fig. ). Network metrics values, such as the number of edges, the clustering coefficient, the average degree, and network density, indicated higher diversity in the microbiome after Δ pltB treatment than after the Pf-5 treatment. Moreover, the network obtained after Δ pltB treatment also featured an increase in the occurrence of Bacillus spp. compared to the network obtained after Pf-5 treatment (Δ pltB /Pf-5 ratio of 0.4%/0.2%) (Fig. ). We observed differences in the composition of the microbiome in Pseudomonas -treated soil compared to natural soil (Fig. and Supplementary Fig. ). As expected, the microbiota developed after the Δ pltB inoculation differed from that after adding of Pf-5 most likely due to the absence of pyoluteorin (Fig. and Supplementary Fig. ). The unconstrained principal coordinate analysis (PCoA) of the Bray–Curtis distance revealed different clustering of the uninoculated rhizosphere soil microbiota (control), the microbiota after Pf-5 inoculation, and the microbiota after Δ pltB inoculation. These clusters were found clearly separated along the second coordinate axis, indicating significant variation (Fig. ). Measurement of the α-diversity showed notable differences between the various treatments. The introduction of the Pseudomonas strains did significantly enhance the microbial diversity of tomato rhizosphere soil (Fig. and Supplementary Fig. c, d). By contrast to Pf-5, adding Δ pltB significantly enriched the representatives of the Bacillus taxon (Fig. d, e and Supplementary Fig. ). Remarkably, the abundance of Bacillus spp. forced by Δ pltB treatment was notably higher than that observed in the control without inoculation of additional strains and in the Pf-5 treatment (Fig. f, g). There was a significant positive correlation in the cell numbers between Bacillus and Pseudomonas in the Δ pltB treatment (Fig. ). In summary, the addition of P. protegens to the tomato rhizosphere microbiome contributed to microbial diversity, which was found further enhanced in the absence of pyoluteorin. Furthermore, in the absence of pyoluteorin synthesis enrichment of beneficial bacteria such as Bacillus spp. residing within the rhizosphere was observed. Pf-5 overcomes Bacillus strains including B. velezensis DMW1 P. protegens Pf-5, the model strain of biocontrol Pseudomonas , was selected as one member of a two-member consortium consisting of representatives of gram-negative and gram-positive biocontrol bacteria. Several Bacillus strains, representing different species of the B. subtilis species complex, were chosen as the second member of the synthetic consortium. It was observed that P. protegens Pf-5 suppressed the growth of B. velezensis DMW1, the model strain B. velezensis FZB42, B. subtilis SYST2, B. paralicheniformis NMSW12, B. safensis GBSW22, B. pumilus NMSW10, and B. halotolerans DGL6 (Fig. ). This indicated that P. protegens Pf-5 contained an active factor or metabolite, which could suppress the growth of all the Bacillus strains used in this study. To explore the active factor(s), a bipartite consortium, consisting of P. protegens Pf-5, and B. velezensis DMW1, chosen due to its excellent biocontrol properties , was established. Pf-5 and DMW1 were differently labeled with either red fluorescence or green fluorescence proteins. The competitive test on solid LB agar showed that Pf-5 entered gradually into the growth region of DMW1, thus inhibiting the growth of DMW1 during the time (Fig. ). The cell number of DMW1 in the contact region with Pf-5 (near Pf-5, Pf-5(N)) was found significantly reduced compared to the DMW1 region not contacted by Pf-5 (far Pf-5, Pf-5(F)). Only less than 3‰ of the cell number, detected in the unaffected control region, were registered after 48 h growth (Fig. b, c). By contrast, the cell number of Pf-5 in the DMW1 contact region (near DMW1, DMW1(N)) was not different compared to the Pf-5 region without contact with DMW1 (far DMW1, DMW1(F)) (Fig. ). B. velezensis DMW1 and its different metabolites (crude extract, surfactin, iturin, and fengycin) did not much impair the growth of P. protegens Pf-5 (Fig. e, f) suggesting that P. protegens Pf-5 strongly suppressed DMW1, but is less inhibited by B. velezensis and its metabolites. P. protegens Pf-5 over B. velezensis DMW1 is regulated by the GacS/GacA two-component system To find out which factor is responsible for Pf-5’s superiority over DMW1 within the two-member consortium, the GacS/GacA two-component system, which globally regulates the synthesis of many secondary metabolites, was considered. The mutant Δ gacA , impaired in expression of the GacS/GacA two-component system, was constructed, and labeled by red fluorescence. The competition experiments showed that Δ gacA lost the ability to overcome DMW1 and displayed no antagonistic activity against DMW1 in the criss-cross experiment (Fig. ). The cell number of DMW1 in the region, adjacent to the Δ gacA mutant growth region (near Δ gacA , Δ gacA (N)), was not significantly affected (Fig. ). As expected, the cell number of the Δ gacA mutant growing in the regions contacted (near DMW1, DMW1(N)) or not contacted (far DMW1, DMW1(F)) by DMW1 was not significantly different (Fig. ). After complementing the mutant gacA strain with the gacA + wild-type gene, resulting in strain Δ gacA-C , the competitive superiority of Δ gacA-C was completely restored (Fig. d, e). DMW1 had no apparent effect on the cell number of Δ gacA-C (Fig. ). Agar-diffusion tests performed with the culture, and the crude extract containing secondary metabolites produced by the Pf-5 wild-type, the Δ gacA mutant, and the complement Δ gacA-C strain corroborated that the GacA/GacS system governs a metabolic factor responsible for the antagonistic activity of Pf-5 exerted against DMW1 (Fig. ). The gacA mutant lost the ability of the wild strain to suppress the growth of DMW1 completely, but this ability was restored by complementing the mutant strain with the gacA + wild-type gene. P. protegens Pf-5 in the two-member community HPLC analysis of the crude extracts obtained from Pf-5 wild-type, Δ gacA , and Δ gacA-C , corroborated earlier findings that synthesis of secondary metabolites, such as pyoluterin, orfamide A, and 2,4-diacetylphloroglucinol (DAPG) is controlled by the GacS/GacA system . The gacA deficient mutant strain, impaired in the synthesis of the GacA/GacS system, produced none of these metabolites (Fig. and Supplement Fig. ). In order to identify the compound(s) responsible for antagonistic activity of Pf-5 against DMW1, the genes responsible for the synthesis of antimicrobial metabolites were knocked out by triparental mating: ofa A (orfamide A), plt B (pyoluteorin), phl A (2,4-diacetylphloroglucinol), prn A (pyrrolnitrin), rzx B (rhizoxin), hcn ABC (hydrogen cyanide), and pFL4656 (non-ribosomal peptide). Only the Δ pltB mutant, unable to synthesize pyoluteorin, lost its ability to antagonize DMW1 (Fig. ). In vitro experiments performed with purified pyoluteorin confirmed that pyoluterin efficiently inhibits the growth of DMW1 (Fig. ). Criss-cross experiments demonstrated that, similar as in the Δ gacA mutant, Δ pltB lost the growth advantage against DMW1 (Fig. d, e). Similar to the Δ gacA mutant strain, the cell number in the Δ pltB mutant was not affected by the presence of DMW1 (Fig. ). Taken together, these results demonstrated that the polyketide pyoluteorin is the critical factor, responsible for the growth advantage of P. protegens Pf-5 over B. velezensis DMW1. P. protegens Pf-5 mutant, thirty-one percent of the differentially expressed genes were significantly upregulated, including those associated with cell motility and metabolite production To further explore the changes of DMW1 after treatment with Pf-5 or Δ pltB , the transcriptome was analyzed. Gene Ontology (GO) term enrichment among differentially expressed genes (DEGs) was performed to comprehensively reveal the biological functions encoded by these genes. This GO enrichment analysis not only highlighted the active roles of DEGs across various biological dimensions but also provided deeper insights into how these genes participate in and influence key functional mechanisms within the bacterium (Fig. , Table ). The results of the KEGG pathway enrichment analysis revealed significant enrichments in multiple key metabolic and biosynthetic pathways, including Carbon metabolism, Biosynthesis of amino acids, Quorum sensing, Carbon fixation pathways in prokaryotes, Biosynthesis of secondary metabolites, Non-ribosomal synthesized-peptides, Propanoate metabolism, RNA degradation, and Valine, leucine, and isoleucine biosynthesis (Fig. ). Based on the KEGG pathway analysis, we focused on the differentially expressed genes (DEGs) enriched under the two major categories of Cellular Processes and Metabolism. In-depth analysis revealed that compared to Pf-5 treatment, Δ pltB treatment significantly enriched DEGs closely related to processes such as flagellin synthesis, amino acid metabolism, and non-ribosomal peptide synthesis. These findings not only highlighted the unique impact of Δ pltB treatment on cell motility, fundamental metabolic activities, and non-ribosomal synthesis of antimicrobial lipopeptides and polyketides but also provided important clues for further exploring its potential biological effects and mechanisms (Table ). We analyzed the expression patterns of DEGs related to flagellin synthesis, and non-ribosomal-peptide synthetases after co-culturing Bacillus with the wild-type Pf-5 and the mutant Δ pltB , respectively (Table ). 21 out of 22 genes involved in flagellin synthesis were upregulated after co-culturing with the mutant Δ pltB . Furthermore, in screening for genes related to non-ribosomal peptide synthetase, 7 out of 11 genes were found to be upregulated (Table ). The data described above demonstrated that B. velezensis DMW1 growing together with P. protegens Pf-5 was suppressed by the pyoluteorin production of Pf-5. In order to evaluate whether the synthesis of antimicrobial secondary metabolites produced by B. velezensis is differently affected by Pf-5, and the Δ pltB mutant strain, unable to secrete pyoluteorin, we analyzed the expression of their transcripts by RT-qPCR. The results showed that the relative expression level of the genes, responsible for the non-ribosomal peptide synthetase of iturin, fengycin, surfactin, difficidin, macrolactin, and bacillaene, and the genes involved in flagellum synthesis in B. velezensis DMW1, were enhanced by 4.24–13.10 times when co-cultivated with the Δ pltB mutant strain compared to the co-cultivation with the Pf-5 wild-type strain (Fig. ). P. protegens Pf-5 mutant promotes the production of secondary metabolites in DMW1 HPLC analysis corroborated that synthesis of the main secondary metabolites of DMW1 was significantly increased when co-cultivated with Δ pltB (Fig. ). Compared with the DMW1 treatment alone, the relative abundance of fengycin, difficidin, and bacillaene in Pf-5 treatment decreased by 10.6%, 10.1%, and 37.3% respectively. Compared with Pf-5 or DMW1 treatment, the relative abundance of the metabolites in Δ pltB treatment significantly increased. Among the main metabolites, the lipopeptides iturin, fengycin, and surfactin involved in plant-induced systemic resistance and antifungal action increased by 1.37, 1.32, and 2.02 times compared to DMW1, while the antibacterial polyketides macrolactin, difficidin, and bacillaene increased by 1.84, 1.21, and 1.39 times compared to DMW1 alone (Fig. ). It has been shown previously that these secondary metabolites are important for the biocontrol exerted by DMW1 against fungal and bacterial plant pathogens . In conclusion, by using a Δ pltB mutant strain, unable to synthesize pyoluteorin, together with the efficient DMW1 biocontrol strain in a bipartite synthetic consortium, it is possible to enhance the efficiency of the consortium without negatively affecting the growth of the gram-positive partner strain. P. protegens Pf-5/ B. velezensis DMW1 consortium to coexist harmoniously, and to be efficient in biofilm formation, root colonization, and plant-disease control Efficient biofilm formation is a precondition for root colonization by plant-associated bacteria. DMW1 formed dense wrinkled biofilms when cultivated solely. However, in the presence of Pf-5, this phenotype disappeared, and the hybrid biofilm appeared thin and sparse, and was characterized by an overwhelming number of Pf-5 cells. By contrast, the replacement of the Pf-5 wild-type by the mutant Δ pltB , restored the dense, wrinkled phenotype, which resembled the biofilm phenotype produced in the absence of Pf-5. Fluorescence microscopy revealed that the number of red-labeled Pseudomonas and green-labeled Bacillus cells was nearly equal (Fig. ). The expression of the biofilm-related genes epsB , and matrix components-related gene blsA , and the flagellar protein-related gene fliC in DMW1 was, compared to the Pf-5 wild type, significantly upregulated in coculture with the Δ pltB mutant (Fig. ). This means that pyoluteorin produced by Pf-5 plays the key role in inhibiting biofilm formation and suppressing of DMW1. Adding purified pyoluteorin (10 µg/mL) had the same inhibiting effect on biofilm formation as adding the Pf-5 wild-type strain (Supplementary Fig. ). If applied solely, the B. velezensis DMW1 efficiently colonized tomato roots. Thereby, the meristem, and the elongation zone of the main root, especially the junction between the epidermal cells, and the root hairs, were preferentially colonized. But, when the P. protegens Pf-5 strain was co-applied with DMW1, both bacteria colonized different sites of the tomato roots. By contrast, when the Pf-5 wild-type was replaced by the Δ pltB mutant strain, B. velezensis and P. protegens (Δ pltB ) did nicely coexist within the same sites (Fig. and Supplementary Fig. ). This result is in accordance with our previous finding that DMW1 and Δ pltB can form mixed biofilms consisting of the members of both species (Fig. ). The degree of colonization was quantified by qPCR followed by absolute quantification of the number of gene copies , , which correspond to the colony forming units (CFU) colonizing the root tissues according to the standard curves (Supplementary Fig. ). The degree of colonization by DMW1 was significantly declined when co-applied with Pf-5 but was not significantly affected when co-applied with Δ pltB (Fig. ). As expected DMW1 had no effect on the colonization rate of Pf-5 (Fig. ). Our results demonstrated that the inhibitory effect of Pf-5 on the root colonizing efficiency of DMW1 can be overcome when expression of pyoluteorin was turned off. There was no statistical difference between the single DMW1 treatment and the treatment of supplementary DMW1 with Δ pltB . Both, P. protegens and B. velezensis could control when applied solely, tomato bacterial wilt disease caused by R. solanacearum , . Here, we found that DMW1, Pf-5, and its mutant Δ pltB efficiently controlled the tomato bacterial wilt disease by 63.46%, 73.08%, and 72.12% (Fig. f, g). Notably, when Pf-5 or Δ pltB were co-applied with DMW1, the control effect was significantly enhanced. The best result was obtained with the bipartite consortium consisting of DMW1 and the Δ pltB mutant strain yielding 91.35% and there was no significant difference from those consisting of DMW1 and Pf-5. These results indicated that in the absence of pyoluteorin production, the bipartite consortium consisting of the P. protegens Pf-5 mutant strain and B. velezensis DMW1 controls tomato bacterial wilt disease in a very efficient manner. Pf-5 lacking pyoluteorin enhances the microbial diversity and abundance of tomato rhizosphere As demonstrated by the chord diagram, representatives of the gamma-proteobacteria, flavobacteria, alpha-proteobacteria, and beta-proteobacteria were the predominant components of the tomato rhizosphere. The addition of P. protegens Pf-5 increased the diversity of the rhizosphere microbial community. In the presence of Pf-5 or Δ pltB , an increase of the members of the Bacteroidetes phylum, and of the Proteobacteria compared to the water control treatment was registered (Supplementary Fig. ). Network metrics values, such as the number of edges, the clustering coefficient, the average degree, and network density, indicated higher diversity in the microbiome after Δ pltB treatment than after the Pf-5 treatment. Moreover, the network obtained after Δ pltB treatment also featured an increase in the occurrence of Bacillus spp. compared to the network obtained after Pf-5 treatment (Δ pltB /Pf-5 ratio of 0.4%/0.2%) (Fig. ). We observed differences in the composition of the microbiome in Pseudomonas -treated soil compared to natural soil (Fig. and Supplementary Fig. ). As expected, the microbiota developed after the Δ pltB inoculation differed from that after adding of Pf-5 most likely due to the absence of pyoluteorin (Fig. and Supplementary Fig. ). The unconstrained principal coordinate analysis (PCoA) of the Bray–Curtis distance revealed different clustering of the uninoculated rhizosphere soil microbiota (control), the microbiota after Pf-5 inoculation, and the microbiota after Δ pltB inoculation. These clusters were found clearly separated along the second coordinate axis, indicating significant variation (Fig. ). Measurement of the α-diversity showed notable differences between the various treatments. The introduction of the Pseudomonas strains did significantly enhance the microbial diversity of tomato rhizosphere soil (Fig. and Supplementary Fig. c, d). By contrast to Pf-5, adding Δ pltB significantly enriched the representatives of the Bacillus taxon (Fig. d, e and Supplementary Fig. ). Remarkably, the abundance of Bacillus spp. forced by Δ pltB treatment was notably higher than that observed in the control without inoculation of additional strains and in the Pf-5 treatment (Fig. f, g). There was a significant positive correlation in the cell numbers between Bacillus and Pseudomonas in the Δ pltB treatment (Fig. ). In summary, the addition of P. protegens to the tomato rhizosphere microbiome contributed to microbial diversity, which was found further enhanced in the absence of pyoluteorin. Furthermore, in the absence of pyoluteorin synthesis enrichment of beneficial bacteria such as Bacillus spp. residing within the rhizosphere was observed. In this study, we explored the interaction within a synthetic consortium formed by a selected gram-negative Pseudomonas and a gram-positive Bacillus strain, both known for their excellent ability to biocontrol pathogens in the plant rhizosphere. The co-cultivation of P. protegens Pf-5 and B. velezensis DMW1 in a solid medium led to the growth inhibition of B. velezensis DMW1, which corresponds to earlier findings that fluorescent pseudomonads such as P. protegens exert negative interactions with plant-associated bacilli . Here, the unknown inhibiting factor was shown to be controlled by the GacS/GacA two-component system, known as a global regulator of secondary metabolism. Further analyses revealed that the chlorinated polyketide pyoluteorin, non-ribosomal synthesized by P. protegens Pf-5 , , and regulated by the GacS/GacA system, caused the antibiosis effect observed in DMW1. Pyoluteorin was shown to inhibit plant pathogens such as Pantoea ananatis , Oomycetes , Chladymonas reinhardtii , and Heterobasidion sp. , Burkholderia glumae , indicating a broad spectrum of antimicrobial activity. We found that P. protegens Pf-5 and B. velezensis DMW1 are incapable of forming together a robust biofilm. Due to the presence of pyoluteorin, the growth of DMW1 became suppressed within the consortium. P. protegens Pf-5 and B. velezensis DMW1 couldn’t coexist in the roots of tomato plants, and the ecological niche was mainly occupied by P. protegens Pf-5. Notably, B. velezensis DMW1 and P. protegens Pf-5 colonized two separate zones of the root system, given that both zones are sufficiently distant from each other. This observation might suggest that DMW1 can only colonize plant roots when not negatively affected by pyoluteorin. Pyoluteorin was identified as the key substance that hindered biofilm formation and affected the expression of genes related to biofilm formation, but its targets and inhibiting mechanisms in bacilli remain to be further studied. Favorable interactions among members can enhance the inhibitory activity of the SynCom against pathogenic bacteria, thereby boosting its biocontrol potential . In order to overcome the inhibiting effect exerted by Pf-5 against B. velezensis DMW1, the pyoluteorin knockout mutant Δ pltB of P. protegens Pf-5 was constructed. Co-cultivating of the P. protegens Δ pltB mutant together with DMW1 in a liquid medium resulted in increased synthesis of the main antimicrobial metabolites. The increased production of Bacillus secondary metabolites might contribute to enhanced biocontrol effects of the bipartite consortium. The efficacy of the two-member DMW1/ P. protegens Δ pltB consortium in controlling tomato bacterial wilt disease surpassed the efficacy of the treatment solely with Pf-5, suggesting that target-directed-engineering of selected members of synthetic consortia can overcome antagonistic interactions, and greatly enhance their efficacy. It is known that interbacterial interaction can trigger the activation of Bacillus biosynthetic gene clusters (BGCs) residing in the root rhizosphere . Another way to overcome inhibition of biofilm formation of plant-associated bacilli in the presence of fluorescent pseudomonads has been reported. Mutations in negative regulators of biofilm formation were generated during directed laboratory evolution resulting in improved competitiveness of B. subtilis . Through a comprehensive comparative analysis of the effects of P. protegens Pf-5 and P. protegens Δ pltB on the microbial community structure in the tomato rhizosphere, we made an intriguing discovery: both P. protegens Pf-5 and P. protegens Δ pltB have the remarkable ability to significantly augment the diversity of the microbial community structure in the soil surrounding tomato roots, thus contributing to a more robust and thriving ecosystem in the tomato rhizosphere. Applying the strategy to overcome the limitation set by the kin boundary by identifying and then eliminating the factor(s) hindering cooperative behavior within synthetic consortia , , we paved the way for developing highly efficient synthetic consortia, e.g. in biocontrol of phytopathogens. Interestingly, after introducing the Δ pltB mutation, P. protegens supported the colonization of Bacillus in the tomato rhizosphere and enhanced microbial biodiversity. This finding highlights the unique contribution of P. protegens Δ pltB in enriching Bacillus spp. (Fig. ). This provides further evidence that positive interaction relationships are particularly beneficial for promoting the diversity of microbial community structure. As an alternative to the use of an engineered SynCom described here, careful selection of beneficial species exhibiting mutualistic interactions offers another way to obtain efficient SynComs. For example, B. velezensis and P. stutzeri can achieve mutualistic symbiosis through the cross-feeding of metabolites . Cross-feeding between bacteria is common and important – . Researchers utilized GutCP to predict numerous new cross-feeding interactions in the human gut microbiota and revealed cross-feeding interactions for nearly 65% of the microorganisms . Leaf bacteria play a significant role in cross-feeding interactions that have functional relevance, and these interactions can be influenced by the phyllosphere environment, indirectly contributing to the diversity of bacterial populations . Mutual interactions among beneficial bacteria can lead to synergistic effects. Here, we showed that by introducing the Δ pltB mutation, P. protegens and B. velezensis developed a cooperative behavior, including the capacity to co-form biofilms and to co-colonize tomato roots. Furthermore, B. velezensis DMW1 enhanced colonization of tomato roots in the presence of P. protegens Δ pltB , showcasing its beneficial effects and biocontrol potential. The efficacy of the two-member DMW1/ P. protegens Δ pltB consortium in controlling tomato bacterial wilt disease, surpassed the effectiveness of the treatment solely with Pf-5, suggesting that target-directed-engineering of selected members of synthetic consortia can overcome antagonistic interactions, and greatly enhance their efficacy. Taken together, the introduction of the targeted mutation connected with the switch off of the pyoluteorin synthesis in the synthetic consortium of two taxonomically very distant plant-growth-promoting bacteria (PGPB), restores their beneficial action on plant growth due to improved interactions on community-level, and enhanced ability to colonize commonly plant roots. Our findings have implications for designing and applying synthetic communities consisting of remote-related PGPB. Microorganisms and cultivation The bacteria used in this study are listed in Table . Bacillus strains, and Escherichia coli strains were cultured by inoculating a single colony in 20 mL lysogeny broth (LB) (10 g L −1 NaCl, 5 g L −1 yeast extract, and 10 g L −1 tryptone) medium, and incubated overnight at 37 °C with shaking (200 rpm); P. protegens strains were cultured by inoculating a single colony in 20 mL King’ B (KB) (20.0 g L −1 tryptone, 10.0 g L −1 NaCl, 1.5 g L −1 MgSO 4 ·7H 2 O, 1.5 g L −1 K 2 HPO 4 ), and incubated overnight at 30 °C with shaking (200 rpm). Plasmid cloning was performed in E. coli DH5α (Table ). The cultured bacteria are regarded as cultures. Primers used for cloning and verification are found in Table . Tagging Pseudomonas and Bacillus with fluorescent proteins To visualize the development of bacterial strains co-cultivated on a solid agar surface, Pf-5 and its mutants were tagged with mCherry red fluorescent protein. The pBBR -mCherry plasmid was transformed into E. coli Top10, and the recombinant Pf-5/pBBR -mCherry , Δ pltB/ pBBR -mCherry , Δ gacA/ pBBR -mCherry strains were generated by triparental mating of Top10/pBBR- mCherry (donor), HB101/pRK - 2013 (helper), and the Pf-5, Δ pltB and Δ gacA strains of interest (recipient) , . In the same way, the mCherry fragment was amplified from the pBBR -mCherry plasmid to construct pUCP26 -mCherry . This plasmid was then transformed into E. coli Top10 and the recombinant Δ gacA-C /pUCP26- mCherry (Δ gacA-C-mcherry ) strain was generated according to the method described above. DMW1 was labeled with green fluorescence protein gfp by transfer of the pAD43-25 plasmid . Detection of competition ability The effect of P. protegens on bioactivity toward Bacillus spp. was determined in an inhibition assay on a solid medium. Cultures of Bacillus spp., as well as P. protegens Pf-5, were adjusted to an OD 600 nm = 1.0. Five μL of the P. protegens Pf-5 culture were placed onto 1.0% LB agar supplemented with 1.0% (v/v) of Bacillus sp. culture. The activity of P. protegens Pf-5 toward Bacillus spp. was determined as described above. The LB agar plates were incubated at 30 °C for 48 h, followed by measuring the size of inhibition zones. For bacterial co-cultivations, 1 mL cultures of P. protegens Pf-5 or its mutants labeled with mCherry , and B. velezensis DMW1 labeled with gfp were washed twice in sterilized water by centrifuging at 5000 rpm for 4 min, followed by discarding the supernatant and resuspending the pellet in sterilized water without the antibiotic. Subsequently, the optical density at 600 nm (OD 600 nm ) was adjusted to 1.0 ( Bacillus ) and 0.1 ( Pseudomonas ). For co-cultivations on agar surfaces, lines were drawn by streaking out P. protegens Pf-5 or its mutants labeled with mCherry , crossed by B. velezensis DMW1 labeled with gfp . The incubation time varied from 24 to 72 h, at 30 °C. Fluorescent colonies were imaged by fluorescent stereo microscopy (Nikon SMZ25, Nikon, Japan). Due to the excitation wavelength for red fluorescent protein being close to that of green protein, strong excitation (at 488 nm) can simultaneously excite both types of fluorescence. To determine the colony number of the bacteria, labeled with different fluorescent proteins, after contacting each other, the bacterial disks were taken at the midpoint of the segment formed by the intersection of two perpendicular lines after 48 h. Three beads (φ = 3 mm) were put in each tube, which was then filled with 1 mL of sterilized water. The bacterial disks in the sterilized water were subjected to grinding for 30 sec at 60 Hz using tissuelyser-64 (Shanghai Jingxin Technology). The grinding steps were repeated twice, and then the bacteria in the medium were released and diluted in sterilized water. According to the gradient dilution method, the bacterial suspensions were then spread on KB medium plates containing ampicillin (100 µg/mL) and gentamicin (25 µg/mL) for Pf-5, Δ gacA and Δ pltB labeled with mCherry , or ampicillin (100 µg/mL) and tetracycline (25 µg/mL) for Δ gacA -C labeled with mCherry . LB medium plates containing chloramphenicol (5 µg/mL) were used for B. velezensis DMW1 labeled with gfp . The plates were incubated at 30 °C ( Pseudomonas ), or 37 °C ( Bacillus ) for 12 h, and then the number of colonies was counted and statistically analyzed. Crude extract preparation To extract secondary metabolites of B. velezensis or P. protegens strains, bacteria were cultured by inoculating a single colony in 20 mL LB medium and incubated overnight at 37 °C under shaking (200 rpm). Then, the bacteria were further cultivated in 200 mL of LB medium for 24 h. Afterward, the pre-treated XAD16 resin was added and adjusted to 2.5%, and the mixture was incubated for another 24 h. Centrifuge at 8000 rpm for 15 min at room temperature, discard the supernatant, resuspend the mixed resin precipitate with 30–40 mL of methanol, and place it at 37 °C at 200 rpm for 4 h, then centrifuge at 8000 rpm at room temperature for 15 min. The supernatant was dissolved in methanol, and the crude extract was obtained. Purification of iturin, surfactin, and fengycin For the purification of iturin, surfactin, and fengycin, crude extract sample was dissolved in a methanol solution and subjected to purification through preparative high-performance liquid chromatography (HPLC) using a C18 column (5 µm, 100.0 mm × 21.2 mm; flow rate: 10 mL/min; gradient: 20–100% MeOH/H 2 O in 30 min, followed by 100% MeOH in 10 min) to generate sub-fractions. The distinctive peaks in the DMW1 sub-fraction underwent semipreparative HPLC (C18 column, 5 µm; 250.0 mm × 10.0 mm; flow rate: 3 mL/min; with 0.1% formic acid in 54% MeOH/H 2 O in 53 min, 65% MeOH/H 2 O in 53–70 min) to obtain a pure compound. NMR spectra were recorded in methanol-d4 on a Bruker AVANCE II 400 MHz instrument, and high-resolution mass spectra were acquired using an Agilent 6530 Accurate-Mass Q-TOF LC/MS coupled to an Agilent 1260 HPLC. Inhibition assay The effect of B. velezensis DMW1, and its secondary metabolites extracts on P. protegens Pf-5 was determined in an inhibition assay on a solid medium. Cultures of B. velezensis DMW1, as well as P. protegens Pf-5 were normalized to an optical density at 600 nm (OD 600 nm ) of 1.0. Five μL of the culture of DMW1 or its secondary metabolites extracts were applied to a paper disk (φ = 5 mm) and then placed onto LB agar containing 1.0% (v/v) of P. protegens Pf-5 culture. The activity of P. protegens Pf-5 or its secondary metabolites extracts towards B. velezensis DMW1 was determined using the same method. Plates were incubated at 30 °C for 48 h, followed by measuring the size of inhibition zones. Construction of deletion mutants in Pf-5 genes involved in the synthesis and regulation of antimicrobial metabolites To explore the effect of secondary metabolites of P. protegens Pf-5 on the activity of B. velezensis DMW1, several Pseudomonas deletion mutants were generated by triparental mating according to the description above , . Mutants of P. protegens Pf-5 were created by the triparental mating. Briefly, the upstream (770 bp) and downstream (770 bp) fragments were amplified by PCR, fused via overlap PCR, and introduced into the pK18mobsacB vector. The recombinant plasmid was transferred from E. coli Top10 to P. protegens Pf-5, with the help of E. coli Top10 (pRK-2013) , . The following mutations were introduced: gacA (secondary metabolite regulatory system) , ofa (orfamide A biosynthetic gene) , pltB (pyoluteorin biosynthetic gene) , phlA (2,4-diacetylphloroglucinol biosynthetic gene) , prnA (pyrrolnitrin biosynthetic gene) , rzxB (rhizoxin biosynthetic gene) , hcnABC (VOC hydrogen cyanide biosynthetic gene) , pFL4656 (non-ribosomal peptide biosynthetic gene). Purification of pyoluteorin For the purification of pyoluteorin, high-performance liquid chromatography (HPLC, Waters), equipped with a reverse-phase column (ZORBAX SB-C18) was performed. The running program was a gradient elution from 5% solvent A (HPLC-grade acetonitrile containing 0.1% trifluoroacetic acid), 95% solvent B (Milli-Q water containing 0.1% trifluoroacetic acid) to 95% A, 5% B for 20 min. A concentration of 95% solvent A and 5% solvent B were then held for 5 min. A gradient elution from 95% solvent A, 5% solvent B to 5% A, 95% B for 1 min; A concentration of 95% solvent A and 5% solvent B were then held for 2 min. For the purification of pyoluteorin, the semipreparative HPLC was used with a flow rate of 4 mL/min and an injection volume of 1 mL. Pyoluteorin was done under UV absorption at 310 nm. In the last step, purified compounds were identified via ultraperformance liquid chromatography-mass spectrometry (UPLC-MS). RNA-seq B. velezensis DMW1 was grown as a single culture or co-cultured with P. protegens Pf-5 or Δ pltB , respectively. Two mL of the culture, containing either the single strain or a mix of B. velezensis DMW1 (OD 600 nm = 1.0) and P. protegens (OD 600 nm = 0.1) (volume ratio = 19:1), were used for inoculation of 200 mL medium as described previously . The cultures were grown at 37 °C under shaking (200 rpm) for 12 h. The bacteria were collected for total RNA extraction and RNA-seq by the DNBSEQ platform (BGI Genomics Co., Ltd., China). The reads were mapped to the DMW1 genome v1.1 ( https://www.ncbi.nlm.nih.gov/nuccore/NZ_CP114180.1 ) using HISAT2. Utilizing the SOAPnuke tool, the expression levels of individual DMW1 genes were quantified through normalization to the “Fragments Per Kilobase of exon per Million mapped reads (FPKM)” metric, enabling the identification of differentially expressed genes (DEGs). The genes with at least two-fold change and p- value ≤ 0.05 were considered DGEs between the two samples. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs was performed using the “KEGG enrichment analysis” tool embedded in the BGI platform ( https://report.bgi.com/ps/mrna/index.html ). RNA isolation and RT-qPCR To investigate srf , itu, fen , mln , bae , dfn , flgD , and fliR gene expression in Bacillus , firstly RNA extraction was carried out using the Bacteria Total RNA Kit (ZP403) (Zoman Biotechnology Co., Ltd; Beijing) following the Gram-positive manufacturer’s protocol. RNA quality and quantity were performed with a Thermo Scientific NanoDrop 2000 UV-vis Spectrophotometer. Primer 3 program available online was used for primer design and primers were synthesized by GenScript. The primers used for this purpose are listed in Table . Reverse transcription-polymerase chain reaction (RT-qPCR) was performed to quantify gene expression of itu (iturin synthesis), fen (fengycin synthesis), srf (surfactin synthesis), bae (bacillaene synthesis), mln (marcolactin synthesis), dfn (difficidin synthesis), flgD (flagellar hook assembly protein FlgD) and fliR (flagellar type III secretion system protein FliR) . Reverse transcriptase and RT-qPCR reactions were conducted using the HiScript III 1st strand cDNA Synthesis Kit (+gDNA) (Vazyme, China). The qPCR reaction was performed in a total volume of 20 μL: 10 μL of ChamQ Universal SYBR qPCR Master Mix, 0.4 μL of each primer (5 μM), 2 μL of cDNA (100 ng), 7.2 µL of Nuclease-free water. The thermal cycling program applied on the ABI StepOne was: preheating at 95 °C for 30 s, (95 °C for 5 s and 60 °C for 30 s) × 40 cycles, and the specificity of the PCR product amplification was verified based on the Tm value for 15 s at 95 °C and 1 min at 60 °C. Finally, the qPCR amplification was run on the ABI StepOne qPCR instrument (Applied Biosystems) with software version 2.3. The relative gene expression analysis was conducted using the 2 –ΔΔCT method with the rpsU gene as a housekeeping gene to normalize mRNA levels between different samples . Analysis of secondary metabolites with HPLC Using a C18 reversed-phase column (ZORBAX SB-C18), we performed a gradient elution starting from 30% solvent A (HPLC-grade acetonitrile with 0.1% trifluoroacetic acid) to 70% solvent B (Milli-Q water with 0.1% trifluoroacetic acid) over a period of 30 min. This was followed by a gradient change to 95% solution A and 5% solution B, maintained for 10 min. The flow rate was set at 1 mL/min. Secondary metabolites were detected at UV wavelengths of 210 nm and 280 nm, respectively, and the peak area represented the production. Biofilm in analysis Bacillus – Pseudomonas interactions For the biofilm assay, 200 μL overnight culture was transferred into a well containing 20 mL liquid LB medium. After incubation at 30 °C/37 °C, 200 rpm for OD 600 nm = 1.0, 10 μL B. velezensis DMW1 (OD 600 nm = 1.0) and 10 μL P. protegens Pf-5/Δ pltB (OD 600 nm = 0.25) culture or sterilized water was added into 2 mL fresh LBGM medium (LB supplemented with 1% glycerol and 100 μM MnSO 4 ) in a flat bottom 12-well microplate . To visualize the distribution of two bacteria in the cultures, the culture was examined with a Zeiss LSM 700 microscope, equipped with a 20× objective. GFP fluorescence was excited at 488 nm and detected between 500 and 530 nm; for mCherry, excitation was at 561 nm with emission detection from 580 to 620 nm after incubation at 30 °C for 24 h without shaking. Biofilm-related gene expression level among the different treatments was analyzed at the transcriptional level by RT-qPCR. Construction of standard curve for qPCR After ligating the target fragment into the pMD-18T vector, the resulting construct was transformed into DH5α competent cells. Plasmids from the transformants were extracted and then diluted in a 10-fold gradient. These diluted plasmids were subsequently used as templates for fluorescence quantitative PCR amplification. The logarithm of the initial copy number of the template was plotted on the x-axis, while the number of cycles reaching the threshold (Threshold cycle, Ct) was plotted on the y-axis, thus generating a standard curve. The CT number was introduced into the standard curve equation to calculate the initial gene copy number of the sample to reveal the gene copy number per gram of dry soil . Root colonization assay To determine the colony number of tomato root colonizing B. velezensis DMW1 and P. protegens Pf-5/Δ pltB in the rhizosphere soil, thirty-day-old tomato-plant seedling roots were dipped into the bacterial suspension containing 5 × 10 7 cells/mL Bacillus and 3 × 10 8 cells/mL Pseudomonas for 40 min. Afterward, all seedlings were transplanted into pots (10 cm in diameter) containing 200 g sterilized peat soil from Danish Pindstrup company (located in the Kingdom of Denmark), and incubated in the greenhouse (30 °C, 75% relative humidity, 14 h light/10 h darkness). Rhizosphere soil samples from these treatments were collected 7 d after transplanting. Five tomato roots were collected from the points of each plot and vigorously shaken to remove excess soil. Then, the soil adhering to the roots (rhizosphere soil) was suspended in phosphate-buffered saline (PBS). The rhizosphere soil suspensions were centrifuged at 10,000 × g for 10 min, and the sediments were stored at −80 °C for later DNA extraction. Soil samples from the pot experiments were chosen for subsequent DNA extraction using Power Soil DNA Isolation Kits (MoBio Laboratories Inc., Carlsbad, CA, USA) following the manufacturer’s protocol. The concentration and quality of the DNA were determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, NC, USA). The abundance of Pseudomonas spp. was quantified by quantitative polymerase chain reaction (qPCR) with specific primers . The abundance of Bacillus spp. was quantified by qPCR with specific primers . For imaging of B. velezensis DMW1/ pAD43-25-gfp , and P. protegens Pf-5 or Δ pltB / pBBR-mCherry colonization in the tomato root, surface-sterilized tomato seeds with roots 10 mm long (3 d) were inoculated with suspensions of B. velezensis DMW1 (5 × 10 7 cells), and P. protegens Pf-5/Δ pltB (3 × 10 8 cells). After incubation for 30 min at 25 °C and shaking with 100 rpm, seedlings were transferred onto square Petri dishes (12 × 12 cm) containing 0.5× Murashige and Skoog (MS) semisolid agar medium (0.7% agar) without sucrose . The square plates were kept in a vertical position during the incubation time of seven days at 25 °C under long daylight conditions (16 h light/8 h darkness) in a plant-growth chamber. To determine the spatial pattern of root colonization by DMW1/ pAD43-25-gfp and Pf-5/ pBBR-mCherry or Δ pltB/pBBR-mCherry , the root system (ca. 10 mm long) was imaged using a Zeiss LSM 700. GFP fluorescence was excited at 488 nm and detected between 500 and 530 nm; for mCherry, excitation was at 561 nm with emission detection from 580 to 620 nm. To determine the root colonization by B. velezensis DMW1 and P. protegens Pf-5/Δ pltB , the root system (ca. 10 mm long) was subjected to grind for 30 sec at 60 Hertz in a 2 mL tube containing 1 mL of distilled water and three beads (φ = 3 mm) using tissuelyser-64 (Shanghai Jingxin Technology). Using the gradient dilution method, the diluted bacterial suspensions were placed onto LB plates. The plates were incubated at 30 °C/37 °C for 12 h to count the number of colonies. Disease control Thirty-day-old tomato seedlings (variety cultivar ‘Mao fen’) roots were inoculated with either: 10 mL 1 × 10 7 CFU/mL DMW1, or 10 mL 1 × 10 8 CFU/mL Pf-5, or 10 mL 1 × 10 8 CFU/mL Δ pltB culture, or a mix of DMW1 and Pf-5 or Δ pltB . Then, 10 mL 5 × 10 7 CFU/mL R. solanacearum suspension was added. Each treatment was repeated in triplicate, and every repeat contained 15 seedlings. All seedlings were maintained in the greenhouse (30 °C, 75% relative humidity, 14 h light/10 h darkness), and inspected periodically until disease symptoms appeared. The disease severity was evaluated according to an empirical scale: Level 0 = tomato plants without visible symptoms; Level 1 = striped necrosis on stems occasionally or less than half of the leaves wilted on unilateral stems; Level 2 = black streaks less than half the height of the stem or between half to two-thirds of the leaves wilted on unilateral stems; Level 3 = more than two-thirds of the leaves wilted on unilateral stems; Level 4 = the plant is dead . The disease severity index was calculated according to the following formula: 1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm{Disease}}\; {\rm{severity}}\; {\rm{index}}\,({\rm{DSI}})=[\Sigma ({\rm{x}}\times {\rm{y}})/({\rm{z}}\times 4)]\times 100$$\end{document} Disease severity index ( DSI ) = [ Σ ( x × y ) / ( z × 4 ) ] × 100 2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm{Disease}}\; {\rm{control}}\; {\rm{effect}}\,=({{\rm{DSI}}}_{{\rm{Treatment}}}-{{\rm{DSI}}}_{{\rm{Control}}})/{{\rm{DSI}}}_{{\rm{Control}}}\times 100 \%$$\end{document} Disease control effect = ( DSI Treatment − DSI Control ) / DSI Control × 100 % Where: x = number of different degrees infected plants in the treatments; y = relative degree value; and z = number of total plants in the treatments. Inoculation of tomatoes and sample collection Tomato seedlings (variety cultivar ‘Mao fen’) grown in natural soil were inoculated with 5% (v/DW) P. protegens in order to explore the effect of Pseudomonas on the microbial community structure. The sample of rhizosphere-associated microbiota was taken four weeks after the transplantation of tomato seedlings, ensuring that the root microbiota had been well established during this period . From each treatment, five root samples from five representative plants were harvested. Rhizosphere soil samples and the corresponding DNA extracts were investigated by using PowerSoil Soil DNA Isolation Kits (MoBio Laboratories Inc., Carlsbad, CA, USA) following the manufacturer’s protocol. Illumina MiSeq sequencing We generated the bacterial community profiles for each sample via PCR amplification of the 16S ribosomal RNA (rRNA) gene targeting regions V4 region using primers 515F and 806R (listed in Table ), followed by Illumina sequencing at Personal MAGIAENE Co. Ltd, Guangdong, China. The raw sequence data was processed using the QIIME2 pipeline, and the operational taxonomic unit (OTU) table was constructed with the UPARSE pipeline . Briefly, reads were truncated at 200 bp and were quality-filtered. After discarding replicates and singletons, the remaining reads were assigned to OTUs at a 97% identity threshold. We obtained 2,219,565 high-quality sequences from 15 samples. High-quality reads were then analyzed with USEARCH, where chimeric and organelle sequences were removed. Statistical analysis For statistical analyses, the software GraphPad PRISM 9.0 with an unpaired T-test was performed. Further, the RStudio 4.3 statistical software environment (R language version 4.3) was used for multiple comparisons. For the drawing of the HPLC peak diagram, Origin 2022 was used. Network analysis was carried out using Gephi 0.10. The screening for differential ASV was performed by using the STAMP software. The bacteria used in this study are listed in Table . Bacillus strains, and Escherichia coli strains were cultured by inoculating a single colony in 20 mL lysogeny broth (LB) (10 g L −1 NaCl, 5 g L −1 yeast extract, and 10 g L −1 tryptone) medium, and incubated overnight at 37 °C with shaking (200 rpm); P. protegens strains were cultured by inoculating a single colony in 20 mL King’ B (KB) (20.0 g L −1 tryptone, 10.0 g L −1 NaCl, 1.5 g L −1 MgSO 4 ·7H 2 O, 1.5 g L −1 K 2 HPO 4 ), and incubated overnight at 30 °C with shaking (200 rpm). Plasmid cloning was performed in E. coli DH5α (Table ). The cultured bacteria are regarded as cultures. Primers used for cloning and verification are found in Table . Pseudomonas and Bacillus with fluorescent proteins To visualize the development of bacterial strains co-cultivated on a solid agar surface, Pf-5 and its mutants were tagged with mCherry red fluorescent protein. The pBBR -mCherry plasmid was transformed into E. coli Top10, and the recombinant Pf-5/pBBR -mCherry , Δ pltB/ pBBR -mCherry , Δ gacA/ pBBR -mCherry strains were generated by triparental mating of Top10/pBBR- mCherry (donor), HB101/pRK - 2013 (helper), and the Pf-5, Δ pltB and Δ gacA strains of interest (recipient) , . In the same way, the mCherry fragment was amplified from the pBBR -mCherry plasmid to construct pUCP26 -mCherry . This plasmid was then transformed into E. coli Top10 and the recombinant Δ gacA-C /pUCP26- mCherry (Δ gacA-C-mcherry ) strain was generated according to the method described above. DMW1 was labeled with green fluorescence protein gfp by transfer of the pAD43-25 plasmid . The effect of P. protegens on bioactivity toward Bacillus spp. was determined in an inhibition assay on a solid medium. Cultures of Bacillus spp., as well as P. protegens Pf-5, were adjusted to an OD 600 nm = 1.0. Five μL of the P. protegens Pf-5 culture were placed onto 1.0% LB agar supplemented with 1.0% (v/v) of Bacillus sp. culture. The activity of P. protegens Pf-5 toward Bacillus spp. was determined as described above. The LB agar plates were incubated at 30 °C for 48 h, followed by measuring the size of inhibition zones. For bacterial co-cultivations, 1 mL cultures of P. protegens Pf-5 or its mutants labeled with mCherry , and B. velezensis DMW1 labeled with gfp were washed twice in sterilized water by centrifuging at 5000 rpm for 4 min, followed by discarding the supernatant and resuspending the pellet in sterilized water without the antibiotic. Subsequently, the optical density at 600 nm (OD 600 nm ) was adjusted to 1.0 ( Bacillus ) and 0.1 ( Pseudomonas ). For co-cultivations on agar surfaces, lines were drawn by streaking out P. protegens Pf-5 or its mutants labeled with mCherry , crossed by B. velezensis DMW1 labeled with gfp . The incubation time varied from 24 to 72 h, at 30 °C. Fluorescent colonies were imaged by fluorescent stereo microscopy (Nikon SMZ25, Nikon, Japan). Due to the excitation wavelength for red fluorescent protein being close to that of green protein, strong excitation (at 488 nm) can simultaneously excite both types of fluorescence. To determine the colony number of the bacteria, labeled with different fluorescent proteins, after contacting each other, the bacterial disks were taken at the midpoint of the segment formed by the intersection of two perpendicular lines after 48 h. Three beads (φ = 3 mm) were put in each tube, which was then filled with 1 mL of sterilized water. The bacterial disks in the sterilized water were subjected to grinding for 30 sec at 60 Hz using tissuelyser-64 (Shanghai Jingxin Technology). The grinding steps were repeated twice, and then the bacteria in the medium were released and diluted in sterilized water. According to the gradient dilution method, the bacterial suspensions were then spread on KB medium plates containing ampicillin (100 µg/mL) and gentamicin (25 µg/mL) for Pf-5, Δ gacA and Δ pltB labeled with mCherry , or ampicillin (100 µg/mL) and tetracycline (25 µg/mL) for Δ gacA -C labeled with mCherry . LB medium plates containing chloramphenicol (5 µg/mL) were used for B. velezensis DMW1 labeled with gfp . The plates were incubated at 30 °C ( Pseudomonas ), or 37 °C ( Bacillus ) for 12 h, and then the number of colonies was counted and statistically analyzed. To extract secondary metabolites of B. velezensis or P. protegens strains, bacteria were cultured by inoculating a single colony in 20 mL LB medium and incubated overnight at 37 °C under shaking (200 rpm). Then, the bacteria were further cultivated in 200 mL of LB medium for 24 h. Afterward, the pre-treated XAD16 resin was added and adjusted to 2.5%, and the mixture was incubated for another 24 h. Centrifuge at 8000 rpm for 15 min at room temperature, discard the supernatant, resuspend the mixed resin precipitate with 30–40 mL of methanol, and place it at 37 °C at 200 rpm for 4 h, then centrifuge at 8000 rpm at room temperature for 15 min. The supernatant was dissolved in methanol, and the crude extract was obtained. For the purification of iturin, surfactin, and fengycin, crude extract sample was dissolved in a methanol solution and subjected to purification through preparative high-performance liquid chromatography (HPLC) using a C18 column (5 µm, 100.0 mm × 21.2 mm; flow rate: 10 mL/min; gradient: 20–100% MeOH/H 2 O in 30 min, followed by 100% MeOH in 10 min) to generate sub-fractions. The distinctive peaks in the DMW1 sub-fraction underwent semipreparative HPLC (C18 column, 5 µm; 250.0 mm × 10.0 mm; flow rate: 3 mL/min; with 0.1% formic acid in 54% MeOH/H 2 O in 53 min, 65% MeOH/H 2 O in 53–70 min) to obtain a pure compound. NMR spectra were recorded in methanol-d4 on a Bruker AVANCE II 400 MHz instrument, and high-resolution mass spectra were acquired using an Agilent 6530 Accurate-Mass Q-TOF LC/MS coupled to an Agilent 1260 HPLC. The effect of B. velezensis DMW1, and its secondary metabolites extracts on P. protegens Pf-5 was determined in an inhibition assay on a solid medium. Cultures of B. velezensis DMW1, as well as P. protegens Pf-5 were normalized to an optical density at 600 nm (OD 600 nm ) of 1.0. Five μL of the culture of DMW1 or its secondary metabolites extracts were applied to a paper disk (φ = 5 mm) and then placed onto LB agar containing 1.0% (v/v) of P. protegens Pf-5 culture. The activity of P. protegens Pf-5 or its secondary metabolites extracts towards B. velezensis DMW1 was determined using the same method. Plates were incubated at 30 °C for 48 h, followed by measuring the size of inhibition zones. To explore the effect of secondary metabolites of P. protegens Pf-5 on the activity of B. velezensis DMW1, several Pseudomonas deletion mutants were generated by triparental mating according to the description above , . Mutants of P. protegens Pf-5 were created by the triparental mating. Briefly, the upstream (770 bp) and downstream (770 bp) fragments were amplified by PCR, fused via overlap PCR, and introduced into the pK18mobsacB vector. The recombinant plasmid was transferred from E. coli Top10 to P. protegens Pf-5, with the help of E. coli Top10 (pRK-2013) , . The following mutations were introduced: gacA (secondary metabolite regulatory system) , ofa (orfamide A biosynthetic gene) , pltB (pyoluteorin biosynthetic gene) , phlA (2,4-diacetylphloroglucinol biosynthetic gene) , prnA (pyrrolnitrin biosynthetic gene) , rzxB (rhizoxin biosynthetic gene) , hcnABC (VOC hydrogen cyanide biosynthetic gene) , pFL4656 (non-ribosomal peptide biosynthetic gene). For the purification of pyoluteorin, high-performance liquid chromatography (HPLC, Waters), equipped with a reverse-phase column (ZORBAX SB-C18) was performed. The running program was a gradient elution from 5% solvent A (HPLC-grade acetonitrile containing 0.1% trifluoroacetic acid), 95% solvent B (Milli-Q water containing 0.1% trifluoroacetic acid) to 95% A, 5% B for 20 min. A concentration of 95% solvent A and 5% solvent B were then held for 5 min. A gradient elution from 95% solvent A, 5% solvent B to 5% A, 95% B for 1 min; A concentration of 95% solvent A and 5% solvent B were then held for 2 min. For the purification of pyoluteorin, the semipreparative HPLC was used with a flow rate of 4 mL/min and an injection volume of 1 mL. Pyoluteorin was done under UV absorption at 310 nm. In the last step, purified compounds were identified via ultraperformance liquid chromatography-mass spectrometry (UPLC-MS). B. velezensis DMW1 was grown as a single culture or co-cultured with P. protegens Pf-5 or Δ pltB , respectively. Two mL of the culture, containing either the single strain or a mix of B. velezensis DMW1 (OD 600 nm = 1.0) and P. protegens (OD 600 nm = 0.1) (volume ratio = 19:1), were used for inoculation of 200 mL medium as described previously . The cultures were grown at 37 °C under shaking (200 rpm) for 12 h. The bacteria were collected for total RNA extraction and RNA-seq by the DNBSEQ platform (BGI Genomics Co., Ltd., China). The reads were mapped to the DMW1 genome v1.1 ( https://www.ncbi.nlm.nih.gov/nuccore/NZ_CP114180.1 ) using HISAT2. Utilizing the SOAPnuke tool, the expression levels of individual DMW1 genes were quantified through normalization to the “Fragments Per Kilobase of exon per Million mapped reads (FPKM)” metric, enabling the identification of differentially expressed genes (DEGs). The genes with at least two-fold change and p- value ≤ 0.05 were considered DGEs between the two samples. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs was performed using the “KEGG enrichment analysis” tool embedded in the BGI platform ( https://report.bgi.com/ps/mrna/index.html ). To investigate srf , itu, fen , mln , bae , dfn , flgD , and fliR gene expression in Bacillus , firstly RNA extraction was carried out using the Bacteria Total RNA Kit (ZP403) (Zoman Biotechnology Co., Ltd; Beijing) following the Gram-positive manufacturer’s protocol. RNA quality and quantity were performed with a Thermo Scientific NanoDrop 2000 UV-vis Spectrophotometer. Primer 3 program available online was used for primer design and primers were synthesized by GenScript. The primers used for this purpose are listed in Table . Reverse transcription-polymerase chain reaction (RT-qPCR) was performed to quantify gene expression of itu (iturin synthesis), fen (fengycin synthesis), srf (surfactin synthesis), bae (bacillaene synthesis), mln (marcolactin synthesis), dfn (difficidin synthesis), flgD (flagellar hook assembly protein FlgD) and fliR (flagellar type III secretion system protein FliR) . Reverse transcriptase and RT-qPCR reactions were conducted using the HiScript III 1st strand cDNA Synthesis Kit (+gDNA) (Vazyme, China). The qPCR reaction was performed in a total volume of 20 μL: 10 μL of ChamQ Universal SYBR qPCR Master Mix, 0.4 μL of each primer (5 μM), 2 μL of cDNA (100 ng), 7.2 µL of Nuclease-free water. The thermal cycling program applied on the ABI StepOne was: preheating at 95 °C for 30 s, (95 °C for 5 s and 60 °C for 30 s) × 40 cycles, and the specificity of the PCR product amplification was verified based on the Tm value for 15 s at 95 °C and 1 min at 60 °C. Finally, the qPCR amplification was run on the ABI StepOne qPCR instrument (Applied Biosystems) with software version 2.3. The relative gene expression analysis was conducted using the 2 –ΔΔCT method with the rpsU gene as a housekeeping gene to normalize mRNA levels between different samples . Using a C18 reversed-phase column (ZORBAX SB-C18), we performed a gradient elution starting from 30% solvent A (HPLC-grade acetonitrile with 0.1% trifluoroacetic acid) to 70% solvent B (Milli-Q water with 0.1% trifluoroacetic acid) over a period of 30 min. This was followed by a gradient change to 95% solution A and 5% solution B, maintained for 10 min. The flow rate was set at 1 mL/min. Secondary metabolites were detected at UV wavelengths of 210 nm and 280 nm, respectively, and the peak area represented the production. Bacillus – Pseudomonas interactions For the biofilm assay, 200 μL overnight culture was transferred into a well containing 20 mL liquid LB medium. After incubation at 30 °C/37 °C, 200 rpm for OD 600 nm = 1.0, 10 μL B. velezensis DMW1 (OD 600 nm = 1.0) and 10 μL P. protegens Pf-5/Δ pltB (OD 600 nm = 0.25) culture or sterilized water was added into 2 mL fresh LBGM medium (LB supplemented with 1% glycerol and 100 μM MnSO 4 ) in a flat bottom 12-well microplate . To visualize the distribution of two bacteria in the cultures, the culture was examined with a Zeiss LSM 700 microscope, equipped with a 20× objective. GFP fluorescence was excited at 488 nm and detected between 500 and 530 nm; for mCherry, excitation was at 561 nm with emission detection from 580 to 620 nm after incubation at 30 °C for 24 h without shaking. Biofilm-related gene expression level among the different treatments was analyzed at the transcriptional level by RT-qPCR. After ligating the target fragment into the pMD-18T vector, the resulting construct was transformed into DH5α competent cells. Plasmids from the transformants were extracted and then diluted in a 10-fold gradient. These diluted plasmids were subsequently used as templates for fluorescence quantitative PCR amplification. The logarithm of the initial copy number of the template was plotted on the x-axis, while the number of cycles reaching the threshold (Threshold cycle, Ct) was plotted on the y-axis, thus generating a standard curve. The CT number was introduced into the standard curve equation to calculate the initial gene copy number of the sample to reveal the gene copy number per gram of dry soil . To determine the colony number of tomato root colonizing B. velezensis DMW1 and P. protegens Pf-5/Δ pltB in the rhizosphere soil, thirty-day-old tomato-plant seedling roots were dipped into the bacterial suspension containing 5 × 10 7 cells/mL Bacillus and 3 × 10 8 cells/mL Pseudomonas for 40 min. Afterward, all seedlings were transplanted into pots (10 cm in diameter) containing 200 g sterilized peat soil from Danish Pindstrup company (located in the Kingdom of Denmark), and incubated in the greenhouse (30 °C, 75% relative humidity, 14 h light/10 h darkness). Rhizosphere soil samples from these treatments were collected 7 d after transplanting. Five tomato roots were collected from the points of each plot and vigorously shaken to remove excess soil. Then, the soil adhering to the roots (rhizosphere soil) was suspended in phosphate-buffered saline (PBS). The rhizosphere soil suspensions were centrifuged at 10,000 × g for 10 min, and the sediments were stored at −80 °C for later DNA extraction. Soil samples from the pot experiments were chosen for subsequent DNA extraction using Power Soil DNA Isolation Kits (MoBio Laboratories Inc., Carlsbad, CA, USA) following the manufacturer’s protocol. The concentration and quality of the DNA were determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, NC, USA). The abundance of Pseudomonas spp. was quantified by quantitative polymerase chain reaction (qPCR) with specific primers . The abundance of Bacillus spp. was quantified by qPCR with specific primers . For imaging of B. velezensis DMW1/ pAD43-25-gfp , and P. protegens Pf-5 or Δ pltB / pBBR-mCherry colonization in the tomato root, surface-sterilized tomato seeds with roots 10 mm long (3 d) were inoculated with suspensions of B. velezensis DMW1 (5 × 10 7 cells), and P. protegens Pf-5/Δ pltB (3 × 10 8 cells). After incubation for 30 min at 25 °C and shaking with 100 rpm, seedlings were transferred onto square Petri dishes (12 × 12 cm) containing 0.5× Murashige and Skoog (MS) semisolid agar medium (0.7% agar) without sucrose . The square plates were kept in a vertical position during the incubation time of seven days at 25 °C under long daylight conditions (16 h light/8 h darkness) in a plant-growth chamber. To determine the spatial pattern of root colonization by DMW1/ pAD43-25-gfp and Pf-5/ pBBR-mCherry or Δ pltB/pBBR-mCherry , the root system (ca. 10 mm long) was imaged using a Zeiss LSM 700. GFP fluorescence was excited at 488 nm and detected between 500 and 530 nm; for mCherry, excitation was at 561 nm with emission detection from 580 to 620 nm. To determine the root colonization by B. velezensis DMW1 and P. protegens Pf-5/Δ pltB , the root system (ca. 10 mm long) was subjected to grind for 30 sec at 60 Hertz in a 2 mL tube containing 1 mL of distilled water and three beads (φ = 3 mm) using tissuelyser-64 (Shanghai Jingxin Technology). Using the gradient dilution method, the diluted bacterial suspensions were placed onto LB plates. The plates were incubated at 30 °C/37 °C for 12 h to count the number of colonies. Thirty-day-old tomato seedlings (variety cultivar ‘Mao fen’) roots were inoculated with either: 10 mL 1 × 10 7 CFU/mL DMW1, or 10 mL 1 × 10 8 CFU/mL Pf-5, or 10 mL 1 × 10 8 CFU/mL Δ pltB culture, or a mix of DMW1 and Pf-5 or Δ pltB . Then, 10 mL 5 × 10 7 CFU/mL R. solanacearum suspension was added. Each treatment was repeated in triplicate, and every repeat contained 15 seedlings. All seedlings were maintained in the greenhouse (30 °C, 75% relative humidity, 14 h light/10 h darkness), and inspected periodically until disease symptoms appeared. The disease severity was evaluated according to an empirical scale: Level 0 = tomato plants without visible symptoms; Level 1 = striped necrosis on stems occasionally or less than half of the leaves wilted on unilateral stems; Level 2 = black streaks less than half the height of the stem or between half to two-thirds of the leaves wilted on unilateral stems; Level 3 = more than two-thirds of the leaves wilted on unilateral stems; Level 4 = the plant is dead . The disease severity index was calculated according to the following formula: 1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm{Disease}}\; {\rm{severity}}\; {\rm{index}}\,({\rm{DSI}})=[\Sigma ({\rm{x}}\times {\rm{y}})/({\rm{z}}\times 4)]\times 100$$\end{document} Disease severity index ( DSI ) = [ Σ ( x × y ) / ( z × 4 ) ] × 100 2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm{Disease}}\; {\rm{control}}\; {\rm{effect}}\,=({{\rm{DSI}}}_{{\rm{Treatment}}}-{{\rm{DSI}}}_{{\rm{Control}}})/{{\rm{DSI}}}_{{\rm{Control}}}\times 100 \%$$\end{document} Disease control effect = ( DSI Treatment − DSI Control ) / DSI Control × 100 % Where: x = number of different degrees infected plants in the treatments; y = relative degree value; and z = number of total plants in the treatments. Tomato seedlings (variety cultivar ‘Mao fen’) grown in natural soil were inoculated with 5% (v/DW) P. protegens in order to explore the effect of Pseudomonas on the microbial community structure. The sample of rhizosphere-associated microbiota was taken four weeks after the transplantation of tomato seedlings, ensuring that the root microbiota had been well established during this period . From each treatment, five root samples from five representative plants were harvested. Rhizosphere soil samples and the corresponding DNA extracts were investigated by using PowerSoil Soil DNA Isolation Kits (MoBio Laboratories Inc., Carlsbad, CA, USA) following the manufacturer’s protocol. We generated the bacterial community profiles for each sample via PCR amplification of the 16S ribosomal RNA (rRNA) gene targeting regions V4 region using primers 515F and 806R (listed in Table ), followed by Illumina sequencing at Personal MAGIAENE Co. Ltd, Guangdong, China. The raw sequence data was processed using the QIIME2 pipeline, and the operational taxonomic unit (OTU) table was constructed with the UPARSE pipeline . Briefly, reads were truncated at 200 bp and were quality-filtered. After discarding replicates and singletons, the remaining reads were assigned to OTUs at a 97% identity threshold. We obtained 2,219,565 high-quality sequences from 15 samples. High-quality reads were then analyzed with USEARCH, where chimeric and organelle sequences were removed. For statistical analyses, the software GraphPad PRISM 9.0 with an unpaired T-test was performed. Further, the RStudio 4.3 statistical software environment (R language version 4.3) was used for multiple comparisons. For the drawing of the HPLC peak diagram, Origin 2022 was used. Network analysis was carried out using Gephi 0.10. The screening for differential ASV was performed by using the STAMP software. Supplementary Material |
Development of water safety risk matrices to improve water safety in Arctic drinking water systems in Nunavut, Canada | fa66529f-ed33-4306-8eae-01dde05fba0e | 11748980 | Microbiology[mh] | Drinking water management in Nunavut Nunavut’s drinking water programme is in the process of modernising, aiming to align with practices in Southern Canada, but faces challenges due to its remote geography, extreme weather and limited resources. Nunavut’s 25 communities are located in the Canadian Arctic with no road connections between them, and populations ranging from 160 to 8,296 residents . Only surface water sources are used and at this stage of the territorial government’s infrastructure upgrade initiative, 10 communities continue to rely solely on chlorine disinfection as the treatment process before drinking water is distributed to users . Water is delivered to individual households and buildings in most communities exclusively by truck or through a combination of trucks and partially piped distribution systems . Nunavut municipalities are responsible for drinking water quality sampling from regulated locations, which include source water and after chlorine disinfection at the end of treatment, either in a treated water storage tank, for communities with this included in their treatment process, or in a delivery truck after sufficient mixing time . Chlorine and turbidity are monitored daily, bacteriological samples are taken monthly and until 2021 the chemical and physical parameter testing was only required biennially based on the existing Public Water Supply Regulations (PWSR) that have not been substantively updated since 1990 . Through an interim directive issued by the Government of Nunavut Department of Health (GN – Health), the mandatory frequency of sampling has since been increased and the list of parameters has been updated to reflect Health Canada’s Guidelines for Canadian Drinking Water Quality (GCDWQ) . This additional data aid in establishing a baseline characterisation of the source water, allow for better monitoring of the efficacy of water treatment upgrades and help in understanding how that may change over time due to climate change impacts on source water quality. However, the sampling programme does not monitor delivered water in domestic cisterns within buildings, and for most treatment systems, the barriers in place may be insufficient to reduce parameters of concern even if exceedances are identified. Water governance in Nunavut In Canada, the regulation of drinking water falls to individual provinces and territories other than on First Nations which are federally managed but not yet regulated . In Nunavut, 24 communities do not generate revenue through collection of taxes. Therefore, water supply infrastructure is funded and constructed by the Government of Nunavut Department of Community and Government Services (GN-CGS) with the goal of transferring ownership to a given municipality after commissioning. In eight municipalities, the local administration has relinquished water infrastructure ownership and/or operations back to GN-CGS . The GN provides the bulk of operational funding to municipalities, with some cost recovery coming from water delivery revenue. The GN also oversees the capital planning for municipal assets, determining the order in which requests for water infrastructure funding are presented to the legislative assembly for approval. Furthermore, the GN also acts as the drinking water regulator through GN-Health. Drinking water quality monitoring in Nunavut The remoteness and limited labour capacity in Nunavut pose challenges for completing drinking water quality samples. GN-CGS provides water operations support to all communities excluding Iqaluit, which includes assisting operators to perform mandatory sampling using correct techniques and improving data management. In recent years, GN-CGS began collating daily free chlorine (FC) and turbidity readings submitted by operators, along with sampling results from accredited laboratories. These records have been used by the authors to understand the known water quality hazards. The focus of GN-Health’s enforcement has historically centred on microbiological parameters sampled after chlorine disinfection, prior to distribution. Testing of individual building tanks and taps is not regulated by GN-Health nor is tank maintenance; therefore, testing and cleaning by public and private building owners in Nunavut is voluntary. The focus on bacteriological monitoring is also seen in peer-reviewed literature but extends past the end-of-treatment sample point, particularly in the Qikiqtaaluk and Kivalliq Regions of Nunavut , as well as neighbouring Inuit-majority jurisdictions in Nunavik (Northern Québec) and Nunatsiavut (Northern Labrador) . Daley et al. found total coliforms in building tap samples taken in Coral Harbour, Iqaluit and Pond Inlet, and E. coli in Coral Harbour and Pond Inlet . Gora et al. also had positive hits of E. coli in building cisterns in Pond Inlet . Masina et al. found that untreated traditional surface water sources in Iqaluit, separate from the municipal source, had Giardia (20.0%) and Cryptosporidium (1.8%) prevalence. Although this was identified as low risk to Inuit residents that consume these alternative sources in Iqaluit, with climate change it is hypothesised that the risk will increase . Currently, there are no available data on the presence of Cryptosporidium in Nunavut’s municipal drinking water sources or on the rates of cryptosporidiosis linked to exposure through these water systems. Martin et al. performed analyses on household water cisterns in four Nunavik communities. Of the 64 cisterns, 21 contained total coliforms, but none tested positive for E. coli . Beyond microbiological parameters, a 2021 hydrocarbon contamination event in Iqaluit underscored the importance of chemical water safety hazard monitoring . For communities with only chlorination, there is a limited reduction in metals, particulate matter and most dissolved organic compounds, all of which can be associated with adverse health impacts at high levels . Currently, few studies have examined the chemical water characteristics in Nunavut, making it difficult to understand the extent of existing hazards and the risk they pose . Water safety planning and water safety risk assessment The World Health Organization (WHO) promotes the use of water safety plans (WSPs) as a proactive approach to manage risks that may threaten drinking water safety . The WSP framework is built on stakeholder system knowledge, hazard identification and risk mitigation . The hazard identification process involves creating a list of actual and potential dangers and their causes. The hazards then undergo a risk assessment using a matrix with defined scoring criteria. An initial WSP step is to assemble a team consisting of members that represent different parts of a water programme such as operators, owners and stakeholders who provide funding or oversight. Forming this team and assigning resources to address identified water safety hazards may be a complicated undertaking in Nunavut due to its inconsistent division of responsibilities and decision-making power. The WSP approach has been adopted and adapted in different jurisdictions throughout the world . To date, Alberta is the only Canadian province using WSPs in its drinking water programme, starting in 2011 . A 2015 study was conducted to evaluate their impact, finding that WSPs improved communication between operators and decision-makers, strengthening long-term planning for infrastructure maintenance and risk mitigation to better protect public health . However, resource constraints emerged as an obstacle to full implementation, particularly in small communities. However, since 2017, there have been a number of studies performed in Canadian Indigenous communities testing the benefits of grassroots management approaches like WSPs, including a review of the potential application of WSPs in Arctic communities and First Nations . Lane et al. have also evaluated how well various WSP matrices have been constructed to quantify risk . This study builds on this previous work by reviewing and comparing existing WSP approaches to develop a risk matrix and WSP framework using Nunavut as a research setting. However, the resource constraints, logistical challenges and identified water safety hazards are not exclusive to Nunavut, suggesting that the framework may be applicable more broadly. Notably, this framework explores a novel approach to hazard risk scoring by considering the presence of hazard barriers, an aspect not typically addressed in existing WSP frameworks. Positionality statement The authors of this paper identify themselves as settler researchers from York University, Lassonde School of Engineering based in Toronto. Both the lead author and second author have worked/are currently working at GN-CGS in the roles of senior municipal planning officer and municipal support, respectively. Nunavut’s drinking water programme is in the process of modernising, aiming to align with practices in Southern Canada, but faces challenges due to its remote geography, extreme weather and limited resources. Nunavut’s 25 communities are located in the Canadian Arctic with no road connections between them, and populations ranging from 160 to 8,296 residents . Only surface water sources are used and at this stage of the territorial government’s infrastructure upgrade initiative, 10 communities continue to rely solely on chlorine disinfection as the treatment process before drinking water is distributed to users . Water is delivered to individual households and buildings in most communities exclusively by truck or through a combination of trucks and partially piped distribution systems . Nunavut municipalities are responsible for drinking water quality sampling from regulated locations, which include source water and after chlorine disinfection at the end of treatment, either in a treated water storage tank, for communities with this included in their treatment process, or in a delivery truck after sufficient mixing time . Chlorine and turbidity are monitored daily, bacteriological samples are taken monthly and until 2021 the chemical and physical parameter testing was only required biennially based on the existing Public Water Supply Regulations (PWSR) that have not been substantively updated since 1990 . Through an interim directive issued by the Government of Nunavut Department of Health (GN – Health), the mandatory frequency of sampling has since been increased and the list of parameters has been updated to reflect Health Canada’s Guidelines for Canadian Drinking Water Quality (GCDWQ) . This additional data aid in establishing a baseline characterisation of the source water, allow for better monitoring of the efficacy of water treatment upgrades and help in understanding how that may change over time due to climate change impacts on source water quality. However, the sampling programme does not monitor delivered water in domestic cisterns within buildings, and for most treatment systems, the barriers in place may be insufficient to reduce parameters of concern even if exceedances are identified. In Canada, the regulation of drinking water falls to individual provinces and territories other than on First Nations which are federally managed but not yet regulated . In Nunavut, 24 communities do not generate revenue through collection of taxes. Therefore, water supply infrastructure is funded and constructed by the Government of Nunavut Department of Community and Government Services (GN-CGS) with the goal of transferring ownership to a given municipality after commissioning. In eight municipalities, the local administration has relinquished water infrastructure ownership and/or operations back to GN-CGS . The GN provides the bulk of operational funding to municipalities, with some cost recovery coming from water delivery revenue. The GN also oversees the capital planning for municipal assets, determining the order in which requests for water infrastructure funding are presented to the legislative assembly for approval. Furthermore, the GN also acts as the drinking water regulator through GN-Health. The remoteness and limited labour capacity in Nunavut pose challenges for completing drinking water quality samples. GN-CGS provides water operations support to all communities excluding Iqaluit, which includes assisting operators to perform mandatory sampling using correct techniques and improving data management. In recent years, GN-CGS began collating daily free chlorine (FC) and turbidity readings submitted by operators, along with sampling results from accredited laboratories. These records have been used by the authors to understand the known water quality hazards. The focus of GN-Health’s enforcement has historically centred on microbiological parameters sampled after chlorine disinfection, prior to distribution. Testing of individual building tanks and taps is not regulated by GN-Health nor is tank maintenance; therefore, testing and cleaning by public and private building owners in Nunavut is voluntary. The focus on bacteriological monitoring is also seen in peer-reviewed literature but extends past the end-of-treatment sample point, particularly in the Qikiqtaaluk and Kivalliq Regions of Nunavut , as well as neighbouring Inuit-majority jurisdictions in Nunavik (Northern Québec) and Nunatsiavut (Northern Labrador) . Daley et al. found total coliforms in building tap samples taken in Coral Harbour, Iqaluit and Pond Inlet, and E. coli in Coral Harbour and Pond Inlet . Gora et al. also had positive hits of E. coli in building cisterns in Pond Inlet . Masina et al. found that untreated traditional surface water sources in Iqaluit, separate from the municipal source, had Giardia (20.0%) and Cryptosporidium (1.8%) prevalence. Although this was identified as low risk to Inuit residents that consume these alternative sources in Iqaluit, with climate change it is hypothesised that the risk will increase . Currently, there are no available data on the presence of Cryptosporidium in Nunavut’s municipal drinking water sources or on the rates of cryptosporidiosis linked to exposure through these water systems. Martin et al. performed analyses on household water cisterns in four Nunavik communities. Of the 64 cisterns, 21 contained total coliforms, but none tested positive for E. coli . Beyond microbiological parameters, a 2021 hydrocarbon contamination event in Iqaluit underscored the importance of chemical water safety hazard monitoring . For communities with only chlorination, there is a limited reduction in metals, particulate matter and most dissolved organic compounds, all of which can be associated with adverse health impacts at high levels . Currently, few studies have examined the chemical water characteristics in Nunavut, making it difficult to understand the extent of existing hazards and the risk they pose . The World Health Organization (WHO) promotes the use of water safety plans (WSPs) as a proactive approach to manage risks that may threaten drinking water safety . The WSP framework is built on stakeholder system knowledge, hazard identification and risk mitigation . The hazard identification process involves creating a list of actual and potential dangers and their causes. The hazards then undergo a risk assessment using a matrix with defined scoring criteria. An initial WSP step is to assemble a team consisting of members that represent different parts of a water programme such as operators, owners and stakeholders who provide funding or oversight. Forming this team and assigning resources to address identified water safety hazards may be a complicated undertaking in Nunavut due to its inconsistent division of responsibilities and decision-making power. The WSP approach has been adopted and adapted in different jurisdictions throughout the world . To date, Alberta is the only Canadian province using WSPs in its drinking water programme, starting in 2011 . A 2015 study was conducted to evaluate their impact, finding that WSPs improved communication between operators and decision-makers, strengthening long-term planning for infrastructure maintenance and risk mitigation to better protect public health . However, resource constraints emerged as an obstacle to full implementation, particularly in small communities. However, since 2017, there have been a number of studies performed in Canadian Indigenous communities testing the benefits of grassroots management approaches like WSPs, including a review of the potential application of WSPs in Arctic communities and First Nations . Lane et al. have also evaluated how well various WSP matrices have been constructed to quantify risk . This study builds on this previous work by reviewing and comparing existing WSP approaches to develop a risk matrix and WSP framework using Nunavut as a research setting. However, the resource constraints, logistical challenges and identified water safety hazards are not exclusive to Nunavut, suggesting that the framework may be applicable more broadly. Notably, this framework explores a novel approach to hazard risk scoring by considering the presence of hazard barriers, an aspect not typically addressed in existing WSP frameworks. Positionality statement The authors of this paper identify themselves as settler researchers from York University, Lassonde School of Engineering based in Toronto. Both the lead author and second author have worked/are currently working at GN-CGS in the roles of senior municipal planning officer and municipal support, respectively. The authors of this paper identify themselves as settler researchers from York University, Lassonde School of Engineering based in Toronto. Both the lead author and second author have worked/are currently working at GN-CGS in the roles of senior municipal planning officer and municipal support, respectively. Nunavut hazards identification and development of hypothetical hazard event scenarios Initial hazard identification for the Nunavut communities was compiled from literature , historical operational data and internal reports provided by GN-CGS and the authors’ professional experiences working with Nunavut water treatment systems. Six hypothetical scenarios were developed to assess the applicability of existing WSP risk matrices and to create and validate a suitable water safety risk evaluation approach. The scenarios are based on observed events in Nunavut, representing both common water quality issues found in many jurisdictions and rarer, more context-specific hazards also observed in other regions . A key consideration was that the scenarios should represent a range of “qualitatively different” hazards as described by Lane and Hrudey . Their work highlighted how qualitatively different hazards can be assigned equivalent risk scores due to ambiguity or inadequate resolution in the definitions of the underlying likelihood and severity scores. For example, a catastrophic (high severity), one-time (low likelihood) microbial outbreak could receive the same risk score as a frequent (high likelihood) exceedance of an aesthetic objective (low severity). In the drinking water context, this could lead to the misallocation of resources, failing to effectively reduce public health risks. Review of WSP approaches in other jurisdictions Twelve water safety matrices from global jurisdictions, as reported and evaluated by Lane and Hrudey , were reviewed for this study. The criteria within each likelihood and consequence definition were initially grouped into broad categories to identify what each jurisdiction considers relevant for assessing risk. Next, the matrices from Alberta, Australia, Ireland, South Africa and the WHO were chosen for a more detailed review because they had sufficiently different score definitions. Three different approaches were used to observe scoring differences between the risk matrices. For the first approach, each rubric was used to score the hazard scenarios; the risk scores are presented in Table S1. The results were then categorised into the matrix’s prescribed risk level based on the ranges in its specific scoring system (Low, Medium, High, Very High). For the second approach, the risk level categories were modified to align with the Cox criteria for construction of risk matrices , where necessary, and the risks were then re-categorised. The final approach focused on observing the alignment between the rubrics when progressing from one score step to the next. Since some jurisdictions use non-linear or non-numerical risk levels, rather than using the actual value of each score step it was necessary to set the progression from one score step to the next as functionally equivalent, such that the different matrices could be compared. For example, in the WHO rubric, the fourth likelihood score is “4”, Australia’s is “B” and Alberta’s is “8”, excluding the “Not Applicable” zero score. These scores were all standardised to “Step 4”. The zero score was disregarded because all hypothetical scenarios had occurred, allowing for only five score steps for likelihood and severity. The score values from approach 1 were then converted to score steps. Development of a Nunavut-specific WSP framework A Nunavut-specific WSP framework was developed based on several key principles derived from insights gained through the jurisdictional rubric evaluation and literature review. The framework: Encompasses potential hazards observed in Nunavut based on government water quality monitoring, previous research studies and input from water system stakeholders Acknowledges territory-specific regulatory and governance structures Can be applied to multiple communities and tailored to individual communities based on community input Limits speculation and bias in scoring:a. Frequency: Numeric criteria that incorporates past observationb. Consequence: Linked to regulation, best practices, potential health outcomes and differentiates between acute and chronic health issues Accounts for how well the existing water system can manage a given hazard (system readiness) Provides a way to track the impact of system improvements on water safety risk Meets the risk-level range criteria as described by Cox and expanded upon in Lane and Hrudey The proposed framework and scoring criteria were developed based on the principles above. The severity scoring criteria have been adapted based on existing or planned updates to regulations and practice as identified in the GN Drinking Water Strategic Framework (DWSF). The likelihood scoring criteria consider both past frequency and the presence of existing barriers as separately weighted factors. A sensitivity analysis was conducted to assess the impact of varying the weights of the two factors on the resulting numeric water safety risk score. Initial hazard identification for the Nunavut communities was compiled from literature , historical operational data and internal reports provided by GN-CGS and the authors’ professional experiences working with Nunavut water treatment systems. Six hypothetical scenarios were developed to assess the applicability of existing WSP risk matrices and to create and validate a suitable water safety risk evaluation approach. The scenarios are based on observed events in Nunavut, representing both common water quality issues found in many jurisdictions and rarer, more context-specific hazards also observed in other regions . A key consideration was that the scenarios should represent a range of “qualitatively different” hazards as described by Lane and Hrudey . Their work highlighted how qualitatively different hazards can be assigned equivalent risk scores due to ambiguity or inadequate resolution in the definitions of the underlying likelihood and severity scores. For example, a catastrophic (high severity), one-time (low likelihood) microbial outbreak could receive the same risk score as a frequent (high likelihood) exceedance of an aesthetic objective (low severity). In the drinking water context, this could lead to the misallocation of resources, failing to effectively reduce public health risks. Twelve water safety matrices from global jurisdictions, as reported and evaluated by Lane and Hrudey , were reviewed for this study. The criteria within each likelihood and consequence definition were initially grouped into broad categories to identify what each jurisdiction considers relevant for assessing risk. Next, the matrices from Alberta, Australia, Ireland, South Africa and the WHO were chosen for a more detailed review because they had sufficiently different score definitions. Three different approaches were used to observe scoring differences between the risk matrices. For the first approach, each rubric was used to score the hazard scenarios; the risk scores are presented in Table S1. The results were then categorised into the matrix’s prescribed risk level based on the ranges in its specific scoring system (Low, Medium, High, Very High). For the second approach, the risk level categories were modified to align with the Cox criteria for construction of risk matrices , where necessary, and the risks were then re-categorised. The final approach focused on observing the alignment between the rubrics when progressing from one score step to the next. Since some jurisdictions use non-linear or non-numerical risk levels, rather than using the actual value of each score step it was necessary to set the progression from one score step to the next as functionally equivalent, such that the different matrices could be compared. For example, in the WHO rubric, the fourth likelihood score is “4”, Australia’s is “B” and Alberta’s is “8”, excluding the “Not Applicable” zero score. These scores were all standardised to “Step 4”. The zero score was disregarded because all hypothetical scenarios had occurred, allowing for only five score steps for likelihood and severity. The score values from approach 1 were then converted to score steps. A Nunavut-specific WSP framework was developed based on several key principles derived from insights gained through the jurisdictional rubric evaluation and literature review. The framework: Encompasses potential hazards observed in Nunavut based on government water quality monitoring, previous research studies and input from water system stakeholders Acknowledges territory-specific regulatory and governance structures Can be applied to multiple communities and tailored to individual communities based on community input Limits speculation and bias in scoring:a. Frequency: Numeric criteria that incorporates past observationb. Consequence: Linked to regulation, best practices, potential health outcomes and differentiates between acute and chronic health issues Accounts for how well the existing water system can manage a given hazard (system readiness) Provides a way to track the impact of system improvements on water safety risk Meets the risk-level range criteria as described by Cox and expanded upon in Lane and Hrudey The proposed framework and scoring criteria were developed based on the principles above. The severity scoring criteria have been adapted based on existing or planned updates to regulations and practice as identified in the GN Drinking Water Strategic Framework (DWSF). The likelihood scoring criteria consider both past frequency and the presence of existing barriers as separately weighted factors. A sensitivity analysis was conducted to assess the impact of varying the weights of the two factors on the resulting numeric water safety risk score. Water safety hazards in Nunavut A preliminary list of hazards was compiled in Table S.2. It includes common hazards for surface water supply systems, not necessarily confirmed in Nunavut but may not have been adequately assessed for elimination, as well as hazards reported in previous academic studies, consultant reports and government records. For example, while the available literature does not confirm the presence of Cryptosporidium or Giardia in Nunavut’s licenced potable water sources , these pathogens are commonly found in surface water. As a precaution, the GN has started incorporating treatment barriers specifically designed to protect against these pathogens in newly constructed water treatment facilities . Conversely, incidences of high manganese in treated water and chlorine residuals below the regulated minimum in treated water distribution trucks have been established within literature and confirmed by GN records. Government oversight, and therefore records, is limited to source water, treatment and distribution, so hazards present in buildings, such as bacteria in cisterns and lead in tap water have been identified solely from academic studies. Beyond quality-focused hazards, service interruption hazards that impact delivery to consumers are recognised in Table S2, as well. Many of the identified hazards are relevant to all Nunavut communities; however, engagement with water professionals and system users in individual communities will be required to confirm the list and subsequently the likelihood and consequence scores of each hazard. The list is expected to change with input from community stakeholders and territorial authorities but can be used to facilitate initial discussions and raise awareness and build capacity and knowledge within the groups. The remedial actions identified in Table S.2 are measures currently in place or actions that could be taken by a community in the interim to improve water safety, prior to completing an upgrade to the existing infrastructure. Scenarios The six hypothetical scenarios below were developed from the hazards listed in Table S.2 but are not directly based upon a specific water system in Nunavut. Rather, the scenarios were used to test the different risk matrix systems. Scenario 1: low chlorine residual in water delivery truck A community water system consists of a surface water supply, an intake pipe and pump, chlorination, three trucks and 120 individual building water cisterns. Regulations dictate that 20 min of contact time with a minimum outlet FC residual of 0.2 mg/L is required to achieve disinfection. At least once a month, the water exiting one or more of the trucks has FC less than 0.2 mg/L after contact time. Municipal truck drivers are required to measure and record FC levels daily, but they do so only once a month, resulting in incomplete records. Scenario 2: manganese in water supply Every spring surface water turns over causing increased levels of manganese to enter the treatment system which does not have a removal process. Manganese remains in the water are delivered to community buildings, causing discolouration and staining, reducing consumer faith in their drinking water. Water is sampled once a year, but manganese does not have an aesthetic limit in the existing regulations. Scenario 3: E. coli in domestic water storage cisterns Opportunistic sampling by GN-Health in a community with a surface water supply, chlorination and trucked distribution discovers E. coli in the cisterns of three public buildings, confirmed through follow-up samples. The cisterns are sampled approximately once a year by the GN Health officer or voluntarily by the building owner, and this is the first time that E. coli has been detected. GN-Health issues a boil water advisory (BWA) and investigates the contamination, which is eventually attributed to inconsistent cistern cleaning, which is in turn linked to the installation of cistern models that are difficult to clean. Because it is determined that the contamination is localised, the BWA is removed from the community. The cisterns are thoroughly cleaned and resampled prior to resuming normal water use in the affected buildings. Scenario 4: lead in tap water in multiple buildings A research study uncovers evidence of lead in tap water within multiple buildings in a community on four occasions over the span of 1 year, with additional findings from a previous study indicating lead was present in the buildings years earlier. Neither study detected lead in source water, at truck outlets, or in building cisterns, indicating that the lead at the tap was associated with the domestic plumbing systems. Potential control measures include treatment to reduce corrosion, replacement of lead-bearing plumbing components and point-of-use lead removal. None of these barriers are in place, and communities are not regulated to monitor lead in tap water. Scenario 5: elevated concentrations of trihalomethanes in water delivery truck Opportunistic sampling by GN-Health indicates that the trihalomethane (THM) level in delivery trucks after 20 min of contact time is above Health Canada’s recommended health-based limit. It is unknown how often THMs are elevated or what the underlying causes are, though it is well established that THMs form when chlorine interacts with natural organic matter (NOM), and there are no removal barriers in place. Scenario 6: interrupted water delivery service At least twice every year, trucks are unable to deliver water due to severe weather. There is some indication that inadequate truck maintenance is a contributing factor, but no records are kept of the disruptions, nor are truck repairs, so this relationship is difficult to confirm. Applicability of existing water safety risk matrices to Nunavut The 12 matrices evaluated by Lane & Hrudey differ in how they define likelihood and consequence but can be broadly categorised as shown in . Notably, absent from the likelihood definition criteria is the presence of a barrier in place to address a given hazard, thereby reducing the future likelihood, a concept previously explored by Walker . World Health Organization The WHO WSP risk matrix uses numeric likelihood and the consequence scores with criteria grounded in compliance with local regulations and health outcomes. This method was developed to be broadly applicable to low-, middle-, or high-income countries, which makes it flexible enough to apply to underserved areas like Nunavut imbedded in a high-income country like Canada. The WHO WSP risk matrix, like all the other matrices, also fails to explicitly account for the water system’s ability to remove contaminants. The consequence scores are not clearly connected to health risks except at the highest consequence and do not differentiate between acute and chronic issues. Alberta A potential strength of the Alberta water safety matrix is the use of a non-linear scoring system that amplifies scores for more serious issues. However, it was difficult to observe this benefit in this analysis due to the reliance on subjective future probability when assigning likelihood scores. Alberta’s system provides no information about how to determine this probability leading to inflation of all the scenario scores used in this paper because the hazards selected had all been previously observed. Alberta’s consequence score definitions are more comprehensive than the WHO by including operational interruptions and by linking compliance and health outcomes, which while recognising a broader range of hazards may lead to confusion when selecting a score. This system includes a “0 – Not Applicable” score, making it easier to develop templates that can be applied across various communities with different water systems within a jurisdiction. In terms of comparability, as a Canadian province, Alberta shares some water safety policies with Nunavut. For example, in both jurisdictions, the current and proposed water quality regulations are informed by the GCDWQ . However, despite being in the same country, Alberta and Nunavut have different governance and regulatory structures. Ireland Ireland’s WSP risk matrix explicitly includes both past occurrences and probability of future reoccurrence in the likelihood score definition, therefore it has the same probability issues as Alberta except when a hazard has only happened once in the past in which case a score reduction can reasonably be applied based on the authors’ interpretation of the scoring definition. The frequency component of the likelihood score is not linked to time intervals, only hazard occurrence. The consequence scores are linked to water quality regulations and differentiate between different types of non-compliance. Like Alberta, the consequence scores include multiple potential criteria but with less precise definitions. This jurisdiction is the least similar to Nunavut based on climate and governance. Australia Australia’s WSP risk matrix has vague future probability criteria for likelihood, and therefore similar weaknesses to Alberta’s, but without time intervals. Australia’s scoring system uses letters for the likelihood scores while preserving numerical consequence scores; each letter–number combination corresponds to a risk level. The main strength of the Australian approach, as presented in Lane and Hrudey, is that the consequence score is directly linked to tangible numeric criteria, such as operating costs and impacted community size . In terms of comparability to Nunavut, although the climates are vastly different, both Australia and Canada are former British colonies currently wrestling with decolonisation with governance structures reflecting this history. Additionally, truck and cistern water systems are widely used in rural and remote Australian communities. South Africa South Africa’s risk system focuses solely on water quality where there is an aesthetic concern or the potential for illness. The non-linear scoring system magnifies hazards with higher risk scores, similar to Alberta’s approach. However when the highest consequence score, defined as ‘death expected from exposure,’ is combined with the lowest likelihood score, the hazard is classified as low risk. In Nunavut, a hazard with a potential consequence of this magnitude would require immediate action to mitigate the risk, based on how water safety policies have historically been applied. This highlights the need to consider both consequence and likelihood scoring criteria in the context of local policy and risk tolerance. Evaluation of risk scoring systems When individual WSP scoring systems were used to score the hazard scenarios, there is low agreement between scoring systems, and low discernment of hazards within a given system, as shown in . The scenarios were developed to represent different risks that would elicit a different magnitude of response based on existing water safety policies in Nunavut. For example, water quality issues with the potential to cause microbial illness outbreak ( E. coli ) should score highest. Only Alberta and Australia achieve the highest risk level in their systems, “High” and “Very High” respective, for this hazardous event, whereas the WHO and South Africa score this as low. The WHO and Ireland systems had the best hazard discernment, but both failed to categorise the presence of E. coli into the highest risk category and therefore do not adequately differentiate between qualitatively different risks. For example, in the Ireland scoring system, both the E. coli and manganese scenarios are categorised as high risk in and , despite E. coli having the potential to cause a major public health emergency, while manganese is primarily an aesthetic concern. After the Alberta, Australia and Ireland scoring rubrics were modified to adhere to the Cox criteria in , there was some improved discernment in the Alberta and Australia hazard risk levels but not enough to adequately differentiate qualitatively different risks. In Approach 3 the likelihood and consequence scores from Approaches 1 and 2 were isolated and converted to score steps 1–5 as shown in and . This revealed that the disagreement between scoring systems can be most attributed to the likelihood scoring criteria. Each scenario was based on the premise that the hazard event had occurred in the past with no indication of the introduction of a barrier to prevent reoccurrence. Therefore, any matrices that considered only future reoccurrence for the likelihood score, such as Alberta and Australia, immediately warranted the highest likelihood score being assigned to the hazard. This accounts for why all resultant risk scores were high or very high in their respective risk-level frameworks, prior to applying the Cox criteria. Conversely, a likelihood score where a frequency of once every 5 years, such as the WHO and South Africa, scores the lowest point allotment even if the consequence receives the highest score. This indicates that past frequency nor future probability alone for likelihood will be appropriate for Nunavut. There is improved scoring alignment between scoring system when looking at consequence score intervals. The notable deviations are: Australia distinguishes between the size of the impacted population South Africa does not consider operational hazards Alberta considers the duration of a service interruption Development of a Nunavut-specific water safety risk framework Based on the above analysis of existing frameworks and their limitations in effectively assessing hazards in remote Arctic communities, we developed a scoring system that adapted these approaches to better suit this context while leveraging their strengths. Additionally, we introduced a new scoring consideration to account for the presence or absence of barriers within a system that reduces risk. Proposed scoring system Our proposed risk matrix includes three scores: the past frequency score, the system readiness score and the consequence score. One important aspect of the proposed system is that it breaks the likelihood score used in other WSP risk matrices into the past frequency score and the system readiness score. The first is grounded in the analysis of past data, while the second is determined based on the presence or absence of infrastructure barriers and monitoring for different potential water safety hazards. These criteria were selected to reduce speculation and bias in scoring and to address criticisms of the WSP systems using future likelihood criteria . Past frequency score The past frequency score is determined by reviewing historical water quality and operational records to determine whether a water safety hazard event has occurred in the past. In some cases, it may be necessary to infer the past frequency of a hazardous water safety event based on external data (e.g. research studies) and/or records from other communities. The score is numeric and as defined in uses time intervals in line with the likelihood scores in the WHO’s WSP Manual Supplementary Tool . Five years was chosen as the upper limit of consideration because very little information exists for most systems in Nunavut before 2019. In the future, this limit could potentially be changed to 20 years, the expected lifetime of a water treatment plant (WTP). Note that in many cases it will be difficult or impossible to assign a past frequency score because of limited past data collection and reporting. This will undoubtedly lead to an underestimation of the past frequency, and thus the potential water safety risk, of important hazardous events. System readiness score Previous research has highlighted the importance of considering the barriers present to manage potential water safety hazards, even if these have not occurred before. We propose adding a system readiness score to Nunavut’s water safety matrix using the criteria identified in to assess the current system’s capacity to address hazardous events that, without proper barriers, could expose users to identified hazards. This score will be informed primarily by design documents, system assessments and operational management documents. Like the WHO, Alberta and Ireland approaches, the system readiness score is also informed by compliance with design and monitoring requirements in the territory as identified by the DWSF . Consequence score The consequence score is informed by the WHO, Alberta and Ireland approaches that assess the consequence of a hazard based on compliance with water quality regulations and recognised potential health outcomes. Our approach, as described in , distinguishes between aesthetic and health-based parameters as well as between preventative health-based requirements, which are indirect indicators of conditions that could give rise to a water safety hazard, and reactive health-based requirements, which indicate the confirmed presence of the hazard. Risk matrix, risk prioritisation and colour coding WSP risk matrices have been identified as having subjective and inconsistent descriptions for the frequency of events and anticipated consequences . As noted, the current matrices are not suitable for performing risk assessments on hazards in Nunavut drinking water systems. Therefore, we developed the following approach. Ranges for different risk levels were chosen such that the resulting risk matrix would be in alignment with the Cox criteria as described by Lane and Hrudey . This resulted in the risk matrix in , which conforms to the Cox Criteria of consistency (no low-risk cell shares a boundary with a high-risk cell, no high-risk cells in the left column or top row), betweenness (an intermediate risk cell exists between low- and high-risk cells), and colour coding (numeric values for different risk levels do not intersect with one another). Assigning past frequency, system readiness and consequence scores for each scenario summarises the past frequency, system readiness and consequence scores assigned to each scenario by the authors. Justifications for each choice are described in the subsections that follow. Scenario 1 Low FC residual is a common hazard that occurs regularly. It has been assigned a past frequency score of 3 (has occurred once a month) and a readiness score of 3 because although the system is in line with current regulations, the monitoring is inadequate, and the system is not designed to meet best practices as outlined in the DWSF. Low FC residual is indicative of conditions that could give rise to the reinfection of treated drinking water by microorganisms, but it is an indicator rather than a direct hazard to human health, so it has been assigned a consequence score of 3 (health – preventative). Scenario 2 Manganese is detected in tap water for a few weeks each year due to seasonal source water quality changes. This was assigned a past frequency score of 2 (occurs at least once a year), a system readiness score of 4 (monitoring in place, no barriers) and a consequence score of 2 (aesthetic impact lasting longer than 1 week). One of the challenges of assigning a consequence score in this case is that if people are concerned about the aesthetic quality of their water, they may switch to untreated sources and/or adopt less healthy practices. A consequence score of 2 may not adequately capture these secondary impacts. Scenario 3 This hazard was assigned a past frequency score of 1 (has occurred at least once in the past 5 years), a system readiness score of 5 (no barriers, no monitoring) and a consequence score of 5 (health – reactive). In this case, the once-a-year sampling is in line with existing voluntary practices in some communities but is not required in territorial regulations and is not adequate to characterise the frequency or extent of the hazard. As a result, the validity of the assigned past frequency score is unclear. Scenario 4 Lead is a serious health hazard for children and foetuses with multiple chronic health impacts, this scenario was assigned a consequence score of 4 (health-based hazard, chronic). No barriers are in place and lead is not regularly monitored at the tap, so this scenario was assigned a system readiness score of 5. It was also assigned a past frequency score of 2 based on the results of past research studies . There are some caveats to this evaluation. The data used to determine past frequency are not from an accredited laboratory and was collected sporadically using multiple different sampling methodologies, which may affect the quality of lead sampling results. Although there are no lead pipes in the community, the presence of lead-bearing components in older the plumbing systems of older buildings suggests a localised but significant risk. This variability between older and newer buildings means the overall risk score may be inflated in newer constructions. Targeted sampling could be used to better understand the extent and frequency of the hazard in the community, helping to inform effective mitigation strategies to reduce exposure. Scenario 5 Opportunistic sampling by territorial regulators indicates that THM levels are above Health Canada’s recommended health-based limit (100 ug/L) in the trucks after 20 min of contact time. However, Health Canada’s limit is meant to be applied to a quarterly sampling average . We assigned this scenario a past frequency score of 1 (has occurred at least once in the past 5 years), a system readiness score of 5 (no barriers in place, no monitoring) and a consequence score of 4 (health based – chronic). The challenge in this scenario is understanding the frequency of the problem and, to a lesser extent, the validity of the health-based maximum acceptable concentration (MAC) recommended by Health Canada, which impacts the consequence score. There is limited evidence that THMs themselves are carcinogenic, though they may be indicative of conditions favourable to the development of more dangerous disinfection byproducts . Thus, benchmarking against the MAC inflates the water safety risk associated with this parameter, potentially resulting in NOM removal being prioritised over other, more dangerous hazards. Scenario 6 Interrupted water delivery due to a winter storm was assigned a past frequency score of 2 (occurs at least once a year), a system readiness score of 5 (no barriers in place) and a consequence score of 3 (health – preventative). The hazardous event in Scenario 6 does not map easily onto the system readiness score or the consequence score, both of which are biased towards measurable water quality parameters, but they do provide a starting point for a high-level assessment of the risk associated with this hazard. Sensitivity analysis of alternative likelihood score equations A key insight from our review is that using the past frequency of a hazard as a proxy for its likelihood of recurrence has both advantages and disadvantages. Using existing data and observations as points of reference helps to ground the likelihood score in real information, reducing the possibility of speculation and human bias during the scoring process. On the other hand, many of the communities that would benefit most from WSPs do not have high-quality operating and water quality records, inadvertently introducing a bias against hazards that have not been observed in the water system in the past. This can result in incorrectly assigning low risk scores to hazards that occur infrequently but have catastrophic consequences . An alternative is to focus instead on the presence or absence of barriers that control the hazard , which we have termed system readiness in this study. In reality, both past frequency and system readiness are important factors that provide explicit and implicit information about the likelihood that users will be exposed to a hazard. We propose a new likelihood score that combines past frequency (PF) and system readiness (SR): Likelihood = x Past Frequency PF + 1 − x System Readiness SR Where x is the fraction assigned to weight each factor. A sensitivity analysis was conducted to explore the impact of the factor weights on the risk score and risk level for the six scenarios developed in this study. The five potential risk equations (REs) used in the sensitivity analysis are summarised in . shows the impact of varying the PF score from 0 to 1 for each of the six scenarios. In five of the six scenarios in this study, the highest risk scores were achieved when the likelihood was only a function of system readiness. The risk associated with one hazard, low chlorine residual, does not appear to be sensitive to the weight assigned to the two components of the likelihood score, but this is an artefact related to a scenario (low chlorine residual in trucks) that was assigned the same score for past frequency (3) and system readiness (3). We recommend that weights be assigned such that such catastrophic events, especially when associated with parts of the water system where only limited data are available, be ranked as high risk. An example of this is Scenario 3 where E. coli is detected in water storage cisterns. The two options that meet this criterium are weights of 0 and 0.25 for past frequency. Past frequency can provide useful information about hazards that are already present in a community and leverage important sources of information such as consulting reports and peer reviewed studies. Weights of 0.25 for past frequency and 0.75 for system readiness acknowledge the utility of past records and experiences while emphasising the importance of system readiness when it comes to water safety risk management. Depending on their needs, stakeholders in Nunavut may choose to discard past frequency score or to use it separately to prioritise hazards that require immediate attention in a given community based on past data collection. Tracking water safety improvements using the water safety risk matrix Scenarios 1, 2, 4 and 5 were chosen for this analysis: Three scenarios representing hazards that fall clearly under the jurisdiction of the municipality and/or the GN (scenarios 1, 2 and 5) and one that demonstrates how incremental improvements in system management can improve water safety at the tap (Scenario 4). In the analysis below, years correspond to the approximate amount of time required to implement new practices and/or infrastructure. We assumed that improved monitoring could be implemented immediately (Year 1) and that actionable results would be available in Year 2. Operational changes could be implemented in Year 2 (no additional resources required) or Year 3 (additional resources required) while infrastructure changes requiring design and construction work could be implemented in Year 4. An additional year of monitoring has been included after each change before confirming that the system is operating according to best practices and changing the system readiness score to ensure that the intervention has had the desired effect. Scenario 1: low chlorine residual in water delivery truck In this scenario, we assumed that upon being informed of the elevated water safety risk associated with FC residual levels in the trucks, in Year 1 the municipality immediately brings their compliance sampling schedule in line with current territorial requirements, reducing their system readiness score from 3 to 2 and their risk score from 9 to 6.75. The risk associated with this hazard remains in the “moderate” category. With increased data collection comes increased awareness of the importance of maintaining FC residuals, and in the following year, there is only one instance of low FC residual in the truck, reducing the past frequency score to from 3 (at least once a month) to 2 (at least once per year), and their total risk score to 6. In Year 3, there are no instances of low chlorine residuals in the truck, and the past frequency score decreases further to 1 (at least once in the past 5 years) with the total risk score decreasing to 5.25. The community works with the GN-CGS in Year 4 to implement best practices to improve water treatment to remove the turbidity, NOM and metals that can contribute to high chlorine demand (e.g. adding coagulation ahead of filtration). After a year of monitoring to ensure that the system upgrades have achieved the desired results, the community’s system readiness score decreases to 1 (all best practices have been implemented) and their total risk score falls to 3, which is in the low-risk category. From this point forward, the community has all the barriers in place to manage this risk but should continue to monitor FC residuals in the truck to meet territorial requirements and to ensure that the barriers remain effective. This thought experiment demonstrates how increased routine monitoring can improve compliance and identify areas of potential long-term improvement to water infrastructure. Scenario 2: manganese in tap water In Scenario 2, seasonally elevated manganese in the surface water supply combined with a lack of manganese removal at the WTP has resulted in annual incidences of high manganese levels at the tap and a risk level of “moderate” or “high” for this hazard depending on whether past frequency is (RE2) or is not (RE1) considered in the calculation. Past records show that sampling has been sporadic, and the samples that have been taken have shown that the high manganese levels fall between the federal guideline AO (20 μg/L) and the MAC (120 μg/L). In Year 1 after the risk assessment, the community implements more targeted sampling and determines that manganese is present at some taps at levels in excess of the MAC. Manganese at levels above 120 μg/L has been linked to chronic health issues , so this new information results in an increase in the consequence score from 2 (aesthetic), to 4 (chronic health issue) and a risk level of “high”. With this knowledge, the community is able to successfully lobby for an upgrade to their WTP to remove seasonal manganese, improving their system readiness score to 2 in Year 4 (total risk score of 7, moderate) and 1 in Year 5 (total risk score of 4, low) after follow-up sampling has demonstrated the effectiveness of the manganese removal process. This shows that improved monitoring of an aesthetic contaminant uncovered a more serious water safety hazard, resulting in a temporarily increased water safety risk score. This more detailed water quality information drives the design and implementation of an effective barrier, ultimately resulting in a low-risk score for this hazard in this community. Scenario 3: E. coli detected in cisterns Scenario 3 highlights the complexity of some water hazards in the decentralised trucked water systems in remote Arctic communities and the importance of having buy-in from all the stakeholders who are impacted by and have influence over the water safety hazard in question. The territorial government does not currently have a regulatory mechanism to enforce more effective barriers (e.g. improved cleaning procedures, improved tank model selection). The stakeholders responsible for building design, and therefore selection of cistern models, and building owners responsible for cistern cleaning could implement improvements in the short term (e.g. new cleaning procedures), the medium term (e.g. point of entry/point of use treatment) and the long term (e.g. adopting best practices for cistern selection and installation). Additionally, the territorial government, recognising the serious and poorly characterised water safety risk associated with cisterns, could work towards the development and implementation of new policies and procedures for monitoring and improving water safety in cisterns. The decentralised nature of truck and cistern systems and the large number of potential stakeholders involved in cistern management make this a complex, long-term task, but one worth pursuing to reduce the safety risk that Nunavummiut are exposed to every day. Scenario 4: lead in tap water in multiple buildings Depending on where lead originates in the system, potential control measures for lead in tap water include flushing the taps, the implementation of processes at the WTP to reduce corrosion (pH control, alkalinity, addition of corrosion inhibitors), replacement of lead-bearing plumbing components in domestic plumbing systems and point of use treatment to remove lead that has entered the tap water . A well-designed lead monitoring programme can help to identify the source(s) of lead in a water system and to determine the most appropriate intervention(s) to minimise lead exposure. The potential impacts of a series of short-term and longer-term interventions to reduce lead at the tap on the calculated risk score are illustrated in . In this hypothetical scenario, a lead monitoring plan is implemented in Year 1 to better understand the extent and frequency of lead exposure in the community that includes targeted weekly sampling. Implementing the monitoring programme lowers the community’s system readiness score from 5 (no barriers, no monitoring) to 4 (no barriers, monitoring in place). The results of the sampling determine that lead is present at least once a week, increasing the past frequency score from 1 (has happened in the past 5 years) to 4 (has happened at least once a week). Assuming that RE2 is used to calculate the risk score, the changes in Year 1 result in a net decrease in the risk score from 17 to 16. In Year 2, the community continues to monitor for lead and institutes regular flushing in buildings that have been identified as having lead issues, reducing the system readiness score to 2. The flushing reduces the frequency of high lead readings from once a week to once a month, reducing the past frequency score to 3. In Year 3, the community undertakes a programme to remove lead bearing components from the plumbing systems in affected buildings, immediately reducing the system readiness score to 2 (barriers exist and are monitored) and reducing the past frequency to once in the past year (2), resulting in a risk score of 8 and a risk level of “moderate”. After a year of monitoring, the interventions appear to have eliminated the sources of lead at the tap and the past frequency and system readiness scores both fall to 1, resulting in a risk level of “low” for both Year 4 and Year 5. When past frequency is not included in the calculation (RE1), Scenario 4 would be assigned a risk level of very high in Year 0, whereas when past frequency makes up 25% of the likelihood score, the scenario is assigned a risk level of high in Year 0. The inclusion of past frequency in the risk score calculation has a smaller impact on the risk levels assigned in later years when the potential interventions are phased in. For example, the risk level is moderate after flushing is implemented in Year 2 irrespective of whether past frequency is included in the likelihood score. Scenario 5: trihalomethanes detected above recommended limits in truck In Year 1 quarterly monitoring is implemented to better understand the frequency of high THMs in the truck. This increased monitoring indicates that THMs are present at least four times a year, increasing the past frequency score from 1 to 2 and increasing the risk score from 16 to 17 (RE2) or keeping it steady at 20 (RE1). Improving NOM removal will require a substantial upgrade to the water treatment system (e.g. adding coagulation, oxidation, or adsorption), which is a long-term endeavour. The alternative is to reduce chlorine addition, but this cannot be done without compromising microbial water safety. In this thought experiment, we have assumed that monitoring continues in Year 2 and Year 3, while a coagulant dosing and mixing system is designed and built at the WTP. The upgrade is completed in Year 4 (risk score then drops to 8, moderate) and the results of the upgrade are monitored and confirmed in Year 5, leading to a final risk score of 4 and a “low” risk level. This scenario demonstrates that some hazards will require interventions that will take multiple years and substantial funding to implement. Scenario 6: Interrupted water delivery Scenario 6 was assigned a risk level of moderate, based on a past frequency of once a year (2), limited system readiness (5) and a consequence score of (3), reflecting a hazard that is indicative of potential health impacts. Improving these scores will be challenging, but some potential interventions include monitoring and truck maintenance, and prioritising road clearing after storms to facilitate water delivery. Furthermore, communities could formalise an emergency preparedness plan for extreme weather events that includes coordinated water conservation communication and prioritisation of water delivery above other municipal activities, while it remains safe to do so. System users can minimise the impact of short-term water shortages by conserving water. Larger tanks could be installed in newly constructed buildings or, in the long term, trucks could be replaced with piped systems. Ultimately, weather is beyond the control of any stakeholder, and many potential improvements will require long-term commitment from multiple parties. A preliminary list of hazards was compiled in Table S.2. It includes common hazards for surface water supply systems, not necessarily confirmed in Nunavut but may not have been adequately assessed for elimination, as well as hazards reported in previous academic studies, consultant reports and government records. For example, while the available literature does not confirm the presence of Cryptosporidium or Giardia in Nunavut’s licenced potable water sources , these pathogens are commonly found in surface water. As a precaution, the GN has started incorporating treatment barriers specifically designed to protect against these pathogens in newly constructed water treatment facilities . Conversely, incidences of high manganese in treated water and chlorine residuals below the regulated minimum in treated water distribution trucks have been established within literature and confirmed by GN records. Government oversight, and therefore records, is limited to source water, treatment and distribution, so hazards present in buildings, such as bacteria in cisterns and lead in tap water have been identified solely from academic studies. Beyond quality-focused hazards, service interruption hazards that impact delivery to consumers are recognised in Table S2, as well. Many of the identified hazards are relevant to all Nunavut communities; however, engagement with water professionals and system users in individual communities will be required to confirm the list and subsequently the likelihood and consequence scores of each hazard. The list is expected to change with input from community stakeholders and territorial authorities but can be used to facilitate initial discussions and raise awareness and build capacity and knowledge within the groups. The remedial actions identified in Table S.2 are measures currently in place or actions that could be taken by a community in the interim to improve water safety, prior to completing an upgrade to the existing infrastructure. The six hypothetical scenarios below were developed from the hazards listed in Table S.2 but are not directly based upon a specific water system in Nunavut. Rather, the scenarios were used to test the different risk matrix systems. Scenario 1: low chlorine residual in water delivery truck A community water system consists of a surface water supply, an intake pipe and pump, chlorination, three trucks and 120 individual building water cisterns. Regulations dictate that 20 min of contact time with a minimum outlet FC residual of 0.2 mg/L is required to achieve disinfection. At least once a month, the water exiting one or more of the trucks has FC less than 0.2 mg/L after contact time. Municipal truck drivers are required to measure and record FC levels daily, but they do so only once a month, resulting in incomplete records. Scenario 2: manganese in water supply Every spring surface water turns over causing increased levels of manganese to enter the treatment system which does not have a removal process. Manganese remains in the water are delivered to community buildings, causing discolouration and staining, reducing consumer faith in their drinking water. Water is sampled once a year, but manganese does not have an aesthetic limit in the existing regulations. Scenario 3: E. coli in domestic water storage cisterns Opportunistic sampling by GN-Health in a community with a surface water supply, chlorination and trucked distribution discovers E. coli in the cisterns of three public buildings, confirmed through follow-up samples. The cisterns are sampled approximately once a year by the GN Health officer or voluntarily by the building owner, and this is the first time that E. coli has been detected. GN-Health issues a boil water advisory (BWA) and investigates the contamination, which is eventually attributed to inconsistent cistern cleaning, which is in turn linked to the installation of cistern models that are difficult to clean. Because it is determined that the contamination is localised, the BWA is removed from the community. The cisterns are thoroughly cleaned and resampled prior to resuming normal water use in the affected buildings. Scenario 4: lead in tap water in multiple buildings A research study uncovers evidence of lead in tap water within multiple buildings in a community on four occasions over the span of 1 year, with additional findings from a previous study indicating lead was present in the buildings years earlier. Neither study detected lead in source water, at truck outlets, or in building cisterns, indicating that the lead at the tap was associated with the domestic plumbing systems. Potential control measures include treatment to reduce corrosion, replacement of lead-bearing plumbing components and point-of-use lead removal. None of these barriers are in place, and communities are not regulated to monitor lead in tap water. Scenario 5: elevated concentrations of trihalomethanes in water delivery truck Opportunistic sampling by GN-Health indicates that the trihalomethane (THM) level in delivery trucks after 20 min of contact time is above Health Canada’s recommended health-based limit. It is unknown how often THMs are elevated or what the underlying causes are, though it is well established that THMs form when chlorine interacts with natural organic matter (NOM), and there are no removal barriers in place. Scenario 6: interrupted water delivery service At least twice every year, trucks are unable to deliver water due to severe weather. There is some indication that inadequate truck maintenance is a contributing factor, but no records are kept of the disruptions, nor are truck repairs, so this relationship is difficult to confirm. A community water system consists of a surface water supply, an intake pipe and pump, chlorination, three trucks and 120 individual building water cisterns. Regulations dictate that 20 min of contact time with a minimum outlet FC residual of 0.2 mg/L is required to achieve disinfection. At least once a month, the water exiting one or more of the trucks has FC less than 0.2 mg/L after contact time. Municipal truck drivers are required to measure and record FC levels daily, but they do so only once a month, resulting in incomplete records. Every spring surface water turns over causing increased levels of manganese to enter the treatment system which does not have a removal process. Manganese remains in the water are delivered to community buildings, causing discolouration and staining, reducing consumer faith in their drinking water. Water is sampled once a year, but manganese does not have an aesthetic limit in the existing regulations. Opportunistic sampling by GN-Health in a community with a surface water supply, chlorination and trucked distribution discovers E. coli in the cisterns of three public buildings, confirmed through follow-up samples. The cisterns are sampled approximately once a year by the GN Health officer or voluntarily by the building owner, and this is the first time that E. coli has been detected. GN-Health issues a boil water advisory (BWA) and investigates the contamination, which is eventually attributed to inconsistent cistern cleaning, which is in turn linked to the installation of cistern models that are difficult to clean. Because it is determined that the contamination is localised, the BWA is removed from the community. The cisterns are thoroughly cleaned and resampled prior to resuming normal water use in the affected buildings. A research study uncovers evidence of lead in tap water within multiple buildings in a community on four occasions over the span of 1 year, with additional findings from a previous study indicating lead was present in the buildings years earlier. Neither study detected lead in source water, at truck outlets, or in building cisterns, indicating that the lead at the tap was associated with the domestic plumbing systems. Potential control measures include treatment to reduce corrosion, replacement of lead-bearing plumbing components and point-of-use lead removal. None of these barriers are in place, and communities are not regulated to monitor lead in tap water. Opportunistic sampling by GN-Health indicates that the trihalomethane (THM) level in delivery trucks after 20 min of contact time is above Health Canada’s recommended health-based limit. It is unknown how often THMs are elevated or what the underlying causes are, though it is well established that THMs form when chlorine interacts with natural organic matter (NOM), and there are no removal barriers in place. At least twice every year, trucks are unable to deliver water due to severe weather. There is some indication that inadequate truck maintenance is a contributing factor, but no records are kept of the disruptions, nor are truck repairs, so this relationship is difficult to confirm. The 12 matrices evaluated by Lane & Hrudey differ in how they define likelihood and consequence but can be broadly categorised as shown in . Notably, absent from the likelihood definition criteria is the presence of a barrier in place to address a given hazard, thereby reducing the future likelihood, a concept previously explored by Walker . The WHO WSP risk matrix uses numeric likelihood and the consequence scores with criteria grounded in compliance with local regulations and health outcomes. This method was developed to be broadly applicable to low-, middle-, or high-income countries, which makes it flexible enough to apply to underserved areas like Nunavut imbedded in a high-income country like Canada. The WHO WSP risk matrix, like all the other matrices, also fails to explicitly account for the water system’s ability to remove contaminants. The consequence scores are not clearly connected to health risks except at the highest consequence and do not differentiate between acute and chronic issues. A potential strength of the Alberta water safety matrix is the use of a non-linear scoring system that amplifies scores for more serious issues. However, it was difficult to observe this benefit in this analysis due to the reliance on subjective future probability when assigning likelihood scores. Alberta’s system provides no information about how to determine this probability leading to inflation of all the scenario scores used in this paper because the hazards selected had all been previously observed. Alberta’s consequence score definitions are more comprehensive than the WHO by including operational interruptions and by linking compliance and health outcomes, which while recognising a broader range of hazards may lead to confusion when selecting a score. This system includes a “0 – Not Applicable” score, making it easier to develop templates that can be applied across various communities with different water systems within a jurisdiction. In terms of comparability, as a Canadian province, Alberta shares some water safety policies with Nunavut. For example, in both jurisdictions, the current and proposed water quality regulations are informed by the GCDWQ . However, despite being in the same country, Alberta and Nunavut have different governance and regulatory structures. Ireland’s WSP risk matrix explicitly includes both past occurrences and probability of future reoccurrence in the likelihood score definition, therefore it has the same probability issues as Alberta except when a hazard has only happened once in the past in which case a score reduction can reasonably be applied based on the authors’ interpretation of the scoring definition. The frequency component of the likelihood score is not linked to time intervals, only hazard occurrence. The consequence scores are linked to water quality regulations and differentiate between different types of non-compliance. Like Alberta, the consequence scores include multiple potential criteria but with less precise definitions. This jurisdiction is the least similar to Nunavut based on climate and governance. Australia’s WSP risk matrix has vague future probability criteria for likelihood, and therefore similar weaknesses to Alberta’s, but without time intervals. Australia’s scoring system uses letters for the likelihood scores while preserving numerical consequence scores; each letter–number combination corresponds to a risk level. The main strength of the Australian approach, as presented in Lane and Hrudey, is that the consequence score is directly linked to tangible numeric criteria, such as operating costs and impacted community size . In terms of comparability to Nunavut, although the climates are vastly different, both Australia and Canada are former British colonies currently wrestling with decolonisation with governance structures reflecting this history. Additionally, truck and cistern water systems are widely used in rural and remote Australian communities. South Africa’s risk system focuses solely on water quality where there is an aesthetic concern or the potential for illness. The non-linear scoring system magnifies hazards with higher risk scores, similar to Alberta’s approach. However when the highest consequence score, defined as ‘death expected from exposure,’ is combined with the lowest likelihood score, the hazard is classified as low risk. In Nunavut, a hazard with a potential consequence of this magnitude would require immediate action to mitigate the risk, based on how water safety policies have historically been applied. This highlights the need to consider both consequence and likelihood scoring criteria in the context of local policy and risk tolerance. When individual WSP scoring systems were used to score the hazard scenarios, there is low agreement between scoring systems, and low discernment of hazards within a given system, as shown in . The scenarios were developed to represent different risks that would elicit a different magnitude of response based on existing water safety policies in Nunavut. For example, water quality issues with the potential to cause microbial illness outbreak ( E. coli ) should score highest. Only Alberta and Australia achieve the highest risk level in their systems, “High” and “Very High” respective, for this hazardous event, whereas the WHO and South Africa score this as low. The WHO and Ireland systems had the best hazard discernment, but both failed to categorise the presence of E. coli into the highest risk category and therefore do not adequately differentiate between qualitatively different risks. For example, in the Ireland scoring system, both the E. coli and manganese scenarios are categorised as high risk in and , despite E. coli having the potential to cause a major public health emergency, while manganese is primarily an aesthetic concern. After the Alberta, Australia and Ireland scoring rubrics were modified to adhere to the Cox criteria in , there was some improved discernment in the Alberta and Australia hazard risk levels but not enough to adequately differentiate qualitatively different risks. In Approach 3 the likelihood and consequence scores from Approaches 1 and 2 were isolated and converted to score steps 1–5 as shown in and . This revealed that the disagreement between scoring systems can be most attributed to the likelihood scoring criteria. Each scenario was based on the premise that the hazard event had occurred in the past with no indication of the introduction of a barrier to prevent reoccurrence. Therefore, any matrices that considered only future reoccurrence for the likelihood score, such as Alberta and Australia, immediately warranted the highest likelihood score being assigned to the hazard. This accounts for why all resultant risk scores were high or very high in their respective risk-level frameworks, prior to applying the Cox criteria. Conversely, a likelihood score where a frequency of once every 5 years, such as the WHO and South Africa, scores the lowest point allotment even if the consequence receives the highest score. This indicates that past frequency nor future probability alone for likelihood will be appropriate for Nunavut. There is improved scoring alignment between scoring system when looking at consequence score intervals. The notable deviations are: Australia distinguishes between the size of the impacted population South Africa does not consider operational hazards Alberta considers the duration of a service interruption Based on the above analysis of existing frameworks and their limitations in effectively assessing hazards in remote Arctic communities, we developed a scoring system that adapted these approaches to better suit this context while leveraging their strengths. Additionally, we introduced a new scoring consideration to account for the presence or absence of barriers within a system that reduces risk. Proposed scoring system Our proposed risk matrix includes three scores: the past frequency score, the system readiness score and the consequence score. One important aspect of the proposed system is that it breaks the likelihood score used in other WSP risk matrices into the past frequency score and the system readiness score. The first is grounded in the analysis of past data, while the second is determined based on the presence or absence of infrastructure barriers and monitoring for different potential water safety hazards. These criteria were selected to reduce speculation and bias in scoring and to address criticisms of the WSP systems using future likelihood criteria . Past frequency score The past frequency score is determined by reviewing historical water quality and operational records to determine whether a water safety hazard event has occurred in the past. In some cases, it may be necessary to infer the past frequency of a hazardous water safety event based on external data (e.g. research studies) and/or records from other communities. The score is numeric and as defined in uses time intervals in line with the likelihood scores in the WHO’s WSP Manual Supplementary Tool . Five years was chosen as the upper limit of consideration because very little information exists for most systems in Nunavut before 2019. In the future, this limit could potentially be changed to 20 years, the expected lifetime of a water treatment plant (WTP). Note that in many cases it will be difficult or impossible to assign a past frequency score because of limited past data collection and reporting. This will undoubtedly lead to an underestimation of the past frequency, and thus the potential water safety risk, of important hazardous events. System readiness score Previous research has highlighted the importance of considering the barriers present to manage potential water safety hazards, even if these have not occurred before. We propose adding a system readiness score to Nunavut’s water safety matrix using the criteria identified in to assess the current system’s capacity to address hazardous events that, without proper barriers, could expose users to identified hazards. This score will be informed primarily by design documents, system assessments and operational management documents. Like the WHO, Alberta and Ireland approaches, the system readiness score is also informed by compliance with design and monitoring requirements in the territory as identified by the DWSF . Consequence score The consequence score is informed by the WHO, Alberta and Ireland approaches that assess the consequence of a hazard based on compliance with water quality regulations and recognised potential health outcomes. Our approach, as described in , distinguishes between aesthetic and health-based parameters as well as between preventative health-based requirements, which are indirect indicators of conditions that could give rise to a water safety hazard, and reactive health-based requirements, which indicate the confirmed presence of the hazard. Our proposed risk matrix includes three scores: the past frequency score, the system readiness score and the consequence score. One important aspect of the proposed system is that it breaks the likelihood score used in other WSP risk matrices into the past frequency score and the system readiness score. The first is grounded in the analysis of past data, while the second is determined based on the presence or absence of infrastructure barriers and monitoring for different potential water safety hazards. These criteria were selected to reduce speculation and bias in scoring and to address criticisms of the WSP systems using future likelihood criteria . The past frequency score is determined by reviewing historical water quality and operational records to determine whether a water safety hazard event has occurred in the past. In some cases, it may be necessary to infer the past frequency of a hazardous water safety event based on external data (e.g. research studies) and/or records from other communities. The score is numeric and as defined in uses time intervals in line with the likelihood scores in the WHO’s WSP Manual Supplementary Tool . Five years was chosen as the upper limit of consideration because very little information exists for most systems in Nunavut before 2019. In the future, this limit could potentially be changed to 20 years, the expected lifetime of a water treatment plant (WTP). Note that in many cases it will be difficult or impossible to assign a past frequency score because of limited past data collection and reporting. This will undoubtedly lead to an underestimation of the past frequency, and thus the potential water safety risk, of important hazardous events. Previous research has highlighted the importance of considering the barriers present to manage potential water safety hazards, even if these have not occurred before. We propose adding a system readiness score to Nunavut’s water safety matrix using the criteria identified in to assess the current system’s capacity to address hazardous events that, without proper barriers, could expose users to identified hazards. This score will be informed primarily by design documents, system assessments and operational management documents. Like the WHO, Alberta and Ireland approaches, the system readiness score is also informed by compliance with design and monitoring requirements in the territory as identified by the DWSF . The consequence score is informed by the WHO, Alberta and Ireland approaches that assess the consequence of a hazard based on compliance with water quality regulations and recognised potential health outcomes. Our approach, as described in , distinguishes between aesthetic and health-based parameters as well as between preventative health-based requirements, which are indirect indicators of conditions that could give rise to a water safety hazard, and reactive health-based requirements, which indicate the confirmed presence of the hazard. WSP risk matrices have been identified as having subjective and inconsistent descriptions for the frequency of events and anticipated consequences . As noted, the current matrices are not suitable for performing risk assessments on hazards in Nunavut drinking water systems. Therefore, we developed the following approach. Ranges for different risk levels were chosen such that the resulting risk matrix would be in alignment with the Cox criteria as described by Lane and Hrudey . This resulted in the risk matrix in , which conforms to the Cox Criteria of consistency (no low-risk cell shares a boundary with a high-risk cell, no high-risk cells in the left column or top row), betweenness (an intermediate risk cell exists between low- and high-risk cells), and colour coding (numeric values for different risk levels do not intersect with one another). summarises the past frequency, system readiness and consequence scores assigned to each scenario by the authors. Justifications for each choice are described in the subsections that follow. Scenario 1 Low FC residual is a common hazard that occurs regularly. It has been assigned a past frequency score of 3 (has occurred once a month) and a readiness score of 3 because although the system is in line with current regulations, the monitoring is inadequate, and the system is not designed to meet best practices as outlined in the DWSF. Low FC residual is indicative of conditions that could give rise to the reinfection of treated drinking water by microorganisms, but it is an indicator rather than a direct hazard to human health, so it has been assigned a consequence score of 3 (health – preventative). Scenario 2 Manganese is detected in tap water for a few weeks each year due to seasonal source water quality changes. This was assigned a past frequency score of 2 (occurs at least once a year), a system readiness score of 4 (monitoring in place, no barriers) and a consequence score of 2 (aesthetic impact lasting longer than 1 week). One of the challenges of assigning a consequence score in this case is that if people are concerned about the aesthetic quality of their water, they may switch to untreated sources and/or adopt less healthy practices. A consequence score of 2 may not adequately capture these secondary impacts. Scenario 3 This hazard was assigned a past frequency score of 1 (has occurred at least once in the past 5 years), a system readiness score of 5 (no barriers, no monitoring) and a consequence score of 5 (health – reactive). In this case, the once-a-year sampling is in line with existing voluntary practices in some communities but is not required in territorial regulations and is not adequate to characterise the frequency or extent of the hazard. As a result, the validity of the assigned past frequency score is unclear. Scenario 4 Lead is a serious health hazard for children and foetuses with multiple chronic health impacts, this scenario was assigned a consequence score of 4 (health-based hazard, chronic). No barriers are in place and lead is not regularly monitored at the tap, so this scenario was assigned a system readiness score of 5. It was also assigned a past frequency score of 2 based on the results of past research studies . There are some caveats to this evaluation. The data used to determine past frequency are not from an accredited laboratory and was collected sporadically using multiple different sampling methodologies, which may affect the quality of lead sampling results. Although there are no lead pipes in the community, the presence of lead-bearing components in older the plumbing systems of older buildings suggests a localised but significant risk. This variability between older and newer buildings means the overall risk score may be inflated in newer constructions. Targeted sampling could be used to better understand the extent and frequency of the hazard in the community, helping to inform effective mitigation strategies to reduce exposure. Scenario 5 Opportunistic sampling by territorial regulators indicates that THM levels are above Health Canada’s recommended health-based limit (100 ug/L) in the trucks after 20 min of contact time. However, Health Canada’s limit is meant to be applied to a quarterly sampling average . We assigned this scenario a past frequency score of 1 (has occurred at least once in the past 5 years), a system readiness score of 5 (no barriers in place, no monitoring) and a consequence score of 4 (health based – chronic). The challenge in this scenario is understanding the frequency of the problem and, to a lesser extent, the validity of the health-based maximum acceptable concentration (MAC) recommended by Health Canada, which impacts the consequence score. There is limited evidence that THMs themselves are carcinogenic, though they may be indicative of conditions favourable to the development of more dangerous disinfection byproducts . Thus, benchmarking against the MAC inflates the water safety risk associated with this parameter, potentially resulting in NOM removal being prioritised over other, more dangerous hazards. Scenario 6 Interrupted water delivery due to a winter storm was assigned a past frequency score of 2 (occurs at least once a year), a system readiness score of 5 (no barriers in place) and a consequence score of 3 (health – preventative). The hazardous event in Scenario 6 does not map easily onto the system readiness score or the consequence score, both of which are biased towards measurable water quality parameters, but they do provide a starting point for a high-level assessment of the risk associated with this hazard. Low FC residual is a common hazard that occurs regularly. It has been assigned a past frequency score of 3 (has occurred once a month) and a readiness score of 3 because although the system is in line with current regulations, the monitoring is inadequate, and the system is not designed to meet best practices as outlined in the DWSF. Low FC residual is indicative of conditions that could give rise to the reinfection of treated drinking water by microorganisms, but it is an indicator rather than a direct hazard to human health, so it has been assigned a consequence score of 3 (health – preventative). Manganese is detected in tap water for a few weeks each year due to seasonal source water quality changes. This was assigned a past frequency score of 2 (occurs at least once a year), a system readiness score of 4 (monitoring in place, no barriers) and a consequence score of 2 (aesthetic impact lasting longer than 1 week). One of the challenges of assigning a consequence score in this case is that if people are concerned about the aesthetic quality of their water, they may switch to untreated sources and/or adopt less healthy practices. A consequence score of 2 may not adequately capture these secondary impacts. This hazard was assigned a past frequency score of 1 (has occurred at least once in the past 5 years), a system readiness score of 5 (no barriers, no monitoring) and a consequence score of 5 (health – reactive). In this case, the once-a-year sampling is in line with existing voluntary practices in some communities but is not required in territorial regulations and is not adequate to characterise the frequency or extent of the hazard. As a result, the validity of the assigned past frequency score is unclear. Lead is a serious health hazard for children and foetuses with multiple chronic health impacts, this scenario was assigned a consequence score of 4 (health-based hazard, chronic). No barriers are in place and lead is not regularly monitored at the tap, so this scenario was assigned a system readiness score of 5. It was also assigned a past frequency score of 2 based on the results of past research studies . There are some caveats to this evaluation. The data used to determine past frequency are not from an accredited laboratory and was collected sporadically using multiple different sampling methodologies, which may affect the quality of lead sampling results. Although there are no lead pipes in the community, the presence of lead-bearing components in older the plumbing systems of older buildings suggests a localised but significant risk. This variability between older and newer buildings means the overall risk score may be inflated in newer constructions. Targeted sampling could be used to better understand the extent and frequency of the hazard in the community, helping to inform effective mitigation strategies to reduce exposure. Opportunistic sampling by territorial regulators indicates that THM levels are above Health Canada’s recommended health-based limit (100 ug/L) in the trucks after 20 min of contact time. However, Health Canada’s limit is meant to be applied to a quarterly sampling average . We assigned this scenario a past frequency score of 1 (has occurred at least once in the past 5 years), a system readiness score of 5 (no barriers in place, no monitoring) and a consequence score of 4 (health based – chronic). The challenge in this scenario is understanding the frequency of the problem and, to a lesser extent, the validity of the health-based maximum acceptable concentration (MAC) recommended by Health Canada, which impacts the consequence score. There is limited evidence that THMs themselves are carcinogenic, though they may be indicative of conditions favourable to the development of more dangerous disinfection byproducts . Thus, benchmarking against the MAC inflates the water safety risk associated with this parameter, potentially resulting in NOM removal being prioritised over other, more dangerous hazards. Interrupted water delivery due to a winter storm was assigned a past frequency score of 2 (occurs at least once a year), a system readiness score of 5 (no barriers in place) and a consequence score of 3 (health – preventative). The hazardous event in Scenario 6 does not map easily onto the system readiness score or the consequence score, both of which are biased towards measurable water quality parameters, but they do provide a starting point for a high-level assessment of the risk associated with this hazard. A key insight from our review is that using the past frequency of a hazard as a proxy for its likelihood of recurrence has both advantages and disadvantages. Using existing data and observations as points of reference helps to ground the likelihood score in real information, reducing the possibility of speculation and human bias during the scoring process. On the other hand, many of the communities that would benefit most from WSPs do not have high-quality operating and water quality records, inadvertently introducing a bias against hazards that have not been observed in the water system in the past. This can result in incorrectly assigning low risk scores to hazards that occur infrequently but have catastrophic consequences . An alternative is to focus instead on the presence or absence of barriers that control the hazard , which we have termed system readiness in this study. In reality, both past frequency and system readiness are important factors that provide explicit and implicit information about the likelihood that users will be exposed to a hazard. We propose a new likelihood score that combines past frequency (PF) and system readiness (SR): Likelihood = x Past Frequency PF + 1 − x System Readiness SR Where x is the fraction assigned to weight each factor. A sensitivity analysis was conducted to explore the impact of the factor weights on the risk score and risk level for the six scenarios developed in this study. The five potential risk equations (REs) used in the sensitivity analysis are summarised in . shows the impact of varying the PF score from 0 to 1 for each of the six scenarios. In five of the six scenarios in this study, the highest risk scores were achieved when the likelihood was only a function of system readiness. The risk associated with one hazard, low chlorine residual, does not appear to be sensitive to the weight assigned to the two components of the likelihood score, but this is an artefact related to a scenario (low chlorine residual in trucks) that was assigned the same score for past frequency (3) and system readiness (3). We recommend that weights be assigned such that such catastrophic events, especially when associated with parts of the water system where only limited data are available, be ranked as high risk. An example of this is Scenario 3 where E. coli is detected in water storage cisterns. The two options that meet this criterium are weights of 0 and 0.25 for past frequency. Past frequency can provide useful information about hazards that are already present in a community and leverage important sources of information such as consulting reports and peer reviewed studies. Weights of 0.25 for past frequency and 0.75 for system readiness acknowledge the utility of past records and experiences while emphasising the importance of system readiness when it comes to water safety risk management. Depending on their needs, stakeholders in Nunavut may choose to discard past frequency score or to use it separately to prioritise hazards that require immediate attention in a given community based on past data collection. Scenarios 1, 2, 4 and 5 were chosen for this analysis: Three scenarios representing hazards that fall clearly under the jurisdiction of the municipality and/or the GN (scenarios 1, 2 and 5) and one that demonstrates how incremental improvements in system management can improve water safety at the tap (Scenario 4). In the analysis below, years correspond to the approximate amount of time required to implement new practices and/or infrastructure. We assumed that improved monitoring could be implemented immediately (Year 1) and that actionable results would be available in Year 2. Operational changes could be implemented in Year 2 (no additional resources required) or Year 3 (additional resources required) while infrastructure changes requiring design and construction work could be implemented in Year 4. An additional year of monitoring has been included after each change before confirming that the system is operating according to best practices and changing the system readiness score to ensure that the intervention has had the desired effect. Scenario 1: low chlorine residual in water delivery truck In this scenario, we assumed that upon being informed of the elevated water safety risk associated with FC residual levels in the trucks, in Year 1 the municipality immediately brings their compliance sampling schedule in line with current territorial requirements, reducing their system readiness score from 3 to 2 and their risk score from 9 to 6.75. The risk associated with this hazard remains in the “moderate” category. With increased data collection comes increased awareness of the importance of maintaining FC residuals, and in the following year, there is only one instance of low FC residual in the truck, reducing the past frequency score to from 3 (at least once a month) to 2 (at least once per year), and their total risk score to 6. In Year 3, there are no instances of low chlorine residuals in the truck, and the past frequency score decreases further to 1 (at least once in the past 5 years) with the total risk score decreasing to 5.25. The community works with the GN-CGS in Year 4 to implement best practices to improve water treatment to remove the turbidity, NOM and metals that can contribute to high chlorine demand (e.g. adding coagulation ahead of filtration). After a year of monitoring to ensure that the system upgrades have achieved the desired results, the community’s system readiness score decreases to 1 (all best practices have been implemented) and their total risk score falls to 3, which is in the low-risk category. From this point forward, the community has all the barriers in place to manage this risk but should continue to monitor FC residuals in the truck to meet territorial requirements and to ensure that the barriers remain effective. This thought experiment demonstrates how increased routine monitoring can improve compliance and identify areas of potential long-term improvement to water infrastructure. Scenario 2: manganese in tap water In Scenario 2, seasonally elevated manganese in the surface water supply combined with a lack of manganese removal at the WTP has resulted in annual incidences of high manganese levels at the tap and a risk level of “moderate” or “high” for this hazard depending on whether past frequency is (RE2) or is not (RE1) considered in the calculation. Past records show that sampling has been sporadic, and the samples that have been taken have shown that the high manganese levels fall between the federal guideline AO (20 μg/L) and the MAC (120 μg/L). In Year 1 after the risk assessment, the community implements more targeted sampling and determines that manganese is present at some taps at levels in excess of the MAC. Manganese at levels above 120 μg/L has been linked to chronic health issues , so this new information results in an increase in the consequence score from 2 (aesthetic), to 4 (chronic health issue) and a risk level of “high”. With this knowledge, the community is able to successfully lobby for an upgrade to their WTP to remove seasonal manganese, improving their system readiness score to 2 in Year 4 (total risk score of 7, moderate) and 1 in Year 5 (total risk score of 4, low) after follow-up sampling has demonstrated the effectiveness of the manganese removal process. This shows that improved monitoring of an aesthetic contaminant uncovered a more serious water safety hazard, resulting in a temporarily increased water safety risk score. This more detailed water quality information drives the design and implementation of an effective barrier, ultimately resulting in a low-risk score for this hazard in this community. Scenario 3: E. coli detected in cisterns Scenario 3 highlights the complexity of some water hazards in the decentralised trucked water systems in remote Arctic communities and the importance of having buy-in from all the stakeholders who are impacted by and have influence over the water safety hazard in question. The territorial government does not currently have a regulatory mechanism to enforce more effective barriers (e.g. improved cleaning procedures, improved tank model selection). The stakeholders responsible for building design, and therefore selection of cistern models, and building owners responsible for cistern cleaning could implement improvements in the short term (e.g. new cleaning procedures), the medium term (e.g. point of entry/point of use treatment) and the long term (e.g. adopting best practices for cistern selection and installation). Additionally, the territorial government, recognising the serious and poorly characterised water safety risk associated with cisterns, could work towards the development and implementation of new policies and procedures for monitoring and improving water safety in cisterns. The decentralised nature of truck and cistern systems and the large number of potential stakeholders involved in cistern management make this a complex, long-term task, but one worth pursuing to reduce the safety risk that Nunavummiut are exposed to every day. Scenario 4: lead in tap water in multiple buildings Depending on where lead originates in the system, potential control measures for lead in tap water include flushing the taps, the implementation of processes at the WTP to reduce corrosion (pH control, alkalinity, addition of corrosion inhibitors), replacement of lead-bearing plumbing components in domestic plumbing systems and point of use treatment to remove lead that has entered the tap water . A well-designed lead monitoring programme can help to identify the source(s) of lead in a water system and to determine the most appropriate intervention(s) to minimise lead exposure. The potential impacts of a series of short-term and longer-term interventions to reduce lead at the tap on the calculated risk score are illustrated in . In this hypothetical scenario, a lead monitoring plan is implemented in Year 1 to better understand the extent and frequency of lead exposure in the community that includes targeted weekly sampling. Implementing the monitoring programme lowers the community’s system readiness score from 5 (no barriers, no monitoring) to 4 (no barriers, monitoring in place). The results of the sampling determine that lead is present at least once a week, increasing the past frequency score from 1 (has happened in the past 5 years) to 4 (has happened at least once a week). Assuming that RE2 is used to calculate the risk score, the changes in Year 1 result in a net decrease in the risk score from 17 to 16. In Year 2, the community continues to monitor for lead and institutes regular flushing in buildings that have been identified as having lead issues, reducing the system readiness score to 2. The flushing reduces the frequency of high lead readings from once a week to once a month, reducing the past frequency score to 3. In Year 3, the community undertakes a programme to remove lead bearing components from the plumbing systems in affected buildings, immediately reducing the system readiness score to 2 (barriers exist and are monitored) and reducing the past frequency to once in the past year (2), resulting in a risk score of 8 and a risk level of “moderate”. After a year of monitoring, the interventions appear to have eliminated the sources of lead at the tap and the past frequency and system readiness scores both fall to 1, resulting in a risk level of “low” for both Year 4 and Year 5. When past frequency is not included in the calculation (RE1), Scenario 4 would be assigned a risk level of very high in Year 0, whereas when past frequency makes up 25% of the likelihood score, the scenario is assigned a risk level of high in Year 0. The inclusion of past frequency in the risk score calculation has a smaller impact on the risk levels assigned in later years when the potential interventions are phased in. For example, the risk level is moderate after flushing is implemented in Year 2 irrespective of whether past frequency is included in the likelihood score. Scenario 5: trihalomethanes detected above recommended limits in truck In Year 1 quarterly monitoring is implemented to better understand the frequency of high THMs in the truck. This increased monitoring indicates that THMs are present at least four times a year, increasing the past frequency score from 1 to 2 and increasing the risk score from 16 to 17 (RE2) or keeping it steady at 20 (RE1). Improving NOM removal will require a substantial upgrade to the water treatment system (e.g. adding coagulation, oxidation, or adsorption), which is a long-term endeavour. The alternative is to reduce chlorine addition, but this cannot be done without compromising microbial water safety. In this thought experiment, we have assumed that monitoring continues in Year 2 and Year 3, while a coagulant dosing and mixing system is designed and built at the WTP. The upgrade is completed in Year 4 (risk score then drops to 8, moderate) and the results of the upgrade are monitored and confirmed in Year 5, leading to a final risk score of 4 and a “low” risk level. This scenario demonstrates that some hazards will require interventions that will take multiple years and substantial funding to implement. Scenario 6: Interrupted water delivery Scenario 6 was assigned a risk level of moderate, based on a past frequency of once a year (2), limited system readiness (5) and a consequence score of (3), reflecting a hazard that is indicative of potential health impacts. Improving these scores will be challenging, but some potential interventions include monitoring and truck maintenance, and prioritising road clearing after storms to facilitate water delivery. Furthermore, communities could formalise an emergency preparedness plan for extreme weather events that includes coordinated water conservation communication and prioritisation of water delivery above other municipal activities, while it remains safe to do so. System users can minimise the impact of short-term water shortages by conserving water. Larger tanks could be installed in newly constructed buildings or, in the long term, trucks could be replaced with piped systems. Ultimately, weather is beyond the control of any stakeholder, and many potential improvements will require long-term commitment from multiple parties. In this scenario, we assumed that upon being informed of the elevated water safety risk associated with FC residual levels in the trucks, in Year 1 the municipality immediately brings their compliance sampling schedule in line with current territorial requirements, reducing their system readiness score from 3 to 2 and their risk score from 9 to 6.75. The risk associated with this hazard remains in the “moderate” category. With increased data collection comes increased awareness of the importance of maintaining FC residuals, and in the following year, there is only one instance of low FC residual in the truck, reducing the past frequency score to from 3 (at least once a month) to 2 (at least once per year), and their total risk score to 6. In Year 3, there are no instances of low chlorine residuals in the truck, and the past frequency score decreases further to 1 (at least once in the past 5 years) with the total risk score decreasing to 5.25. The community works with the GN-CGS in Year 4 to implement best practices to improve water treatment to remove the turbidity, NOM and metals that can contribute to high chlorine demand (e.g. adding coagulation ahead of filtration). After a year of monitoring to ensure that the system upgrades have achieved the desired results, the community’s system readiness score decreases to 1 (all best practices have been implemented) and their total risk score falls to 3, which is in the low-risk category. From this point forward, the community has all the barriers in place to manage this risk but should continue to monitor FC residuals in the truck to meet territorial requirements and to ensure that the barriers remain effective. This thought experiment demonstrates how increased routine monitoring can improve compliance and identify areas of potential long-term improvement to water infrastructure. In Scenario 2, seasonally elevated manganese in the surface water supply combined with a lack of manganese removal at the WTP has resulted in annual incidences of high manganese levels at the tap and a risk level of “moderate” or “high” for this hazard depending on whether past frequency is (RE2) or is not (RE1) considered in the calculation. Past records show that sampling has been sporadic, and the samples that have been taken have shown that the high manganese levels fall between the federal guideline AO (20 μg/L) and the MAC (120 μg/L). In Year 1 after the risk assessment, the community implements more targeted sampling and determines that manganese is present at some taps at levels in excess of the MAC. Manganese at levels above 120 μg/L has been linked to chronic health issues , so this new information results in an increase in the consequence score from 2 (aesthetic), to 4 (chronic health issue) and a risk level of “high”. With this knowledge, the community is able to successfully lobby for an upgrade to their WTP to remove seasonal manganese, improving their system readiness score to 2 in Year 4 (total risk score of 7, moderate) and 1 in Year 5 (total risk score of 4, low) after follow-up sampling has demonstrated the effectiveness of the manganese removal process. This shows that improved monitoring of an aesthetic contaminant uncovered a more serious water safety hazard, resulting in a temporarily increased water safety risk score. This more detailed water quality information drives the design and implementation of an effective barrier, ultimately resulting in a low-risk score for this hazard in this community. Scenario 3 highlights the complexity of some water hazards in the decentralised trucked water systems in remote Arctic communities and the importance of having buy-in from all the stakeholders who are impacted by and have influence over the water safety hazard in question. The territorial government does not currently have a regulatory mechanism to enforce more effective barriers (e.g. improved cleaning procedures, improved tank model selection). The stakeholders responsible for building design, and therefore selection of cistern models, and building owners responsible for cistern cleaning could implement improvements in the short term (e.g. new cleaning procedures), the medium term (e.g. point of entry/point of use treatment) and the long term (e.g. adopting best practices for cistern selection and installation). Additionally, the territorial government, recognising the serious and poorly characterised water safety risk associated with cisterns, could work towards the development and implementation of new policies and procedures for monitoring and improving water safety in cisterns. The decentralised nature of truck and cistern systems and the large number of potential stakeholders involved in cistern management make this a complex, long-term task, but one worth pursuing to reduce the safety risk that Nunavummiut are exposed to every day. Depending on where lead originates in the system, potential control measures for lead in tap water include flushing the taps, the implementation of processes at the WTP to reduce corrosion (pH control, alkalinity, addition of corrosion inhibitors), replacement of lead-bearing plumbing components in domestic plumbing systems and point of use treatment to remove lead that has entered the tap water . A well-designed lead monitoring programme can help to identify the source(s) of lead in a water system and to determine the most appropriate intervention(s) to minimise lead exposure. The potential impacts of a series of short-term and longer-term interventions to reduce lead at the tap on the calculated risk score are illustrated in . In this hypothetical scenario, a lead monitoring plan is implemented in Year 1 to better understand the extent and frequency of lead exposure in the community that includes targeted weekly sampling. Implementing the monitoring programme lowers the community’s system readiness score from 5 (no barriers, no monitoring) to 4 (no barriers, monitoring in place). The results of the sampling determine that lead is present at least once a week, increasing the past frequency score from 1 (has happened in the past 5 years) to 4 (has happened at least once a week). Assuming that RE2 is used to calculate the risk score, the changes in Year 1 result in a net decrease in the risk score from 17 to 16. In Year 2, the community continues to monitor for lead and institutes regular flushing in buildings that have been identified as having lead issues, reducing the system readiness score to 2. The flushing reduces the frequency of high lead readings from once a week to once a month, reducing the past frequency score to 3. In Year 3, the community undertakes a programme to remove lead bearing components from the plumbing systems in affected buildings, immediately reducing the system readiness score to 2 (barriers exist and are monitored) and reducing the past frequency to once in the past year (2), resulting in a risk score of 8 and a risk level of “moderate”. After a year of monitoring, the interventions appear to have eliminated the sources of lead at the tap and the past frequency and system readiness scores both fall to 1, resulting in a risk level of “low” for both Year 4 and Year 5. When past frequency is not included in the calculation (RE1), Scenario 4 would be assigned a risk level of very high in Year 0, whereas when past frequency makes up 25% of the likelihood score, the scenario is assigned a risk level of high in Year 0. The inclusion of past frequency in the risk score calculation has a smaller impact on the risk levels assigned in later years when the potential interventions are phased in. For example, the risk level is moderate after flushing is implemented in Year 2 irrespective of whether past frequency is included in the likelihood score. In Year 1 quarterly monitoring is implemented to better understand the frequency of high THMs in the truck. This increased monitoring indicates that THMs are present at least four times a year, increasing the past frequency score from 1 to 2 and increasing the risk score from 16 to 17 (RE2) or keeping it steady at 20 (RE1). Improving NOM removal will require a substantial upgrade to the water treatment system (e.g. adding coagulation, oxidation, or adsorption), which is a long-term endeavour. The alternative is to reduce chlorine addition, but this cannot be done without compromising microbial water safety. In this thought experiment, we have assumed that monitoring continues in Year 2 and Year 3, while a coagulant dosing and mixing system is designed and built at the WTP. The upgrade is completed in Year 4 (risk score then drops to 8, moderate) and the results of the upgrade are monitored and confirmed in Year 5, leading to a final risk score of 4 and a “low” risk level. This scenario demonstrates that some hazards will require interventions that will take multiple years and substantial funding to implement. Scenario 6 was assigned a risk level of moderate, based on a past frequency of once a year (2), limited system readiness (5) and a consequence score of (3), reflecting a hazard that is indicative of potential health impacts. Improving these scores will be challenging, but some potential interventions include monitoring and truck maintenance, and prioritising road clearing after storms to facilitate water delivery. Furthermore, communities could formalise an emergency preparedness plan for extreme weather events that includes coordinated water conservation communication and prioritisation of water delivery above other municipal activities, while it remains safe to do so. System users can minimise the impact of short-term water shortages by conserving water. Larger tanks could be installed in newly constructed buildings or, in the long term, trucks could be replaced with piped systems. Ultimately, weather is beyond the control of any stakeholder, and many potential improvements will require long-term commitment from multiple parties. Form WSP Tteam Assemble a team of representatives from territorial and local governments, operational staff, system users and potentially consultants, researchers and federal government representatives. Identify hazards through document review Gather and analyse design documents, operational records, water quality data and other relevant information to identify potential hazards that should be considered. Tour water system to confirm and identify hazards Physically inspect the water system from source to tap to gather additional information on hazards and identify any new ones. Score hazards according to matrix criteria Apply the risk matrix to evaluate each identified hazard, scoring them based on the past frequency, system readiness and consequence of the hazard. Assign a risk level to each hazard score Use the risk scores to categorise hazards into different risk levels (e.g. low, medium, high, very high). Identify interventions to reduce hazard risk Based on the risk assessment, prioritise and recommend short- and long-term interventions to address identified hazards, considering resource availability. Implement interventions Implement the identified interventions to mitigate the risks associated with the hazards. Monitor and evaluate interventions Continuously monitor the effectiveness of the interventions and adjust as necessary. This cyclical process is meant to be iterative, allowing for continuous improvement as new data and circumstances arise. The steps will be guided by government policies and local priorities, requiring collaboration among stakeholders at the municipal, territorial and federal levels. Specific recommendations are beyond the scope of this process, as they depend on the unique characteristics of each community and system. Strengths and limitations of proposed approach The proposed approach, like other WSP matrices from different jurisdictions, has strengths and weaknesses. Key strengths include the use of past data to reduce speculation, consideration of system readiness and barriers, and recognition of a multi-barrier approach. The system readiness score is based on regulation-informed criteria and best practices, while the consequence score reflects water quality standards and potential health impacts. This encourages evidence-based decision-making and supports technological interventions, continuous improvement and the ability to numerically track risk reduction over time. Weaknesses largely stem from current conditions in the territory that are expected to improve over time. Limited or incomplete historical water quality data, operational records and design documents may bias analysis. As water treatment and record-keeping improve, these issues should be resolved. However, hazards, like cisterns in buildings, remain outside the current regulatory framework. The WSP paradigm’s limitations, particularly when using a quantitative risk lens, also present challenges. Assigning scores for indirect factors such as staff recruitment, training, funding and asset management is difficult. Additionally, the approach does not account for the extent of a given hazard or future climate change impacts. Future improvements to the framework could include: Conducting case studies to assess its application in Nunavut communities Collecting more water quality and operational data to understand hazards and system readiness Making the template more dynamic for planning and tracking interventions Incorporating factors like funding, labour, remoteness and building design to better understand risk drivers Engaging more stakeholders to identify and prioritise community-specific hazards Grounding consequence scores in a more quantitative system, such as quantitative microbial risk assessment Assemble a team of representatives from territorial and local governments, operational staff, system users and potentially consultants, researchers and federal government representatives. Gather and analyse design documents, operational records, water quality data and other relevant information to identify potential hazards that should be considered. Physically inspect the water system from source to tap to gather additional information on hazards and identify any new ones. Apply the risk matrix to evaluate each identified hazard, scoring them based on the past frequency, system readiness and consequence of the hazard. Use the risk scores to categorise hazards into different risk levels (e.g. low, medium, high, very high). Based on the risk assessment, prioritise and recommend short- and long-term interventions to address identified hazards, considering resource availability. Implement the identified interventions to mitigate the risks associated with the hazards. Continuously monitor the effectiveness of the interventions and adjust as necessary. This cyclical process is meant to be iterative, allowing for continuous improvement as new data and circumstances arise. The steps will be guided by government policies and local priorities, requiring collaboration among stakeholders at the municipal, territorial and federal levels. Specific recommendations are beyond the scope of this process, as they depend on the unique characteristics of each community and system. The proposed approach, like other WSP matrices from different jurisdictions, has strengths and weaknesses. Key strengths include the use of past data to reduce speculation, consideration of system readiness and barriers, and recognition of a multi-barrier approach. The system readiness score is based on regulation-informed criteria and best practices, while the consequence score reflects water quality standards and potential health impacts. This encourages evidence-based decision-making and supports technological interventions, continuous improvement and the ability to numerically track risk reduction over time. Weaknesses largely stem from current conditions in the territory that are expected to improve over time. Limited or incomplete historical water quality data, operational records and design documents may bias analysis. As water treatment and record-keeping improve, these issues should be resolved. However, hazards, like cisterns in buildings, remain outside the current regulatory framework. The WSP paradigm’s limitations, particularly when using a quantitative risk lens, also present challenges. Assigning scores for indirect factors such as staff recruitment, training, funding and asset management is difficult. Additionally, the approach does not account for the extent of a given hazard or future climate change impacts. Future improvements to the framework could include: Conducting case studies to assess its application in Nunavut communities Collecting more water quality and operational data to understand hazards and system readiness Making the template more dynamic for planning and tracking interventions Incorporating factors like funding, labour, remoteness and building design to better understand risk drivers Engaging more stakeholders to identify and prioritise community-specific hazards Grounding consequence scores in a more quantitative system, such as quantitative microbial risk assessment A series of WSP matrices were evaluated and deconstructed to assess their suitability for assessing the water safety risk of common or likely hazard scenarios in Nunavut. This exercise revealed where the existing WSP matrices could be adapted to better fit the realities of Nunavut. Retaining past frequency while also recontextualizing future probability in terms of the presence and successful application of barriers in a system allows for a more straightforward assignment of a score to a hazard. Relying entirely on past frequency results in inappropriate risk scores for rare but potentially catastrophic events. Refining the consequence score definitions to distinguish between water quality issues that may cause acute or chronic illness and those that indirectly increase the risk of such illnesses is expected to improve the categorisation of hazards based on their qualitatively different risks. This, in turn, should promote better allocation of resources to reduce water safety risks. Ultimately, the choice of risk matrix will fall to the governing bodies and other stakeholders with direct responsibilities related to drinking water provision from source to tap. Supplementary Information revised.docx |
The neuropsychology of healthy aging: the positive context of the University of the Third Age during the COVID-19 pandemic | 1e1e832e-9363-4305-a85c-76f25283575b | 10115807 | Physiology[mh] | The COVID-19 pandemic had a significant impact on older adults, as they were more vulnerable to both the adverse effects of SARS-CoV-2 contagion and the measures taken by governments to contain the spread of infection . While the infection curve has been effectively contained, restrictive measures have increased social isolation and loneliness , which in turn affects well-being and increases physical frailty, cognitive impairment, and mood swings . Nursing home residents and patients with Alzheimer's disease or other dementias are those most likely to suffer the effects of the COVID-19 pandemic , , placing them at highest risk of a fatal outcome or long COVID-19 impact . A substantial amount of data collected on older adults during the pandemic came mostly from online cross-sectional surveys. For example, self-reported data from an online survey were used to examine the impact of the COVID-19 period on well-being, activity levels, sleep quality, and cognitive function . These studies allowed for large samples to be analyzed, but with notable limitations in terms of the generalizability of the results. First, the fact that not all older adults could be reached via online technologies suggests possible biases in the recruitment of the sample. In addition, the use of self-report questionnaires instead of in-depth neuropsychological assessments made it difficult to clearly classify participants as healthy or cognitively impaired. Finally, the lack of longitudinal studies with neuropsychological data collected before the pandemic made it impossible to detect potential changes in cognitive functioning and/or mood, limiting the resulting findings to short-term effects and neglecting the crucial role of "baseline" performance. Indeed, in assessing the multifaceted consequences of the COVID-19 pandemic, it is important to consider that cognitive decline may be part of normal aging (e.g. – ), in addition to a gradual decline in physical abilities that may limit functional abilities of daily living and quality of life (e.g. ). However, little is known about the impact of the COVID-19 pandemic on healthy aging in individuals who reported not being infected with SARS-CoV-2. Taking advantage of available pre-pandemic measurements, it has been previously reported longitudinal evidence of early neuropsychological changes during the pandemic period in healthy older individuals , . A longitudinal study first examined the role of neuropsychogeriatric factors in lockdown fatigue, by comparing data collected in healthy, cognitively aging, individuals before and during the pandemic. Participants were assessed at three different time points during the pandemic, i.e., during the first lockdown period (T1), immediately afterward (T2), and during the second lockdown period (T3). Results highlighted interacting changes of physical functioning, executive attention, and mood deflactions in the COVID-19 pandemic. Subjects with moderate fatigue reported more depressive and anxious symptoms than subjects with mild fatigue. Cognitive performance in terms of psychomotor speed also appears to play an important role in the perception of fatigue associated with COVID-19 restrictive measures. Specifically, the results of principal component and multiple regression analyses demonstrated the contribution of "cognitive" and "psychological" factors (i.e., attentional and executive performance, as well as mood deflections) in explaining handgrip strength and gait speed as two of five determinants of the Fried frailty model . At T3, lockdown fatigue was explained by higher scores on the Beck Depression Inventory and lower performance on the Trail Making Test Part A. The results of a moderated mediation model showed that the effect of psychomotor speed on lockdown fatigue was mediated by depression, with gait speed having a moderating effect on this relationship . Furthermore, fear of infection could be an additional source of concern for this at-risk population, exacerbating anxiety and thus affecting quality of life . Consistent with this hypothesis, we have previously reported longitudinal data demonstrating that perceived threat associated with the consequences of SARS-CoV-2 infection was predicted by a combination of baseline physical, cognitive, and mood measures, i.e., anxiety and frailty in addition to lower information processing speed and language comprehension performance . To date, however, no longitudinal study has examined whether healthy older people may also have responded to the COVID-19 pandemic with improvements in neuropsychological functioning. This important but under-researched topic could shed light on how the cognitive and functional abilities that enable the well-being of older subjects are maintained during a very difficult time such as a pandemic. Identifying the variables that promote or hinder such maintenance will contribute to the development of programs aimed at preventing cognitive decline in healthy older people, as recommended by the American Geriatrics Society . We aimed to fill this gap by examining neuropsychological measures before and during the pandemic in a group of cognitively healthy older people. Prior to the COVID-19 outbreak, participants had joined the initiatives of an innovative laboratory for active and healthy aging dedicated to training, education, and research at the University of the Third Age (UNITRE) in Turin, Italy. This was the ideal context to identify possible positive responses to the pandemic, in addition to the negative outcomes that have been widely reported (e.g. ). A healthy lifestyle that includes cognitive, social, and physical activities was positively correlated with global cognition , which was achieved through informal learning programs and activities such as the University of the Third Age (U3A) . The combination of positive life experiences such as education and participation in cognitive and socially stimulating leisure activities is thought to increase the effectiveness of cognitive processing in aging individuals, also referred to as cognitive reserve (CR) . It improves cognitive function in healthy aging subjects; in particular, CR has been associated with global cognition and has been reported to enhance executive function and attention , . CR has also been shown to be protective against brain damage and dementia , slowing the cognitive aging process or reducing the risk of psychiatric disorders . Indeed, CR can be altered or improved by cognitive, mental, and physical stimulation activities . Based on these assumptions, participants in our study underwent a thorough neuropsychological assessment before and during the pandemic to identify possible changes in (a) global cognition, memory, language, executive, and attentional functions; (b) physical status; and (c) mood changes. At the final time point (i.e., 21 months after the outbreak of the pandemic), subjects' heart rate variability (HRV), which is an indicator of physiological mechanisms of (dys)regulation , was recorded during the presentation of images depicting the initial and most stressful phase of the pandemic. We predicted that CR would not only help participants maintain good neuropsychological and physical function to cope with the negative experiences of the pandemic, but also play a protective role in pandemic emotional dysregulation. However, given the potential negative impact of the COVID-19 pandemic, we predicted a possible decline in motivation in the form of apathy and increased anxiety also due to fear of SARS-CoV-2 infection. Moreover, the extent of pandemic-related apathy might reflect the combination of anxiety and emotional (dys)regulation. Tables and provide a detailed overview of the characteristics of the 39 participants and their cognitive, affective, and physical status before and during the pandemic. All neuropsychological variables are normally distributed, with no missing data. Regarding the sociodemographic characteristics of the participants, 79% were women and 21% were men, their mean age was 70 years (age range: 62–82), and their mean educational level was 13 years (range: 8–17). They attended middle school (N = 2), high school (N = 28), and college or university (N = 9). Forty-one percent of the sample lived alone and the remaining 59% lived with at least one person (in most cases, their partner). According to the Hollingshead index (HI , ), most of the subjects belonged to the medium–high socioeconomic status (SES). Specifically, 8% of them fell into the highest status (HI range: 66–55), 51% into the second (HI range: 54–40), 36% into the third (HI range: 39–30), 5% into the fourth (HI range: 29–20), and none into the lowest (HI range: 19–18). During the pandemic, only one of the participants had been diagnosed with COVID-19 infection, while the others stated that they had not contracted the virus. Participants reported high compliance with most of the Italian Ministry of Health recommendations to contain infection, i.e., maintaining safe distances and frequent hand washing (100% of participants), using face masks (95%), avoiding crowded places (90%), and wearing latex gloves when outdoors (87%). Regarding immunization against SARS-CoV-2, 95% of participants had been vaccinated against SARS-CoV-2, and 77% had completed two vaccination cycles according to the recommendations of the Italian Ministry of Health. Longitudinal analyses of cognitive functioning Prior to the pandemic, subjects' performance on both the Addenbrooke's Cognitive Examination-Revised Version (ACE-R ) and the Mini Mental State Examination (MMSE ) did not indicate the presence of mild cognitive impairment (MCI, i.e. ), as all subjects scored above cut-off scores and did not report any concerns for subjective cognitive decline (SCD ). Their performance was above the reference cut-off on the following tests: the Montreal Cognitive Assessment (MoCA ) (100%), Rey Memory Test (RMT )-15 instant words (97%), RMT-15 delayed words (95%), Trail-Making Test (TMT )-part A (100%), TMT-part B (100%), TMT B-A (97%), and Token Test (TT (100%). Notably, the percentages of results below the cut-off were consistent with the margin of error of the neuropsychological tests in the normative population (see Table ). During the pandemic, subjects' performance on both ACE-R and the MMSE did not indicate the presence of MCI (i.e. ), as all subjects scored above the regulatory cut-off and did not report SCD , which was also assessed with the Cognitive Function Instrument (CFI ). Their performance was above the reference cut-off on the following tests: MoCA (100%), RMT—15 immediate words (97%), RMT—15 delayed words (97%), TMT—Part A (100%), TMT—Part B (100%), TMT B-A (100%), and TT (97%). Again, although some deficits were found in the neuropsychological tests (see Table ), the percentages of scores below the cut-off were consistent with the margin of error of the tests in the normative population. To test the stability of cognitive functions, Bayes factors (BFs) were calculated by comparing individual performance before and during the COVID-19 pandemic. In particular, we used the BF with the paired-samples t-test to evaluate the relationship between the probability of the data under the null hypothesis and that under the alternative hypothesis – . When comparisons were made with BF > 3 (in terms of significantly equal scores), the results showed no longitudinal changes in memory (RMT—15 immediate words, BF 01 = 4.14; RMT—15 delayed words, BF 01 = 4.33), attention (TMT time scores: part A, BF 01 = 5.46; part B, BF 01 = 3.53; and B-A, BF 01 = 3.15), and global cognition, as measured by MMSE (BF 01 = 4.22). The BF suggests that the data have exactly the same probability of occurrence under the hypothesis that repeated measures are the same, as under the hypothesis that they are different . We found evidence of differential performance with a significant increase in raw test scores across time points in global cognition, as measured by ACE-R (t = −2.952, p = 0.006, α = 0.05), executive functions (MoCA, t = −2.609, p = 0.013, α = 0.05), and comprehension of linguistic utterances (TT, t = −4.70, p = 0.001, α = 0.05). Longitudinal analyses of mood The subjects showed longitudinal changes in mood in the form of increasing apathy and anxiety. Prior to the pandemic, none of the subjects scored below the cut-off on the Apathy Rating Scale (AES ) and the Hamilton Rating Scale for Anxiety (HARS ). Instead, a significant increase in scale scores was observed during the pandemic. Thus, 36% of the sample achieved a below cut-off score on AES and 13% on HARS, respectively. It should be noted that in paired t-tests, there was a significant increase in the mean scores of AES (t = −7.43, p < 0.001, α = 0.05) and HARS (t = −3.71, p < 0.001, α = 0.005), as shown in Fig. , which also shows the improvement in ACE-R task performance. On the other hand, if we consider the comparison with BF > 3 (in terms of significantly equal scores), the results show no longitudinal changes in depression (Beck Depression Inventory, BDI , BF 01 = 5.33), hypomania (Mania Scale, MAS , BF 01 = 5.51) and disinhibition (Disinhibition Scale, DIS , BF 01 = 3.07). Longitudinal analyses of physical functioning As shown in Table , most participants were classified as "robust" (72% and 64% before and during the pandemic, respectively); the rest of the sample had a prefrail status (28% and 36% before and during the pandemic, respectively). None of the participants met Fried et al.'s inclusion criteria for frailty status, either before or during the pandemic. Based on the McNemar hypothesis test, no changes in physical frailty status were observed (X 2 = 2.29, p = 0.515, α = 0.05) across timepoints, as shown in contingency Table . Follow-up: correlation matrix and multiple regression analysis To investigate the relationship between neuropsychological performance at follow-up and emotional (dys)regulation as assessed by the low-frequency/high-frequency (LF/HF) ratio of HRV, we first generated a correlation matrix considering the variables that showed a significant difference before and during the pandemic, based on t-tests and BFs. As shown in Table , significant correlations were found between LF/HF and both AES (p < 0.05) and ACE-R (p < 0.05), but not HARS, MoCA, or TT. AES was selected as the dependent variable of a multiple regression model because it is the measure showing the greatest change across timepoints (t = −7.43, p < 0.001, α = 0.05; Fig. ). In addition to the LF/HF ratio, which reflects emotional (dys)regulation at follow-up, we modeled as predictors the variables that showed significant change longitudinally, i.e., ACE-R and HARS, which capture global cognitive function and mood anxiety. As shown in Table , a strongly significant model (R = 0.711, R 2 = 0.506) indicated that AES was significantly predicted by lower ACE-R scores (t = −2.391, p = 0.024) and by higher HARS scores (t = 3.704, p = 0.001) and higher LF/HF ratio (t = 3.011, p = 0.006), gender (t = 0.621, p = 0.540) and age (t = −0.630, p = 0.534). Prior to the pandemic, subjects' performance on both the Addenbrooke's Cognitive Examination-Revised Version (ACE-R ) and the Mini Mental State Examination (MMSE ) did not indicate the presence of mild cognitive impairment (MCI, i.e. ), as all subjects scored above cut-off scores and did not report any concerns for subjective cognitive decline (SCD ). Their performance was above the reference cut-off on the following tests: the Montreal Cognitive Assessment (MoCA ) (100%), Rey Memory Test (RMT )-15 instant words (97%), RMT-15 delayed words (95%), Trail-Making Test (TMT )-part A (100%), TMT-part B (100%), TMT B-A (97%), and Token Test (TT (100%). Notably, the percentages of results below the cut-off were consistent with the margin of error of the neuropsychological tests in the normative population (see Table ). During the pandemic, subjects' performance on both ACE-R and the MMSE did not indicate the presence of MCI (i.e. ), as all subjects scored above the regulatory cut-off and did not report SCD , which was also assessed with the Cognitive Function Instrument (CFI ). Their performance was above the reference cut-off on the following tests: MoCA (100%), RMT—15 immediate words (97%), RMT—15 delayed words (97%), TMT—Part A (100%), TMT—Part B (100%), TMT B-A (100%), and TT (97%). Again, although some deficits were found in the neuropsychological tests (see Table ), the percentages of scores below the cut-off were consistent with the margin of error of the tests in the normative population. To test the stability of cognitive functions, Bayes factors (BFs) were calculated by comparing individual performance before and during the COVID-19 pandemic. In particular, we used the BF with the paired-samples t-test to evaluate the relationship between the probability of the data under the null hypothesis and that under the alternative hypothesis – . When comparisons were made with BF > 3 (in terms of significantly equal scores), the results showed no longitudinal changes in memory (RMT—15 immediate words, BF 01 = 4.14; RMT—15 delayed words, BF 01 = 4.33), attention (TMT time scores: part A, BF 01 = 5.46; part B, BF 01 = 3.53; and B-A, BF 01 = 3.15), and global cognition, as measured by MMSE (BF 01 = 4.22). The BF suggests that the data have exactly the same probability of occurrence under the hypothesis that repeated measures are the same, as under the hypothesis that they are different . We found evidence of differential performance with a significant increase in raw test scores across time points in global cognition, as measured by ACE-R (t = −2.952, p = 0.006, α = 0.05), executive functions (MoCA, t = −2.609, p = 0.013, α = 0.05), and comprehension of linguistic utterances (TT, t = −4.70, p = 0.001, α = 0.05). The subjects showed longitudinal changes in mood in the form of increasing apathy and anxiety. Prior to the pandemic, none of the subjects scored below the cut-off on the Apathy Rating Scale (AES ) and the Hamilton Rating Scale for Anxiety (HARS ). Instead, a significant increase in scale scores was observed during the pandemic. Thus, 36% of the sample achieved a below cut-off score on AES and 13% on HARS, respectively. It should be noted that in paired t-tests, there was a significant increase in the mean scores of AES (t = −7.43, p < 0.001, α = 0.05) and HARS (t = −3.71, p < 0.001, α = 0.005), as shown in Fig. , which also shows the improvement in ACE-R task performance. On the other hand, if we consider the comparison with BF > 3 (in terms of significantly equal scores), the results show no longitudinal changes in depression (Beck Depression Inventory, BDI , BF 01 = 5.33), hypomania (Mania Scale, MAS , BF 01 = 5.51) and disinhibition (Disinhibition Scale, DIS , BF 01 = 3.07). As shown in Table , most participants were classified as "robust" (72% and 64% before and during the pandemic, respectively); the rest of the sample had a prefrail status (28% and 36% before and during the pandemic, respectively). None of the participants met Fried et al.'s inclusion criteria for frailty status, either before or during the pandemic. Based on the McNemar hypothesis test, no changes in physical frailty status were observed (X 2 = 2.29, p = 0.515, α = 0.05) across timepoints, as shown in contingency Table . To investigate the relationship between neuropsychological performance at follow-up and emotional (dys)regulation as assessed by the low-frequency/high-frequency (LF/HF) ratio of HRV, we first generated a correlation matrix considering the variables that showed a significant difference before and during the pandemic, based on t-tests and BFs. As shown in Table , significant correlations were found between LF/HF and both AES (p < 0.05) and ACE-R (p < 0.05), but not HARS, MoCA, or TT. AES was selected as the dependent variable of a multiple regression model because it is the measure showing the greatest change across timepoints (t = −7.43, p < 0.001, α = 0.05; Fig. ). In addition to the LF/HF ratio, which reflects emotional (dys)regulation at follow-up, we modeled as predictors the variables that showed significant change longitudinally, i.e., ACE-R and HARS, which capture global cognitive function and mood anxiety. As shown in Table , a strongly significant model (R = 0.711, R 2 = 0.506) indicated that AES was significantly predicted by lower ACE-R scores (t = −2.391, p = 0.024) and by higher HARS scores (t = 3.704, p = 0.001) and higher LF/HF ratio (t = 3.011, p = 0.006), gender (t = 0.621, p = 0.540) and age (t = −0.630, p = 0.534). The present study took advantage of the availability of pre-pandemic data on the cognitive, physical, and mental status of a cohort of older people who had participated in the UNITRE Healthy and Active Aging Lab initiatives. The opportunity to directly compare neuropsychological data collected before and during the COVID-19 pandemic, and to investigate possible changes (positive or negative), is unique. Although the results of some surveys have shown that older people responded better to the COVID-19 pandemic than younger people in terms of emotional well-being, lower reactivity to stressors, and better mental health – , our study, however, showed for the first time a positive response of healthy older people to the pandemic in terms of neuropsychological performance at different timepoints. Specifically, the data showed (1) significant increases in raw scores on global cognitive ability (ACE-R), executive function (MoCA), and comprehension of linguistic utterances (TT); (2) stable performance on immediate and delayed memory (RMT), executive attention (TMT parts A and B, in addition to TMT B-A), and global cognitive abilities as measured by the MMSE; and (3) stable physical status as measured by phenotypic frailty determinants. Together with the absence of participants classified as "frail," this pattern of findings underscores the maintenance of good physical performance despite inactivity due to the restrictive lockdown measures. Evidence of stable or even improved measures at different timepoints provides new insight into widely held claims about the negative effects of the pandemic in cognitively preserved older people, based on self-reported observations of deterioration in sleep quality, mental and physical health and functioning, symptoms of depression and apathy, and limitations in social relationships . Although improved performance on attention and working memory tasks has also been reported compared before and during the pandemic measurements, the results of this study were obtained by administering questionnaires adapted for remote administration . Interestingly, there is evidence that cognitive decline may be mitigated by so-called CR, i.e., the mind/brain's resilience to age- or disease-related changes . Specifically, CR has been shown to mitigate the cognitive and psychological effects of physiological aging and, most importantly, its accompanying circumstances such as social isolation and loneliness , which are often cited as major contributors to physical and mental changes during the pandemic . The American Geriatrics Society suggested strategies that could promote healthy aging in the COVID-19 era based on domains that consider health and the promotion of cognitive, physical, and socio-relational aspects . As reported by WHO , U3A identified as an informal learning program and activity worldwide , and prior to the pandemic, showed a positive impact on the psychological well-being and quality of life of the elderly population. Indeed, U3A programs emphasized improvements in physical health status, emotional balance (e.g., decreases in depressive symptoms and negative affect), social support, and coping activities and strategies, thus increasing positive self-perception and sense of control as key effects for the decade of aging . Importantly, our results highlight how cognitive functioning can be stimulated and even improved by appropriate interaction with a positive context during the pandemic, as experienced by UNITRE participants who continued to participate in their distance learning courses and shared common experiences via social media. Our study shows that older people can maintain their cognitive levels if they continue to engage in social and educational activities. Moreover, CR appears to play a protective role against the emotional dysregulation associated with the pandemic. We also observed across different timepoints stable mood scores on the BDI, MAS, and DIS scales, which measure depression, hypomania, and disinhibition, respectively. In contrast, pandemic-related effects appeared to be more associated with mood deflections measured by increased apathy (AES) and anxiety (HARS) during the pandemic. The results at the correlation matrix, regarding the relationship between neuropsychological follow-up data and emotional (dys)regulation during HRV recording, showed significant associations between the LF/HF ratio and both AES and ACE-R, but not with HARS, MoCA, or TT. In particular, apathy was associated with greater emotional (dys)regulation, and ACE-R scores were related to an increase in the LF/HF ratio, while suggestive pandemic images were presented. Because apathy showed the greatest change across timepoints, it was selected as the dependent variable of a multiple regression model. On this basis, higher apathy was predicted by poorer global cognitive performance, increased anxiety, and emotional (dys)regulation as measured by a higher ratio of low to high frequencies of HRV. Thus, preserved global cognitive performance appears to play a protective role against the effects of pandemic-related anxiety and emotional dysregulation on apathy. These findings highlight the importance of CR in attenuating the negative associations between lowered mood and cognition and their joint contribution to well-being, even in challenging situations such as a pandemic. The growing concern triggered by a major health emergency such as the COVID-19 pandemic primarily affects the most vulnerable, for example, the older population with cognitive impairment. In contrast, our study shows for the first time in the literature how some healthy older subjects responded positively to the pandemic emergency despite some mood deflections, showing 'the bright side of the moon ' through our study. Therefore, the UNITRE model is a good way to promote active and healthy aging, especially in the face of complex and negative events such as the COVID-19 pandemic, even taking into account demographic changes. In fact, by 2050, one in five people will be over 60 years old, which is a total of two billion people worldwide . Thus, there is an urgent need to plan and implement evidence-based intervention strategies consistent with our study to promote healthy aging that can overcome the adversity of potential future health emergencies. Therefore, UNITRE programs have the potential to accelerate impact by 2023 on healthy aging. Despite the small sample size, our findings are based on objective measures collected before and during the pandemic through in-depth neuropsychological assessments. This rare opportunity provided new insights into the maintenance of cognitive function and its protective role in relation to changes in mood and emotional (dys)regulation in socially active individuals of moderate to high socioeconomic status. Participants From April to October 2019, about 100 socially active older people from the University of the Third Age (UNITRE) in Turin (Italy) joined the initiatives of an innovative laboratory for active and healthy aging, dedicated to training, education and research and development. Before the outbreak of the pandemic, they participated in teaching modules on cognitive functions, physical activity, nutrition and social integration to promote active and healthy aging. Specifically, they were invited to undergo an in-depth neuropsychological assessment to investigate possible associations between cognitive functioning, mood changes, and physical health status over a 3-year period. A group of 81 volunteers (64 women; 60–80 years) agreed to participate in the study, but pandemic restrictions necessitated the establishment of an online assessment. After the lockdown measures were relaxed, 39 subjects agreed to undergo the final date of neuropsychological assessment between June and October 2021. Thirty-five subjects (27 women; 62–82 years) also agreed to participate in a psychophysiological study from October to December 2021 (i.e., 21 months after the pandemic outbreak) to assess emotional reactivity to images associated with the COVID-19 pandemic. All 39 subjects who participated in the full longitudinal study were > 60 years old and thus can be classified as "older adults" . They were not taking any psychotropic medications that could have affected their cognitive abilities or mood, and none of them complained of cognitive decline . The study was conducted in accordance with the Declaration of Helsinki, having been approved by the Ethics Committee of the University of Turin before (Prot. No. 10038) and during (Prot. No. 151786) the pandemic. All participants gave written informed consent before participating in the study. Sociodemographic assessment In addition to the usual sociodemographic characteristics such as gender, age, and education, we collected participants' socioeconomic status (SES) using the Four Factor Index of Social Status , , which takes into account educational level, gender, occupation, and marital status. The Hollingshead Index (HI) combines these four factors to assess the social status of individuals or nuclear families. The HI scores can be aggregated into a set of values covering the social strata, ranging from 66 (the highest social stratum: 66–55) to 18 (the lowest social stratum: 19–18). The higher the score, the higher the SES. For example, social stratum 30–39, which reflects the average socioeconomic level, includes skilled craftsmen, clerical, and sales workers. Sociodemographic characteristics of participants are shown in Table , along with information on the pandemic. The participants were asked whether they had been (a) previously diagnosed with infection by COVID-19 and (b) vaccinated against SARS-CoV-2 (as well as the number of doses administered). In addition, the subjects were asked whether they had followed the recommendations of the Italian Ministry of Health to control infection, i.e., wearing face masks, wearing latex gloves when outdoors, washing hands frequently, keeping safe distances, and avoiding crowded places. Neuropsychological assessment The study included two time points: before the pandemic, from April to October 2019 (baseline), and during the pandemic, from June to December 2021 (follow-up). To avoid fatigue, a detailed neuropsychological assessment was divided into two sections, lasting approximately 45 min and administered on the same day. The neuropsychological battery was designed to assess three domains, namely cognitive performance, mood, as well as physical and health status. We assessed global cognitive performance with ACE-R , which includes the MMSE score. In addition, SCD was assessed before the pandemic using Jessen's criteria and at follow-up using CFI , which includes both self-report and partner-report to provide a more accurate measure. Executive function was assessed using the MoCA and the TMT—part A and B . TMT—Part A was used to assess speed of information processing (i.e., psychomotor speed). We assessed language comprehension with TT , while instant and delayed recall were analyzed with RMT . Different facets of mood were assessed with the following scales: AES , BDI , DIS , HARS , and MAS . We used the Phenotypic Frailty Model and the Cumulative Illness Rating Scale (CIRS ) to assess physical condition or frailty status and medical history. Five criteria were considered to examine the determinants of a possible physical frailty state: (a) weight loss, (b) grip strength, (c) self-reported fatigue, (d) decreased walking speed, and (e) reduced physical activity. Depending on the presence of the five criteria, participants could be classified as robust (none), prefrail (1 or 2), or frail, (3 or more) . COVID-19-related picture stimuli To assess individual differences in physiological arousal associated with the pandemic, we first searched the "Google images" website ( https://images.google.com/ ) for images showing the most critical period between the virus outbreak and the lockdown in Italy. Using the keyword "COVID”, we obtained an initial selection of 124 images, which were then evaluated separately by all authors to select those that met specific inclusion criteria: presence of people regardless of their age, sex, and sociodemographic status in the pandemic context (e.g., healthy individuals, patients, deceased individuals, medical personnel); presence of people in the most typical COVID-19 scenarios (e.g., everyday contexts and hospitals). Based on these criteria, a subset of images was selected for further evaluation based on the following criteria: presence of at least 2 people (i.e., "social" images); absence of redundant images, i.e., same content but different formats and resolutions (i.e., "unique" images); presence of unique signals of the pandemic, such as: masks that had to be worn by at least some of the people depicted; slogans (e.g., "Everything will be fine") and/or banners from balconies; the presence of individuals with exclusively Caucasian facial features (i.e., with the same ethnicity as the study participants) to convey greater familiarity and emotional relevance ("ethnicity"). These criteria resulted in a final data set of 75 images randomly presented to each participant during a psychophysiological recording of heart rate variability with self-determined duration. The entire data set had a duration between 45 and 60 min. Psychophysiological evaluation Thirty-five previously studied subjects (27 women; 62–82 years) underwent further psychophysiological recording of HRV while viewing images reminiscent of the initial and most stressful pandemic phase. HRV reflects cardiac autonomic activity (i.e., sympathovagal balance), which indicates the ability of the autonomic nervous system to respond flexibly to external stimuli and to respond to psychophysiological stressors . The aim of this assessment was to identify possible associations between pandemic-related changes in mood and/or cognitive performance and a proxy for emotional (dys)regulation represented by the ratio of low-frequency to high-frequency HRV, which in turn reflects cardiac sympathovagal balance. These data were acquired with the Nexus-4 blood volume pulse (BVP) and heartbeat measurement device and subsequently processed with custom-designed software in MATLAB 7.10.0 (R2010a) (The Mathworks, Inc; Natick, MA, USA). Each channel was synchronously recorded at 2048 Hz and extracted at 256 Hz to calculate the indices. Signal processing Cardiovascular and respiratory activities were monitored to assess both voluntary and autonomic effects of breathing on heart rate. The IBI (Inter-Beat-Interval extracted from the BVP sensor, recognized as a measure equivalent to the R-R interval from the electrocardiogram) was analyzed. Following the guidelines of the Task Force of the European Society of Cardiology and the North American Society for Pacing and Electrophysiology, typical indices of spectral HRV were used to assess the autonomic nervous system response – . Spectral analysis was performed using Fourier spectral methods and dedicated software. The rhythms were classified as very low frequency (VLF, < 0.04 Hz), low frequency (LF, from 0.04 to 0.15 Hz), and high frequency (HF, from 0.15 to 0.5 Hz) oscillations. This procedure allowed us to calculate the LF/HF ratio, also known as the index of sympathovagal balance. Statistical analysis Analyses were performed using Jamovi Statistics software (version 2.2.5.0). Two normality tests (i.e., Kolmogorov–Smirnov and Shapiro–Wilk) were performed to determine whether the variables were normally distributed. We computed BFs and paired-sample t-tests to compare the data collected before and during the pandemic. The BFs were designed to determine whether variables related to cognition, mood, and physical status remained stable between time points. We tested this hypothesis using the BF, which is a ratio between the probability of the data in the null hypothesis and the alternative hypothesis – . The evidence for similarity of the measures is considered substantial when 2.5 < BF < 10 – . It is noteworthy that in this case the measures considered are statistically similar, as opposed to the hypothesis that they are different. BF is generally considered a more reliable statistical test compared to p-value of a t-test. In particular, the American Psychological Association acknowledged the recommendations of the American Statistical Association and emphasized the importance of using BF, among other methods . In this study, we used BF to compare repeated measures and assume similarities or differences between them. Using BF, we were able to determine whether a model that predicted similarities was significantly better than a model that assumed differences. The BF provides the likelihood ratio for this comparison. To examine the relationship between neuropsychological performance at follow-up and emotional (dys)regulation as indexed by HRV recording, we computed a correlation matrix representing the relationship between (a) the variables associated with a statistically significant difference in the paired t-tests (comparison between before and during the pandemic) and (b) the LF/HF ratio index. The variables that showed a significant difference in the t-tests, in addition to the LF/HF ratio, were then selected as predictors in a multiple regression analysis (adjusted for age and gender), and the neuropsychological scale that showed the greatest change between timepoints was selected as the dependent variable. From April to October 2019, about 100 socially active older people from the University of the Third Age (UNITRE) in Turin (Italy) joined the initiatives of an innovative laboratory for active and healthy aging, dedicated to training, education and research and development. Before the outbreak of the pandemic, they participated in teaching modules on cognitive functions, physical activity, nutrition and social integration to promote active and healthy aging. Specifically, they were invited to undergo an in-depth neuropsychological assessment to investigate possible associations between cognitive functioning, mood changes, and physical health status over a 3-year period. A group of 81 volunteers (64 women; 60–80 years) agreed to participate in the study, but pandemic restrictions necessitated the establishment of an online assessment. After the lockdown measures were relaxed, 39 subjects agreed to undergo the final date of neuropsychological assessment between June and October 2021. Thirty-five subjects (27 women; 62–82 years) also agreed to participate in a psychophysiological study from October to December 2021 (i.e., 21 months after the pandemic outbreak) to assess emotional reactivity to images associated with the COVID-19 pandemic. All 39 subjects who participated in the full longitudinal study were > 60 years old and thus can be classified as "older adults" . They were not taking any psychotropic medications that could have affected their cognitive abilities or mood, and none of them complained of cognitive decline . The study was conducted in accordance with the Declaration of Helsinki, having been approved by the Ethics Committee of the University of Turin before (Prot. No. 10038) and during (Prot. No. 151786) the pandemic. All participants gave written informed consent before participating in the study. In addition to the usual sociodemographic characteristics such as gender, age, and education, we collected participants' socioeconomic status (SES) using the Four Factor Index of Social Status , , which takes into account educational level, gender, occupation, and marital status. The Hollingshead Index (HI) combines these four factors to assess the social status of individuals or nuclear families. The HI scores can be aggregated into a set of values covering the social strata, ranging from 66 (the highest social stratum: 66–55) to 18 (the lowest social stratum: 19–18). The higher the score, the higher the SES. For example, social stratum 30–39, which reflects the average socioeconomic level, includes skilled craftsmen, clerical, and sales workers. Sociodemographic characteristics of participants are shown in Table , along with information on the pandemic. The participants were asked whether they had been (a) previously diagnosed with infection by COVID-19 and (b) vaccinated against SARS-CoV-2 (as well as the number of doses administered). In addition, the subjects were asked whether they had followed the recommendations of the Italian Ministry of Health to control infection, i.e., wearing face masks, wearing latex gloves when outdoors, washing hands frequently, keeping safe distances, and avoiding crowded places. The study included two time points: before the pandemic, from April to October 2019 (baseline), and during the pandemic, from June to December 2021 (follow-up). To avoid fatigue, a detailed neuropsychological assessment was divided into two sections, lasting approximately 45 min and administered on the same day. The neuropsychological battery was designed to assess three domains, namely cognitive performance, mood, as well as physical and health status. We assessed global cognitive performance with ACE-R , which includes the MMSE score. In addition, SCD was assessed before the pandemic using Jessen's criteria and at follow-up using CFI , which includes both self-report and partner-report to provide a more accurate measure. Executive function was assessed using the MoCA and the TMT—part A and B . TMT—Part A was used to assess speed of information processing (i.e., psychomotor speed). We assessed language comprehension with TT , while instant and delayed recall were analyzed with RMT . Different facets of mood were assessed with the following scales: AES , BDI , DIS , HARS , and MAS . We used the Phenotypic Frailty Model and the Cumulative Illness Rating Scale (CIRS ) to assess physical condition or frailty status and medical history. Five criteria were considered to examine the determinants of a possible physical frailty state: (a) weight loss, (b) grip strength, (c) self-reported fatigue, (d) decreased walking speed, and (e) reduced physical activity. Depending on the presence of the five criteria, participants could be classified as robust (none), prefrail (1 or 2), or frail, (3 or more) . To assess individual differences in physiological arousal associated with the pandemic, we first searched the "Google images" website ( https://images.google.com/ ) for images showing the most critical period between the virus outbreak and the lockdown in Italy. Using the keyword "COVID”, we obtained an initial selection of 124 images, which were then evaluated separately by all authors to select those that met specific inclusion criteria: presence of people regardless of their age, sex, and sociodemographic status in the pandemic context (e.g., healthy individuals, patients, deceased individuals, medical personnel); presence of people in the most typical COVID-19 scenarios (e.g., everyday contexts and hospitals). Based on these criteria, a subset of images was selected for further evaluation based on the following criteria: presence of at least 2 people (i.e., "social" images); absence of redundant images, i.e., same content but different formats and resolutions (i.e., "unique" images); presence of unique signals of the pandemic, such as: masks that had to be worn by at least some of the people depicted; slogans (e.g., "Everything will be fine") and/or banners from balconies; the presence of individuals with exclusively Caucasian facial features (i.e., with the same ethnicity as the study participants) to convey greater familiarity and emotional relevance ("ethnicity"). These criteria resulted in a final data set of 75 images randomly presented to each participant during a psychophysiological recording of heart rate variability with self-determined duration. The entire data set had a duration between 45 and 60 min. Thirty-five previously studied subjects (27 women; 62–82 years) underwent further psychophysiological recording of HRV while viewing images reminiscent of the initial and most stressful pandemic phase. HRV reflects cardiac autonomic activity (i.e., sympathovagal balance), which indicates the ability of the autonomic nervous system to respond flexibly to external stimuli and to respond to psychophysiological stressors . The aim of this assessment was to identify possible associations between pandemic-related changes in mood and/or cognitive performance and a proxy for emotional (dys)regulation represented by the ratio of low-frequency to high-frequency HRV, which in turn reflects cardiac sympathovagal balance. These data were acquired with the Nexus-4 blood volume pulse (BVP) and heartbeat measurement device and subsequently processed with custom-designed software in MATLAB 7.10.0 (R2010a) (The Mathworks, Inc; Natick, MA, USA). Each channel was synchronously recorded at 2048 Hz and extracted at 256 Hz to calculate the indices. Cardiovascular and respiratory activities were monitored to assess both voluntary and autonomic effects of breathing on heart rate. The IBI (Inter-Beat-Interval extracted from the BVP sensor, recognized as a measure equivalent to the R-R interval from the electrocardiogram) was analyzed. Following the guidelines of the Task Force of the European Society of Cardiology and the North American Society for Pacing and Electrophysiology, typical indices of spectral HRV were used to assess the autonomic nervous system response – . Spectral analysis was performed using Fourier spectral methods and dedicated software. The rhythms were classified as very low frequency (VLF, < 0.04 Hz), low frequency (LF, from 0.04 to 0.15 Hz), and high frequency (HF, from 0.15 to 0.5 Hz) oscillations. This procedure allowed us to calculate the LF/HF ratio, also known as the index of sympathovagal balance. Analyses were performed using Jamovi Statistics software (version 2.2.5.0). Two normality tests (i.e., Kolmogorov–Smirnov and Shapiro–Wilk) were performed to determine whether the variables were normally distributed. We computed BFs and paired-sample t-tests to compare the data collected before and during the pandemic. The BFs were designed to determine whether variables related to cognition, mood, and physical status remained stable between time points. We tested this hypothesis using the BF, which is a ratio between the probability of the data in the null hypothesis and the alternative hypothesis – . The evidence for similarity of the measures is considered substantial when 2.5 < BF < 10 – . It is noteworthy that in this case the measures considered are statistically similar, as opposed to the hypothesis that they are different. BF is generally considered a more reliable statistical test compared to p-value of a t-test. In particular, the American Psychological Association acknowledged the recommendations of the American Statistical Association and emphasized the importance of using BF, among other methods . In this study, we used BF to compare repeated measures and assume similarities or differences between them. Using BF, we were able to determine whether a model that predicted similarities was significantly better than a model that assumed differences. The BF provides the likelihood ratio for this comparison. To examine the relationship between neuropsychological performance at follow-up and emotional (dys)regulation as indexed by HRV recording, we computed a correlation matrix representing the relationship between (a) the variables associated with a statistically significant difference in the paired t-tests (comparison between before and during the pandemic) and (b) the LF/HF ratio index. The variables that showed a significant difference in the t-tests, in addition to the LF/HF ratio, were then selected as predictors in a multiple regression analysis (adjusted for age and gender), and the neuropsychological scale that showed the greatest change between timepoints was selected as the dependent variable. |
Shenmai injection revives cardiac function in rats with hypertensive heart failure: involvement of microbial-host co-metabolism | b74536a1-ed59-43dc-973e-36e405f5048d | 11761217 | Biochemistry[mh] | Heart failure (HF) represents the most advanced stage of cardiac illnesses . The elderly population is predominantly burdened with chronic illnesses such as heart failure, and this causes substantial economic strain on the patient’s family and society . In up to 85% of cases, hypertension (HTN) is the primary modifiable risk factor for the onset of HF . The current estimated population of HF in the US is 6 million , making hypertension-induced HF a serious public health concern. New studies have demonstrated a link between intestinal microbial illnesses and the onset and progression of HF . It has been hypothesized that HF is caused by an increase in intestinal bacteria, resulting in increased inflammation and an increased number of bacteria in the bloodstream . Cui et al. used metabolomic and metagenomic analyses of feces and blood from HF patients to reveal an imbalance among intestinal microflora . According to Pasini et al., the severity of HF may be correlated with the proliferation of pathogenic intestinal microflora and increased intestinal permeability . The discovery of a heart-gut axis provides new approaches to the therapy of HF . Trimethylamine N-oxide (TMAO) is formed when trimethylamine (TMA), produced by gut microbes acting on dietary compounds, is further oxidized by liver flavin-containing mono-oxygenase (FMO) . A systematic review of 19,256 subjects has previously demonstrated that raised levels of TMAO and its precursors were reported to be related to major detrimental cardiovascular complications, such as HF, and an increased risk of death from any cause . An investigation using mice fed a diet high in choline or TMAO revealed heightened serum levels of TMAO and a more severe HF . These findings support the notion that a better prognosis for HF patients is associated with lower TMAO level . Therefore, methods for treating HF that lower the level of serum TMAO or TMAO-producing microbes are ideal. Numerous studies have provided ample evidence of the efficacy of Traditional Chinese medicine (TCM) in HF therapy . Based on recent research, TCM has been proven to regulate gut microbiota to inhibit the onset of cardiovascular diseases . TCM successfully maintains a healthy intestinal environment, encourages the propagation of beneficial bacteria, and balances the gut microbiota. While also inhibiting the proliferation of harmful bacteria . Shenmai Injection (SMI) is a type of Traditional Chinese Medicine injection (TCMI) prepared from Ginseng Radix et Rhizoma Rubra (Panax ginseng C. A. Mey, Hongshen) and Radix Ophiopogonis ( Ophiopogon japonicus (Linn. f.) Ker-Gawl, Maidong) using modern technology. Pharmacological studies established that ophiopogonin D, ginsenoside Rg1, ginsenoside Rb1, and ginsenoside Re are the compounds of SMI. It has been demonstrated that SMI protects cardiomyocytes through the regulation of the activity of enzymes related to energy metabolism, ATP production, and mitochondrial function . Research has demonstrated that SMI can protect against the cardiotoxicity caused by doxorubicin by maintaining mitochondrial homeostasis and the miR-30a/Beclin pathway . By activating Nrf2/GPX4 signaling-mediated ferroptosis, pretreatment with SMI minimized myocardial I/R damage and presented a therapeutic approach to treating and preventing ischemic heart diseases . SMI has been shown in clinical trials to improve energy metabolism in HF patients . Our previous research has demonstrated that SMI regulates the TGF-β 1/Smad signaling pathway, thereby preventing myocardial fibrosis and effectively improving H-HF , but the potential involvement of gut microbiota in these therapeutic effects remains unclear. Studies have shown that ginsenoside Rg1 alleviates acute ulcerative colitis by modulating gut microbiota and microbial tryptophan metabolism , while ginsenoside Rh4 inhibits colorectal cancer through the regulation of gut microbiota-mediated bile acid metabolism . However, there is currently no research exploring the impact of Shenmai Injection on gut microbiota.Thus, we aim to investigate the mechanisms behind the improvement of cardiac activity in chronic heart failure by using SMI. Animals and treatment The Institutional Animal Care and Use Committee (IACUC) at the Hunan University of Chinese Medicine (HUMC) approved the experimental protocol. Salt-sensitive rats ( n = 24), aged six weeks and weighing 200–220 g, were obtained from Beijing Weitong Lihua Animal Co., Ltd., license number: SCXK(Beijing)2016-0011, animal batch number: N1100111911056755.All of the animals were housed in a standard husbandry environment. Following the acclimation period of seven days, eight rats were allocated to three groups at random: Control (CON), H-HF Model (MOD), and H-HF Model with Shenmai injection (SM). The rat model of H-HF was created utilizing the procedures previously reported . For 20 weeks, the Control (CON) group had a regular diet containing 0.3% NaCl(normal diet), while the MOD and SM groups received a diet high in salt (8% NaCl). Animals were supplied with an unlimited supply of food and water. The CON and MOD groups received intraperitoneal injections of sterile water (6.0 mL/kg), whereas the SMI group received Shenmai injections (6.0 mL/kg) for a period of 15 days. Samples The rats were anesthetized after 15 days using urethane (1.0 g/kg, i.p.). Blood was drawn from the abdominal aorta, and euthanasia was via dislocation of the neck. Briefly, the rats were held securely by the body, with one hand gripping the back or base of the tail. The head was quickly pulled downward and backward to separate the cervical vertebrae, causing immediate loss of consciousness. Blood was left for 3 h at room temperature and then centrifuged to separate and obtain serum. Myocardial and colonic tissues preserved with 4% paraformaldehyde underwent histopathological evaluation. After that, sections embedded in paraffin were stained with hematoxylin and eosin (HE). ELISA was conducted using commercial ELISA kits. NT-proBNP ELISA Kit(CUSABIO, CSB-E08752r). CRP ELISA Kit(CUSABIO, CSB-E07922r). IL-1β ELISA Kit(CUSABIO, CSB-E08055r). Zonulin ELISA Kit(mlbio, ml059419. LPS ELISA Kit(CUSABIO, CSB-E14247r). Samples from the colon were obtained by firmly compressing the inner contents into a clean tube, which was then frozen with liquid nitrogen and stored at -80 °C. For 16S rRNA sequencing and microbiome analysis, 6 rats were picked from each group at random. Echocardiography and blood pressure measurement Echocardiography was employed to assess cardiac function with an ultrasound color Doppler diagnostic equipment(S2N, Shenzhen Kaili Technology Co., Ltd., China). The dimensions of the left atrium and ventricle, as well as the left ventricular ejection fraction (LVEF), were assessed using M-mode echocardiography in the parasternal long-axis view, following the American Society of Echocardiography’s M-mode technique. Three consecutive cardiac cycles were examined to calculate the mean value. The Teichholtz formula was utilized to calculate the left ventricular fractional shortening (LVFS) and LVEF. The echocardiography parameters are as follows: frame rate = 54, dynamic range/gain = 100/3, gain = 150, frequency = 8.0–12.0 MHz. Blood pressure was assessed using a Volume Pressure Recording (VPR) system (CODA; Kent Scientific). For each animal, systolic blood pressure (SBP) and diastolic blood pressure (DBP) were calculated as the average of three independent measurements. Quality control of Shenmai injection Shenmai injection (lot number,1909288) was manufactured by Chiatai Qingchunbao(CTQ) Pharmaceutical Co. Ltd. (Hanzhou, China) with a China FDA drug ratification number of GuoYaoZhunZi- Z33020019. It is a solution extracted from Ginseng Radix et Rhizoma Rubra (Panax ginseng C. A. Mey, Hongshen) and Radix Ophiopogonis ( Ophiopogon japonicus (Linn. f.) Ker-Gawl, Maidong), as described in Table . Its quality meets the standard of China Food and Drug Administation (approval No: WS3-B-3428-98-2004).According to the CTQ Pharmaceutical Group Co. Ltd, the quality control standards for SMI require that the total concentration of GinsenosideRg1 (C42H72O14), GinsenosideRe (C48H82O18), and GinsenosideRb1(C54H92O23) must not be lower than 100 µg/mL and that the overall concentration of the three agents should be between 300 and 600 µg/mL . To ensure the quality of the Shenmai injection, High-Performance Liquid Chromatography (HPLC) analysis was performed.Following filtration through a 0.22 μm nylon membrane, the components of SMI were analyzed using an HPLC System (U3000, ThermoFisher Scientific). Detailed amplification conditions can be found in the Supplementary Material and Methods file. The HPLC analysis demonstrated the presence of Ginsenoside Rg1, Ginsenoside Re, and GinsenosideRb1 in SMI, which agreed with the results reported previously . The outputs of HPLC are summarized in Fig. . Network pharmacology methodology Following the requirements of oral bioavailability (OB) of ≥ 30% and drug-likeness (DL) of ≥ 0.18, we were able to determine all of the active ingredients in red ginseng (hong shen) using the Traditional Chinese Medicine Database Analysis Platform (TCMSP, https://tcmsp-e.com/ ) . The BATMAN-TCM database ( http://bionet.ncpsb.org.cn/batman-tcm/ ) provided high-confidence proteins for Ophiopogon japonicus (Maidong) . Disease-related targets were identified by searching for the term “heart failure” in the databases OMIM ( https://www.omim.org/ ) and GeneCards ( https://www.genecards.org/ ). We used Cytoscape 3.7.2 to construct the “drug component-target” network by mapping the targets of the drug component to the targets of the disease. Based on the STRING database ( https://string-db.org/ ), a protein-protein interaction network was constructed, with a minimum interaction score of 0.7. The drug-disease intersecting genes were uploaded to the DAVID database ( https://david.ncifcrf.gov/summary.jsp ), with gene identifiers set to OFFICIAL_GENE_SYMBOL and the species set to Homo sapiens. DAVID 6.8 was utilized to annotate GO gene functions into three categories: Molecular Function (MF), Cellular Component (CC), and Biological Process (BP) to describe the function of active proteins in Shenmai injection therapy for heart failure. Fecal metabolic profiling Fecal Metabolic Profiling was carried out using the procedures described in our earlier study . A QC sample was generated by combining an equivalent amount of sample supernatant (Fig. A, B).Analysis of the negative and positive modes identified 12,356 and 14,389 peaks, respectively, identifying 344 and 1,058 metabolites. The same software package was used for multivariate analysis, where normalized peak area data was imported into SIMCA16.0.2 . Online databases such as HMDB, ChemSpider, and KEGG were searched to identify metabolites with a VIP greater than 1 and a P-value of less than 0.05 (ascertained by Student’s t-test). Metabolomics analyses were conducted by Biotree Biomedical Company (Shanghai, China). 16S rRNA sequencing The 16S rRNA sequencing analysis was carried out using the procedures described in our earlier research . In brief, PCR amplification was carried out, and the purified amplicons were pooled and sequenced using paired-end sequencing. The raw data was subsequently evaluated. Detailed sequencing analysis procedures are provided in the Supplementary Material and Methods. Biotree Biomedical Company (Shanghai, China) was responsible for sequencing and analysis. Quantification of serum TMAO Using UHPLC-MRM-MS/MS, the Agilent 1290 Infinity II series UHPLC System (Agilent Technologies) was employed to analyze the supernatant (80 µL). The Agilent 6460 triple quadrupole mass spectrometer, outfitted with an AJS electrospray ionization interface, was utilized to create an assay. Biotree Biomedical Company performed the analysis while Agilent MassHunter Workstation Software (B.08.00, Agilent Technologies) was utilized for the MRM data processing and capture. Detailed amplification conditions can be found in the Supplementary Material and Methods file. Metabolomics analyses were conducted by Biotree Biomedical Company (Shanghai, China). Correlation network among “compounds-targets-metabolites- microbiota” Correlation coefficients between the different gut microbiotas and metabolites were computed using the Spearman correlation analysis method. To identify the correlation between metabolites and targets, more SMI differential metabolites and targets were integrated into the metaboanalyst platform . A “components-targets-metabolites-microbes” interaction network was created by integrating the aforementioned results to further reveal the regulatory function of SMI against H-HF. OmicShare, an online tool, was used for the visualization process. Statistical analysis The data analysis was carried out with SPSS 22.0 (IBM, USA). Data with equal variances and normal distribution were assessed for significance using one-way ANOVA and Tukey’s post hoc test. Otherwise, the Mann-Whitney U test was used. A significance threshold of p < 0.05 was established. Additionally, Metorigin ( http://metorigin.metbioinformatics.cn/ ) was used to analyze the traceability of differential metabolites. Sankey network generation, origin analysis, and function analysis were all carried out utilizing the basic Metorigin analysis mode that is accessible on the official website. The Institutional Animal Care and Use Committee (IACUC) at the Hunan University of Chinese Medicine (HUMC) approved the experimental protocol. Salt-sensitive rats ( n = 24), aged six weeks and weighing 200–220 g, were obtained from Beijing Weitong Lihua Animal Co., Ltd., license number: SCXK(Beijing)2016-0011, animal batch number: N1100111911056755.All of the animals were housed in a standard husbandry environment. Following the acclimation period of seven days, eight rats were allocated to three groups at random: Control (CON), H-HF Model (MOD), and H-HF Model with Shenmai injection (SM). The rat model of H-HF was created utilizing the procedures previously reported . For 20 weeks, the Control (CON) group had a regular diet containing 0.3% NaCl(normal diet), while the MOD and SM groups received a diet high in salt (8% NaCl). Animals were supplied with an unlimited supply of food and water. The CON and MOD groups received intraperitoneal injections of sterile water (6.0 mL/kg), whereas the SMI group received Shenmai injections (6.0 mL/kg) for a period of 15 days. The rats were anesthetized after 15 days using urethane (1.0 g/kg, i.p.). Blood was drawn from the abdominal aorta, and euthanasia was via dislocation of the neck. Briefly, the rats were held securely by the body, with one hand gripping the back or base of the tail. The head was quickly pulled downward and backward to separate the cervical vertebrae, causing immediate loss of consciousness. Blood was left for 3 h at room temperature and then centrifuged to separate and obtain serum. Myocardial and colonic tissues preserved with 4% paraformaldehyde underwent histopathological evaluation. After that, sections embedded in paraffin were stained with hematoxylin and eosin (HE). ELISA was conducted using commercial ELISA kits. NT-proBNP ELISA Kit(CUSABIO, CSB-E08752r). CRP ELISA Kit(CUSABIO, CSB-E07922r). IL-1β ELISA Kit(CUSABIO, CSB-E08055r). Zonulin ELISA Kit(mlbio, ml059419. LPS ELISA Kit(CUSABIO, CSB-E14247r). Samples from the colon were obtained by firmly compressing the inner contents into a clean tube, which was then frozen with liquid nitrogen and stored at -80 °C. For 16S rRNA sequencing and microbiome analysis, 6 rats were picked from each group at random. Echocardiography was employed to assess cardiac function with an ultrasound color Doppler diagnostic equipment(S2N, Shenzhen Kaili Technology Co., Ltd., China). The dimensions of the left atrium and ventricle, as well as the left ventricular ejection fraction (LVEF), were assessed using M-mode echocardiography in the parasternal long-axis view, following the American Society of Echocardiography’s M-mode technique. Three consecutive cardiac cycles were examined to calculate the mean value. The Teichholtz formula was utilized to calculate the left ventricular fractional shortening (LVFS) and LVEF. The echocardiography parameters are as follows: frame rate = 54, dynamic range/gain = 100/3, gain = 150, frequency = 8.0–12.0 MHz. Blood pressure was assessed using a Volume Pressure Recording (VPR) system (CODA; Kent Scientific). For each animal, systolic blood pressure (SBP) and diastolic blood pressure (DBP) were calculated as the average of three independent measurements. Shenmai injection (lot number,1909288) was manufactured by Chiatai Qingchunbao(CTQ) Pharmaceutical Co. Ltd. (Hanzhou, China) with a China FDA drug ratification number of GuoYaoZhunZi- Z33020019. It is a solution extracted from Ginseng Radix et Rhizoma Rubra (Panax ginseng C. A. Mey, Hongshen) and Radix Ophiopogonis ( Ophiopogon japonicus (Linn. f.) Ker-Gawl, Maidong), as described in Table . Its quality meets the standard of China Food and Drug Administation (approval No: WS3-B-3428-98-2004).According to the CTQ Pharmaceutical Group Co. Ltd, the quality control standards for SMI require that the total concentration of GinsenosideRg1 (C42H72O14), GinsenosideRe (C48H82O18), and GinsenosideRb1(C54H92O23) must not be lower than 100 µg/mL and that the overall concentration of the three agents should be between 300 and 600 µg/mL . To ensure the quality of the Shenmai injection, High-Performance Liquid Chromatography (HPLC) analysis was performed.Following filtration through a 0.22 μm nylon membrane, the components of SMI were analyzed using an HPLC System (U3000, ThermoFisher Scientific). Detailed amplification conditions can be found in the Supplementary Material and Methods file. The HPLC analysis demonstrated the presence of Ginsenoside Rg1, Ginsenoside Re, and GinsenosideRb1 in SMI, which agreed with the results reported previously . The outputs of HPLC are summarized in Fig. . Following the requirements of oral bioavailability (OB) of ≥ 30% and drug-likeness (DL) of ≥ 0.18, we were able to determine all of the active ingredients in red ginseng (hong shen) using the Traditional Chinese Medicine Database Analysis Platform (TCMSP, https://tcmsp-e.com/ ) . The BATMAN-TCM database ( http://bionet.ncpsb.org.cn/batman-tcm/ ) provided high-confidence proteins for Ophiopogon japonicus (Maidong) . Disease-related targets were identified by searching for the term “heart failure” in the databases OMIM ( https://www.omim.org/ ) and GeneCards ( https://www.genecards.org/ ). We used Cytoscape 3.7.2 to construct the “drug component-target” network by mapping the targets of the drug component to the targets of the disease. Based on the STRING database ( https://string-db.org/ ), a protein-protein interaction network was constructed, with a minimum interaction score of 0.7. The drug-disease intersecting genes were uploaded to the DAVID database ( https://david.ncifcrf.gov/summary.jsp ), with gene identifiers set to OFFICIAL_GENE_SYMBOL and the species set to Homo sapiens. DAVID 6.8 was utilized to annotate GO gene functions into three categories: Molecular Function (MF), Cellular Component (CC), and Biological Process (BP) to describe the function of active proteins in Shenmai injection therapy for heart failure. Fecal Metabolic Profiling was carried out using the procedures described in our earlier study . A QC sample was generated by combining an equivalent amount of sample supernatant (Fig. A, B).Analysis of the negative and positive modes identified 12,356 and 14,389 peaks, respectively, identifying 344 and 1,058 metabolites. The same software package was used for multivariate analysis, where normalized peak area data was imported into SIMCA16.0.2 . Online databases such as HMDB, ChemSpider, and KEGG were searched to identify metabolites with a VIP greater than 1 and a P-value of less than 0.05 (ascertained by Student’s t-test). Metabolomics analyses were conducted by Biotree Biomedical Company (Shanghai, China). The 16S rRNA sequencing analysis was carried out using the procedures described in our earlier research . In brief, PCR amplification was carried out, and the purified amplicons were pooled and sequenced using paired-end sequencing. The raw data was subsequently evaluated. Detailed sequencing analysis procedures are provided in the Supplementary Material and Methods. Biotree Biomedical Company (Shanghai, China) was responsible for sequencing and analysis. Using UHPLC-MRM-MS/MS, the Agilent 1290 Infinity II series UHPLC System (Agilent Technologies) was employed to analyze the supernatant (80 µL). The Agilent 6460 triple quadrupole mass spectrometer, outfitted with an AJS electrospray ionization interface, was utilized to create an assay. Biotree Biomedical Company performed the analysis while Agilent MassHunter Workstation Software (B.08.00, Agilent Technologies) was utilized for the MRM data processing and capture. Detailed amplification conditions can be found in the Supplementary Material and Methods file. Metabolomics analyses were conducted by Biotree Biomedical Company (Shanghai, China). Correlation coefficients between the different gut microbiotas and metabolites were computed using the Spearman correlation analysis method. To identify the correlation between metabolites and targets, more SMI differential metabolites and targets were integrated into the metaboanalyst platform . A “components-targets-metabolites-microbes” interaction network was created by integrating the aforementioned results to further reveal the regulatory function of SMI against H-HF. OmicShare, an online tool, was used for the visualization process. The data analysis was carried out with SPSS 22.0 (IBM, USA). Data with equal variances and normal distribution were assessed for significance using one-way ANOVA and Tukey’s post hoc test. Otherwise, the Mann-Whitney U test was used. A significance threshold of p < 0.05 was established. Additionally, Metorigin ( http://metorigin.metbioinformatics.cn/ ) was used to analyze the traceability of differential metabolites. Sankey network generation, origin analysis, and function analysis were all carried out utilizing the basic Metorigin analysis mode that is accessible on the official website. Pharmacodynamic study of SMI against H-HF rats SMI improved cardiac function We observed that the blood pressure of the MOD and SM groups increased to 190/150mmHg at 12 weeks (Fig. A-B). The control group was found to have significantly lower SBP and DBP readings than the MOD and SM groups. Following the treatment, no substantial changes in blood pressure were observed between the groups (Fig. C-D). Furthermore, no alterations in the weight of rats were observed in each group after the intervention (Fig. E). To validate the H-HF rat model, we first assessed the serum level of NT-proBNP. It was observed that the MOD group had a higher NT-proBNP serum level than the CON group (Fig. A). Comparing the MOD groups to the CON group, the MOD groups showed lower levels of LVEF and LVFS (Fig. B and C), and the MOD group’s M-mode echocardiogram showed impaired cardiac performance (Fig. F). MOD cardiomyocytes were observed by HE staining to be enlarged, irregularly shaped, with a disordered arrangement; the interstitial space between the cells was also filled with fibrous tissue and heavily infiltrated with inflammatory cells (Fig. G). CRP and IL-1β are cytokines used to identify inflammation . The MOD group displayed elevated serum CRP and IL-1β levels compared to controls (Fig. D and E). These observations corroborated the H-HF model , indicating that establishing the H-HF rat model had succeeded. Administering SMI to H-HF rats reduced the elevation of NT-proBNP, CRP, and IL-1β levels. The SMI treatment restored decreased levels of LVEF and LVFS in the MOD group (Fig. B and C). The cardiac functions were also restored by the administration of SMI, as manifested by M-mode echocardiogram (Fig. F) and HE staining (Fig. G). SMI improved intestinal barrier function HE staining of colonic tissue indicated a reduction in mucosal integrity and increased inflammatory cells in the MOD group (Fig. I). This impairment of mucosal functions was reversed with SMI treatment. The presence of Lipopolysaccharide (LPS) is indicative of damage to the intestinal mucosa , and Zonulin is used to assess intestinal permeability . In H-HF rats, the MOD group showed noticeably higher serum concentrations of LPS and Zonulin than the control group, indicating a breakdown of the intestinal mucosal barrier and increased intestinal permeability (Fig. H and J). LPS and Zonulin levels were lower in the SMI group compared to the MOD group, suggesting that SMI effectively improved intestinal permeability and intestinal barrier function. Network pharmacology analysis Three compounds were collected from Red Ginseng (Hong Shen) and ten compounds were obtained from Ophiopogon japonicus (Mai Dong) (Table ). Based on searches of the GeneCards and OMIM disease databases, 4158 heart failure (HF)-related disease targets and 122 overlapping targets were discovered (Fig. A). TNF, IL-6, IL-1β, AKT1, STAT3, NFκΒ, IFNG, IL-10, TP53, and TLR4 were identified as the primary targets by PPI protein interaction analysis (Fig. B). Figure C illustrates the findings of the “drug-component-disease-target” network. As suggested by KEGG analysis, the TNF, IL-17, and Toll-like receptor signaling pathways may be involved in the mechanism through which SMI prevents and treats HF (Fig. D). Apoptotic processes, inflammatory responses, response to external biotic stimuli, negative regulation of cell proliferation, negative regulation of the apoptotic process, cellular response to lipopolysaccharide, G protein-coupled receptor signaling pathway, and negative regulation of gene expression are among the main biological processes predicted by GO analysis and included 474 significantly enriched biological function entries for treating heart failure (Fig. E). There are 52 entries related to cellular components (CC), involving the plasma membrane, membrane, cytoplasm, extracellular space, extracellular region, extracellular exosome, cell surface, mitochondrion, endoplasmic reticulum membrane, and endoplasmic reticulum. Additionally, there are 80 entries related to molecular functions, involving protein binding, identical protein binding, enzyme binding, protein homodimerization activity, DNA binding, zinc ion binding, heme binding, signaling receptor activity, sequence-specific DNA binding, and receptor binding. SMI restored the gut microbiota of H-HF rats Sequencing analysis of gut microbiota The sequencing of 18 fecal samples yielded 1,440,437 raw reads, which were merged and filtered to produce 1,408,229 clean tags. On average, 67,749 clean tags were obtained. To determine whether the sequencing data adequately reflected the diversity of species in the sample, a rarefaction curve was employed. Overall consistency in the results revealed that the sequencing data was adequate (Fig. A). A Venn diagram depicting the OTU distributions was shown in Fig. B. Across the three groups, 607 OTUs were identified, with 518 being shared by all of them. Alpha diversity analysis was carried out to assess the disparities in the structural complexity of the gut microbiota. Chao 1 and Shannon indices did not uncover any significant discrepancies in diversity across the three groups (Fig. C, D). However, a distinct divergence of profiles was found between the CON, MOD, and SM groups according to weighted unifrac PCoA of beta diversity (Fig. C). ANOSIM analysis (ANOSIM: R = 0.732, p = 0.001) demonstrated that the three groups were distinctly segregated. The proximity of the CON and SM populations indicated that their gut bacteria profiles were similar. The points representing the MOD group were further away from the points of the CON and SM groups, implying that the MOD group’s colony structure was markedly different from the other two groups. The PCoA outcomes (model stress = 0.0959 < 0.2) were supported by the NMDS analysis (Fig. D). Composition of gut microbiota and its difference analysis The microbial community composition was evaluated at the phylum and genus level (Fig. E and F). Results revealed that the MOD group had a reduced abundance of Bacteroidetes, Patescibacteria, Spirochaetes, and Elusimicrobia, and an elevated proportion of Firmicutes and Proteobacteria in comparison to the controls (Fig. A). As shown in Fig. A(g), the F/B ratio of the MOD group was substantially higher in comparison to that of the other group. The MOD group was noticed to be deficient in Muribaculaceae and Lachnospiraceae_NK4A136_group in comparison to the controls while having an increased abundance of Romboutsia and Ruminococcaceae at the genus level. These findings suggest that H-HF rats had an alteration in their gut bacterial equilibrium, which SMI treatment partially reversed (Fig. A). The heatmaps in Fig. B further illustrate the variations in the gut microbiota between the three groups. Prediction of the function of the gut microbiota By utilizing PICRUSt, a KEGG pathway analysis technique, we were able to evaluate the functional composition of the bacterial communities in the metagenome. All functional genes were shown at level III (Fig. ). The genes associated with energy and amino acid metabolism were more abundant in the MOD group, indicating that these metabolic pathways were disturbed in H-HF. Furthermore, after receiving SMI treatment, several genes related to energy and amino acid metabolism were altered (Fig. ). SMI improved disordered metabolism in H-HF Regulation of SMI on the differential metabolites Principal component analysis (PCA) and orthogonal partial least square discriminate analysis (OPLS-DA), which scaled and log-translated the data to reduce noise and high variance effects, were successful in differentiating between the three groups. PCA demonstrated a clear distinction between the three groups, with the SM group being closer to the CON group than the MOD group (Fig. A and B). OPLS-DA analysis verified the distinctiveness among the three groups identified by PCA (Fig. ). Metabolic analysis revealed that 17 metabolites had significantly altered levels in positive mode and other 17 metabolites in negative ion mode, while 29 biomarkers were significantly restored after SMI treatment (Table , Fig. C-D). Our results revealed that these metabolites were linked to energy, methylamine, bile acid, and amino acid metabolisms (Table ). Metorgin tracing analysis The analysis of source-based metabolic function and metabolite traceability found 29 differential metabolites linked to SMI: 6 bacterial metabolites, 1 host-specific metabolite, and 24 bacteria-host cometabolites (Fig. A-B). According to metabolites pathway enrichment analysis (MPEA), the databases for the host, bacterial, and co-metabolism metabolic pathways were paired with 1, 4, and 28 relevant metabolic pathways, respectively (Fig. C-D). Of these pathways, 1, 1, and 10 revealed a significant ( p < 0.05) association with SMI. The origin-based functional analysis revealed that the microbial community was specific to phenylalanine metabolism and that the host was specific to primary bile acid biosynthesis. Histidine metabolism, tryptophan metabolism, valine, leucine, and isoleucine biosynthesis, beta-Alanine metabolism, butanoate metabolism, inositol phosphate metabolism, ascorbate and aldarate metabolism, nicotinate and nicotinamide metabolism, aminoacyl-tRNA biosynthesis and pentose and glucuronate interconversions were pathways of co-metabolism between microbes and hosts. The primary mechanism linked to SMI was histidine metabolism. A Bio-Sankey network based on MetOrigin analysis further visualized the biological relationships and statistical correlations between microbiota and metabolites to better depict the co-metabolic relationships between microbiota and hosts (Fig. A-B). Quantification of serum TMAO Studies have demonstrated a positive association between TMAO levels and cardiovascular conditions . In comparison to the controls, the serum levels of TMAO and TMA levels in the MOD group were considerably higher (Fig. A–B), which supports earlier findings . The drops in serum levels of TMAO and TMA after SMI treatment were not statistically significant, which may have been due to the small sample size. Correlation analysis Correlations between the gut microbiota and fecal metabolic phenotype The relationship between metabolites and gut genera was evaluated by the Spearman correlation coefficient. A strong correlation is indicated by a value of r greater than 0.7. Figure C displays the network diagram with strong correlations. TMAO was strongly positively correlated with Elusimicrobium , _xylanophilum_group , oxidoreducens_group and negatively correlated with Catabacter , Defluviitaleaceae_UCG-011 , Parvibacter , _f_Atopobiaceae , Peptococcus , Coriobacteriaceae_UCG-002 , Staphylococcus , and Romboutsi . Therefore, methylamine metabolism must be impacted by the gut bacteria mentioned above. Similarly, D-glucuronic acid and D-xylitol correlated positively with the xylanophilum group , whereas creatinine correlated negatively with Coriobacteriaceae_UCG-002 . These findings suggest that Coriobacteriaceae_UCG-002 , Dubosiella , and the xylanophilum_group have an impact on energy metabolism. Bile acid metabolites (including chenodeoxycholic acid, cholic acid, and glycocholic acid) and amino acids (such as norvaline and L-threonine) have also been found to have a strong correlation with gut microbe composition. Our correlation data demonstrated modifications to the gut microbiome, resulting in a significantly altered metabolomic profile. Therefore, our current findings suggest that the mechanism by which SMI can improve heart function in an H-HF rat model may include effects on microbial energy, methylamine, bile acid, and amino acid metabolism in the intestine. Correlations between TMAO and gut microbiota Tables and present the results of Spearman’s correlation analysis used to evaluate the associations between the composition of the gut and the levels of TMAO metabolites. A positive correlation was uncovered between serum TMAO levels and the proportion of Actinobacteria. In contrast, a negative correlation was observed with Elusimicrobia at the phylum level (Fig. D). The analysis showed that serum TMAO had a direct relationship with Romboutsia , and an inverse relationship with Ruminococcaceae_UCG_014 and _f_Muribaculaceae . Furthermore, serum TMA levels had a strong negative correlation with Ruminococcaceae_UCG_014 (Fig. E). Based on these results, it is possible that SMI administration could alter TMAO levels by affecting the relevant microflora. Correlations between differential metabolites and targets of SMI Figure shows the connections between the targets of SMI and differential metabolites. Additionally, the regulatory role of SMI in preventing heart failure is highlighted by the “components-targets-metabolites-microbes” interaction network depicted in Fig. . By controlling for 46 proteins, network integration analysis shows that the 11 potentially active components of SMI can affect the differential expression of 8 metabolites and 24 gut microbes. SMI improved cardiac function We observed that the blood pressure of the MOD and SM groups increased to 190/150mmHg at 12 weeks (Fig. A-B). The control group was found to have significantly lower SBP and DBP readings than the MOD and SM groups. Following the treatment, no substantial changes in blood pressure were observed between the groups (Fig. C-D). Furthermore, no alterations in the weight of rats were observed in each group after the intervention (Fig. E). To validate the H-HF rat model, we first assessed the serum level of NT-proBNP. It was observed that the MOD group had a higher NT-proBNP serum level than the CON group (Fig. A). Comparing the MOD groups to the CON group, the MOD groups showed lower levels of LVEF and LVFS (Fig. B and C), and the MOD group’s M-mode echocardiogram showed impaired cardiac performance (Fig. F). MOD cardiomyocytes were observed by HE staining to be enlarged, irregularly shaped, with a disordered arrangement; the interstitial space between the cells was also filled with fibrous tissue and heavily infiltrated with inflammatory cells (Fig. G). CRP and IL-1β are cytokines used to identify inflammation . The MOD group displayed elevated serum CRP and IL-1β levels compared to controls (Fig. D and E). These observations corroborated the H-HF model , indicating that establishing the H-HF rat model had succeeded. Administering SMI to H-HF rats reduced the elevation of NT-proBNP, CRP, and IL-1β levels. The SMI treatment restored decreased levels of LVEF and LVFS in the MOD group (Fig. B and C). The cardiac functions were also restored by the administration of SMI, as manifested by M-mode echocardiogram (Fig. F) and HE staining (Fig. G). SMI improved intestinal barrier function HE staining of colonic tissue indicated a reduction in mucosal integrity and increased inflammatory cells in the MOD group (Fig. I). This impairment of mucosal functions was reversed with SMI treatment. The presence of Lipopolysaccharide (LPS) is indicative of damage to the intestinal mucosa , and Zonulin is used to assess intestinal permeability . In H-HF rats, the MOD group showed noticeably higher serum concentrations of LPS and Zonulin than the control group, indicating a breakdown of the intestinal mucosal barrier and increased intestinal permeability (Fig. H and J). LPS and Zonulin levels were lower in the SMI group compared to the MOD group, suggesting that SMI effectively improved intestinal permeability and intestinal barrier function. We observed that the blood pressure of the MOD and SM groups increased to 190/150mmHg at 12 weeks (Fig. A-B). The control group was found to have significantly lower SBP and DBP readings than the MOD and SM groups. Following the treatment, no substantial changes in blood pressure were observed between the groups (Fig. C-D). Furthermore, no alterations in the weight of rats were observed in each group after the intervention (Fig. E). To validate the H-HF rat model, we first assessed the serum level of NT-proBNP. It was observed that the MOD group had a higher NT-proBNP serum level than the CON group (Fig. A). Comparing the MOD groups to the CON group, the MOD groups showed lower levels of LVEF and LVFS (Fig. B and C), and the MOD group’s M-mode echocardiogram showed impaired cardiac performance (Fig. F). MOD cardiomyocytes were observed by HE staining to be enlarged, irregularly shaped, with a disordered arrangement; the interstitial space between the cells was also filled with fibrous tissue and heavily infiltrated with inflammatory cells (Fig. G). CRP and IL-1β are cytokines used to identify inflammation . The MOD group displayed elevated serum CRP and IL-1β levels compared to controls (Fig. D and E). These observations corroborated the H-HF model , indicating that establishing the H-HF rat model had succeeded. Administering SMI to H-HF rats reduced the elevation of NT-proBNP, CRP, and IL-1β levels. The SMI treatment restored decreased levels of LVEF and LVFS in the MOD group (Fig. B and C). The cardiac functions were also restored by the administration of SMI, as manifested by M-mode echocardiogram (Fig. F) and HE staining (Fig. G). HE staining of colonic tissue indicated a reduction in mucosal integrity and increased inflammatory cells in the MOD group (Fig. I). This impairment of mucosal functions was reversed with SMI treatment. The presence of Lipopolysaccharide (LPS) is indicative of damage to the intestinal mucosa , and Zonulin is used to assess intestinal permeability . In H-HF rats, the MOD group showed noticeably higher serum concentrations of LPS and Zonulin than the control group, indicating a breakdown of the intestinal mucosal barrier and increased intestinal permeability (Fig. H and J). LPS and Zonulin levels were lower in the SMI group compared to the MOD group, suggesting that SMI effectively improved intestinal permeability and intestinal barrier function. Three compounds were collected from Red Ginseng (Hong Shen) and ten compounds were obtained from Ophiopogon japonicus (Mai Dong) (Table ). Based on searches of the GeneCards and OMIM disease databases, 4158 heart failure (HF)-related disease targets and 122 overlapping targets were discovered (Fig. A). TNF, IL-6, IL-1β, AKT1, STAT3, NFκΒ, IFNG, IL-10, TP53, and TLR4 were identified as the primary targets by PPI protein interaction analysis (Fig. B). Figure C illustrates the findings of the “drug-component-disease-target” network. As suggested by KEGG analysis, the TNF, IL-17, and Toll-like receptor signaling pathways may be involved in the mechanism through which SMI prevents and treats HF (Fig. D). Apoptotic processes, inflammatory responses, response to external biotic stimuli, negative regulation of cell proliferation, negative regulation of the apoptotic process, cellular response to lipopolysaccharide, G protein-coupled receptor signaling pathway, and negative regulation of gene expression are among the main biological processes predicted by GO analysis and included 474 significantly enriched biological function entries for treating heart failure (Fig. E). There are 52 entries related to cellular components (CC), involving the plasma membrane, membrane, cytoplasm, extracellular space, extracellular region, extracellular exosome, cell surface, mitochondrion, endoplasmic reticulum membrane, and endoplasmic reticulum. Additionally, there are 80 entries related to molecular functions, involving protein binding, identical protein binding, enzyme binding, protein homodimerization activity, DNA binding, zinc ion binding, heme binding, signaling receptor activity, sequence-specific DNA binding, and receptor binding. Sequencing analysis of gut microbiota The sequencing of 18 fecal samples yielded 1,440,437 raw reads, which were merged and filtered to produce 1,408,229 clean tags. On average, 67,749 clean tags were obtained. To determine whether the sequencing data adequately reflected the diversity of species in the sample, a rarefaction curve was employed. Overall consistency in the results revealed that the sequencing data was adequate (Fig. A). A Venn diagram depicting the OTU distributions was shown in Fig. B. Across the three groups, 607 OTUs were identified, with 518 being shared by all of them. Alpha diversity analysis was carried out to assess the disparities in the structural complexity of the gut microbiota. Chao 1 and Shannon indices did not uncover any significant discrepancies in diversity across the three groups (Fig. C, D). However, a distinct divergence of profiles was found between the CON, MOD, and SM groups according to weighted unifrac PCoA of beta diversity (Fig. C). ANOSIM analysis (ANOSIM: R = 0.732, p = 0.001) demonstrated that the three groups were distinctly segregated. The proximity of the CON and SM populations indicated that their gut bacteria profiles were similar. The points representing the MOD group were further away from the points of the CON and SM groups, implying that the MOD group’s colony structure was markedly different from the other two groups. The PCoA outcomes (model stress = 0.0959 < 0.2) were supported by the NMDS analysis (Fig. D). Composition of gut microbiota and its difference analysis The microbial community composition was evaluated at the phylum and genus level (Fig. E and F). Results revealed that the MOD group had a reduced abundance of Bacteroidetes, Patescibacteria, Spirochaetes, and Elusimicrobia, and an elevated proportion of Firmicutes and Proteobacteria in comparison to the controls (Fig. A). As shown in Fig. A(g), the F/B ratio of the MOD group was substantially higher in comparison to that of the other group. The MOD group was noticed to be deficient in Muribaculaceae and Lachnospiraceae_NK4A136_group in comparison to the controls while having an increased abundance of Romboutsia and Ruminococcaceae at the genus level. These findings suggest that H-HF rats had an alteration in their gut bacterial equilibrium, which SMI treatment partially reversed (Fig. A). The heatmaps in Fig. B further illustrate the variations in the gut microbiota between the three groups. Prediction of the function of the gut microbiota By utilizing PICRUSt, a KEGG pathway analysis technique, we were able to evaluate the functional composition of the bacterial communities in the metagenome. All functional genes were shown at level III (Fig. ). The genes associated with energy and amino acid metabolism were more abundant in the MOD group, indicating that these metabolic pathways were disturbed in H-HF. Furthermore, after receiving SMI treatment, several genes related to energy and amino acid metabolism were altered (Fig. ). The sequencing of 18 fecal samples yielded 1,440,437 raw reads, which were merged and filtered to produce 1,408,229 clean tags. On average, 67,749 clean tags were obtained. To determine whether the sequencing data adequately reflected the diversity of species in the sample, a rarefaction curve was employed. Overall consistency in the results revealed that the sequencing data was adequate (Fig. A). A Venn diagram depicting the OTU distributions was shown in Fig. B. Across the three groups, 607 OTUs were identified, with 518 being shared by all of them. Alpha diversity analysis was carried out to assess the disparities in the structural complexity of the gut microbiota. Chao 1 and Shannon indices did not uncover any significant discrepancies in diversity across the three groups (Fig. C, D). However, a distinct divergence of profiles was found between the CON, MOD, and SM groups according to weighted unifrac PCoA of beta diversity (Fig. C). ANOSIM analysis (ANOSIM: R = 0.732, p = 0.001) demonstrated that the three groups were distinctly segregated. The proximity of the CON and SM populations indicated that their gut bacteria profiles were similar. The points representing the MOD group were further away from the points of the CON and SM groups, implying that the MOD group’s colony structure was markedly different from the other two groups. The PCoA outcomes (model stress = 0.0959 < 0.2) were supported by the NMDS analysis (Fig. D). The microbial community composition was evaluated at the phylum and genus level (Fig. E and F). Results revealed that the MOD group had a reduced abundance of Bacteroidetes, Patescibacteria, Spirochaetes, and Elusimicrobia, and an elevated proportion of Firmicutes and Proteobacteria in comparison to the controls (Fig. A). As shown in Fig. A(g), the F/B ratio of the MOD group was substantially higher in comparison to that of the other group. The MOD group was noticed to be deficient in Muribaculaceae and Lachnospiraceae_NK4A136_group in comparison to the controls while having an increased abundance of Romboutsia and Ruminococcaceae at the genus level. These findings suggest that H-HF rats had an alteration in their gut bacterial equilibrium, which SMI treatment partially reversed (Fig. A). The heatmaps in Fig. B further illustrate the variations in the gut microbiota between the three groups. By utilizing PICRUSt, a KEGG pathway analysis technique, we were able to evaluate the functional composition of the bacterial communities in the metagenome. All functional genes were shown at level III (Fig. ). The genes associated with energy and amino acid metabolism were more abundant in the MOD group, indicating that these metabolic pathways were disturbed in H-HF. Furthermore, after receiving SMI treatment, several genes related to energy and amino acid metabolism were altered (Fig. ). Regulation of SMI on the differential metabolites Principal component analysis (PCA) and orthogonal partial least square discriminate analysis (OPLS-DA), which scaled and log-translated the data to reduce noise and high variance effects, were successful in differentiating between the three groups. PCA demonstrated a clear distinction between the three groups, with the SM group being closer to the CON group than the MOD group (Fig. A and B). OPLS-DA analysis verified the distinctiveness among the three groups identified by PCA (Fig. ). Metabolic analysis revealed that 17 metabolites had significantly altered levels in positive mode and other 17 metabolites in negative ion mode, while 29 biomarkers were significantly restored after SMI treatment (Table , Fig. C-D). Our results revealed that these metabolites were linked to energy, methylamine, bile acid, and amino acid metabolisms (Table ). Metorgin tracing analysis The analysis of source-based metabolic function and metabolite traceability found 29 differential metabolites linked to SMI: 6 bacterial metabolites, 1 host-specific metabolite, and 24 bacteria-host cometabolites (Fig. A-B). According to metabolites pathway enrichment analysis (MPEA), the databases for the host, bacterial, and co-metabolism metabolic pathways were paired with 1, 4, and 28 relevant metabolic pathways, respectively (Fig. C-D). Of these pathways, 1, 1, and 10 revealed a significant ( p < 0.05) association with SMI. The origin-based functional analysis revealed that the microbial community was specific to phenylalanine metabolism and that the host was specific to primary bile acid biosynthesis. Histidine metabolism, tryptophan metabolism, valine, leucine, and isoleucine biosynthesis, beta-Alanine metabolism, butanoate metabolism, inositol phosphate metabolism, ascorbate and aldarate metabolism, nicotinate and nicotinamide metabolism, aminoacyl-tRNA biosynthesis and pentose and glucuronate interconversions were pathways of co-metabolism between microbes and hosts. The primary mechanism linked to SMI was histidine metabolism. A Bio-Sankey network based on MetOrigin analysis further visualized the biological relationships and statistical correlations between microbiota and metabolites to better depict the co-metabolic relationships between microbiota and hosts (Fig. A-B). Quantification of serum TMAO Studies have demonstrated a positive association between TMAO levels and cardiovascular conditions . In comparison to the controls, the serum levels of TMAO and TMA levels in the MOD group were considerably higher (Fig. A–B), which supports earlier findings . The drops in serum levels of TMAO and TMA after SMI treatment were not statistically significant, which may have been due to the small sample size. Principal component analysis (PCA) and orthogonal partial least square discriminate analysis (OPLS-DA), which scaled and log-translated the data to reduce noise and high variance effects, were successful in differentiating between the three groups. PCA demonstrated a clear distinction between the three groups, with the SM group being closer to the CON group than the MOD group (Fig. A and B). OPLS-DA analysis verified the distinctiveness among the three groups identified by PCA (Fig. ). Metabolic analysis revealed that 17 metabolites had significantly altered levels in positive mode and other 17 metabolites in negative ion mode, while 29 biomarkers were significantly restored after SMI treatment (Table , Fig. C-D). Our results revealed that these metabolites were linked to energy, methylamine, bile acid, and amino acid metabolisms (Table ). The analysis of source-based metabolic function and metabolite traceability found 29 differential metabolites linked to SMI: 6 bacterial metabolites, 1 host-specific metabolite, and 24 bacteria-host cometabolites (Fig. A-B). According to metabolites pathway enrichment analysis (MPEA), the databases for the host, bacterial, and co-metabolism metabolic pathways were paired with 1, 4, and 28 relevant metabolic pathways, respectively (Fig. C-D). Of these pathways, 1, 1, and 10 revealed a significant ( p < 0.05) association with SMI. The origin-based functional analysis revealed that the microbial community was specific to phenylalanine metabolism and that the host was specific to primary bile acid biosynthesis. Histidine metabolism, tryptophan metabolism, valine, leucine, and isoleucine biosynthesis, beta-Alanine metabolism, butanoate metabolism, inositol phosphate metabolism, ascorbate and aldarate metabolism, nicotinate and nicotinamide metabolism, aminoacyl-tRNA biosynthesis and pentose and glucuronate interconversions were pathways of co-metabolism between microbes and hosts. The primary mechanism linked to SMI was histidine metabolism. A Bio-Sankey network based on MetOrigin analysis further visualized the biological relationships and statistical correlations between microbiota and metabolites to better depict the co-metabolic relationships between microbiota and hosts (Fig. A-B). Studies have demonstrated a positive association between TMAO levels and cardiovascular conditions . In comparison to the controls, the serum levels of TMAO and TMA levels in the MOD group were considerably higher (Fig. A–B), which supports earlier findings . The drops in serum levels of TMAO and TMA after SMI treatment were not statistically significant, which may have been due to the small sample size. Correlations between the gut microbiota and fecal metabolic phenotype The relationship between metabolites and gut genera was evaluated by the Spearman correlation coefficient. A strong correlation is indicated by a value of r greater than 0.7. Figure C displays the network diagram with strong correlations. TMAO was strongly positively correlated with Elusimicrobium , _xylanophilum_group , oxidoreducens_group and negatively correlated with Catabacter , Defluviitaleaceae_UCG-011 , Parvibacter , _f_Atopobiaceae , Peptococcus , Coriobacteriaceae_UCG-002 , Staphylococcus , and Romboutsi . Therefore, methylamine metabolism must be impacted by the gut bacteria mentioned above. Similarly, D-glucuronic acid and D-xylitol correlated positively with the xylanophilum group , whereas creatinine correlated negatively with Coriobacteriaceae_UCG-002 . These findings suggest that Coriobacteriaceae_UCG-002 , Dubosiella , and the xylanophilum_group have an impact on energy metabolism. Bile acid metabolites (including chenodeoxycholic acid, cholic acid, and glycocholic acid) and amino acids (such as norvaline and L-threonine) have also been found to have a strong correlation with gut microbe composition. Our correlation data demonstrated modifications to the gut microbiome, resulting in a significantly altered metabolomic profile. Therefore, our current findings suggest that the mechanism by which SMI can improve heart function in an H-HF rat model may include effects on microbial energy, methylamine, bile acid, and amino acid metabolism in the intestine. Correlations between TMAO and gut microbiota Tables and present the results of Spearman’s correlation analysis used to evaluate the associations between the composition of the gut and the levels of TMAO metabolites. A positive correlation was uncovered between serum TMAO levels and the proportion of Actinobacteria. In contrast, a negative correlation was observed with Elusimicrobia at the phylum level (Fig. D). The analysis showed that serum TMAO had a direct relationship with Romboutsia , and an inverse relationship with Ruminococcaceae_UCG_014 and _f_Muribaculaceae . Furthermore, serum TMA levels had a strong negative correlation with Ruminococcaceae_UCG_014 (Fig. E). Based on these results, it is possible that SMI administration could alter TMAO levels by affecting the relevant microflora. Correlations between differential metabolites and targets of SMI Figure shows the connections between the targets of SMI and differential metabolites. Additionally, the regulatory role of SMI in preventing heart failure is highlighted by the “components-targets-metabolites-microbes” interaction network depicted in Fig. . By controlling for 46 proteins, network integration analysis shows that the 11 potentially active components of SMI can affect the differential expression of 8 metabolites and 24 gut microbes. The relationship between metabolites and gut genera was evaluated by the Spearman correlation coefficient. A strong correlation is indicated by a value of r greater than 0.7. Figure C displays the network diagram with strong correlations. TMAO was strongly positively correlated with Elusimicrobium , _xylanophilum_group , oxidoreducens_group and negatively correlated with Catabacter , Defluviitaleaceae_UCG-011 , Parvibacter , _f_Atopobiaceae , Peptococcus , Coriobacteriaceae_UCG-002 , Staphylococcus , and Romboutsi . Therefore, methylamine metabolism must be impacted by the gut bacteria mentioned above. Similarly, D-glucuronic acid and D-xylitol correlated positively with the xylanophilum group , whereas creatinine correlated negatively with Coriobacteriaceae_UCG-002 . These findings suggest that Coriobacteriaceae_UCG-002 , Dubosiella , and the xylanophilum_group have an impact on energy metabolism. Bile acid metabolites (including chenodeoxycholic acid, cholic acid, and glycocholic acid) and amino acids (such as norvaline and L-threonine) have also been found to have a strong correlation with gut microbe composition. Our correlation data demonstrated modifications to the gut microbiome, resulting in a significantly altered metabolomic profile. Therefore, our current findings suggest that the mechanism by which SMI can improve heart function in an H-HF rat model may include effects on microbial energy, methylamine, bile acid, and amino acid metabolism in the intestine. Tables and present the results of Spearman’s correlation analysis used to evaluate the associations between the composition of the gut and the levels of TMAO metabolites. A positive correlation was uncovered between serum TMAO levels and the proportion of Actinobacteria. In contrast, a negative correlation was observed with Elusimicrobia at the phylum level (Fig. D). The analysis showed that serum TMAO had a direct relationship with Romboutsia , and an inverse relationship with Ruminococcaceae_UCG_014 and _f_Muribaculaceae . Furthermore, serum TMA levels had a strong negative correlation with Ruminococcaceae_UCG_014 (Fig. E). Based on these results, it is possible that SMI administration could alter TMAO levels by affecting the relevant microflora. Figure shows the connections between the targets of SMI and differential metabolites. Additionally, the regulatory role of SMI in preventing heart failure is highlighted by the “components-targets-metabolites-microbes” interaction network depicted in Fig. . By controlling for 46 proteins, network integration analysis shows that the 11 potentially active components of SMI can affect the differential expression of 8 metabolites and 24 gut microbes. In recent years, there has been an increasing amount of research on the connection between alterations in the gut microbiota and metabolites and the onset of heart failure . However, the mechanisms by which SMI affects chronic heart failure from this perspective remain largely unknown. This study employs metabolomics, 16S rRNA high-throughput sequencing, and network pharmacology to investigate the influence of Shenmai injection on gut microbiota and metabolites in hypertensive heart failure rats. Moreover, the MetaboAnalyst platform was employed to clarify the connection between metabolites and targets, while the MetOrigin platform was used to examine the origin and function of metabolites. To establish a comprehensive analysis of the systematic relationships between the components, targets, metabolites, and gut microbiota influenced by SMI, a “component-target-metabolite-microbiota” interaction network was constructed. This provided new information about the mechanisms underlying SMI in heart failure therapy. SMI is recognized for its effects of invigorating Qi to prevent collapse, nourishing Yin, and promoting saliva production. Both red ginseng and Radix Ophiopogonis have been shown in basic experiments and clinical studies to possess immune-regulating, blood circulation-improving, antioxidant, anti-inflammatory, and anticancer properties . Radix Ophiopogonis has also been shown to exhibit anti-atherosclerotic effects. Research has shown that SMI has antioxidant properties and can reduce oxidative stress . Based on systematic review and meta-analysis, SMI has demonstrated efficacy in treating anthracycline-induced cardiotoxicity and is consequently a possible course of therapy for this condition . In this study, first, we evaluated cardiac function indicators, myocardial tissue HE staining, echocardiographic parameters (LVFS and LVEF), serum NT-proBNP levels, and inflammatory markers CRP and IL-1β to establish that SMI significantly improves cardiac function. In addition, SMI may be able to improve gut barrier function and decrease intestinal permeability based on its effects on intestinal permeability marker Zonulin and gut barrier function indicators (LPS). Indicating regulating the homeostasis of gut microbiota could be one of the primary ways that SMI enhances cardiac function. We evaluated alterations in the composition and functionality of gut microbiota in salt-sensitive hypertensive heart failure rats to learn more about the effect of SMI on gut microbiota. The gut microbiome profiles of the CON and MOD groups differed significantly, as demonstrated by our results, suggesting that the H-HF modeling had changed the microbial structure thus offering insight into how SMI affected the gut microbiota of the H-HF model. The SMI administration successfully revived the microbiota’s structure and functions in H-HF rats. It has been observed that an augmented proportion of Proteobacteria is a potential indicator of epithelial dysfunction and can also be used to diagnose gut dysbiosis and associated health risks . According to our research, the proportions of Proteobacteria in H-HF rats were successfully decreased by SMI treatment, bringing the F/B ratio back to par with the CON group. This suggests that SMI has a positive effect on reestablishing the equilibrium of intestinal flora. In addition, a previous study found that the gut microbiome of patients with chronic heart failure included fewer butyrate-producing bacteria . Butyrate and SCFA are known to be produced by bacteria in the Lachnospiraceae family . Research has demonstrated a correlation between Muribaculaceae and propionic acid levels, an indicator of SCFA concentration . Our research detected that the proportion of Lachnospiraceae_NK4A136 and _ f_Muribaculaceae augmented after SMI treatment, suggesting that SMI treatment reinstates bacteria that generate SCFAs. Assessing co-metabolic relationships between the host and gut microbiota can provide fresh perspectives on the critical function of the gut microbiome in host health . Combining 16S high-throughput sequencing with metabolomics provides a powerful approach to exploring the mechanisms underlying disease development. Compared to omics approaches that employ biofluids, such as urine and serum, the fecal metabolome offers a more comprehensive view as it reflects the combined effects of genetic, environmental, and dietary factors . Microbiome sequencing can therefore be used to understand the relationships between bacterial populations by using non-targeted metabolomics analysis of fecal samples. In the H-HF rat model, 34 metabolites had been significantly altered; SMI treatment restored 29 of these biomarkers. Our study demonstrates that Shenmai injection can significantly improve metabolic disorders in hypertensive heart failure rats. To determine if the host or the microbial community is the source of differential metabolites, we employed MetOrigin. Numerous identified metabolites participate in co-metabolism activities shared between the host and its resident gut microbiota. Energy metabolism, amino acid metabolism, methylamine metabolism, and bile acid metabolism are the key metabolic pathways engaged in these processes. It is well-established that energy metabolism is disturbed in heart failure (HF), and modulating cardiac energy metabolism has been proposed as a therapeutic strategy for HF .Recent studies have indicated that energy metabolism dysfunction plays a critical role in the pathophysiology of HF, with alterations in metabolic pathways contributing to the progression of the disease . Our study further supports this notion, as the metabolites associated with energy metabolism, such as gamma-aminobutyric acid, glutaric acid, D-glucuronic acid, 2-hydroxybutyric acid, and creatinine, were significantly lower in the MOD group. These findings align with previous research showing that impaired cardiac energetics is a major contributor to HF . Moreover, the 16S functional prediction analysis demonstrated that H-HF had a notable association with energy metabolism, consistent with other studies linking metabolic disturbances to HF. In particular, studies have highlighted how altered energy metabolism in HF affects mitochondrial function and cellular ATP production, contributing to cardiac dysfunction . Our correlation analysis further demonstrated that metabolites like creatinine were inversely associated with Coriobacteriaceae_UCG-002 , while D-glucuronic acid and D-xylitol showed significant positive correlations with the [Eubacterium]_xylanophilum_group but were negatively related to Coriobacteriaceae_UCG-002 . These results suggest a close relationship between gut microbiota and energy metabolism in H-HF, which is consistent with emerging evidence on the interplay between gut microbiota and metabolic disturbances in cardiovascular diseases . Notably, the levels of key bacteria such as Coriobacteriaceae_UCG-002 and [Eubacterium]_xylanophilum_group returned to normal following SMI treatment, accompanied by an increase in metabolites linked to energy metabolism. This suggests that the therapeutic mechanism of SMI may involve the regulation of both microbiota and metabolites related to energy metabolism, an idea supported by similar findings in other studies on TCM and its effects on metabolic regulation.Metabolizing amino acids is indispensable to the energy supply, as it facilitates the conversion of amino acids into glucose through gluconeogenesis. Studies have shown that the administration of amino acids can be advantageous to people with HF, with improvements seen in various clinical endpoints . Our study revealed a substantial decline in the metabolism of several amino acids (e.g., norvaline, ketoleucine, L-threonine, L-Valine, and N-acetylornithine) in the MOD group. Additionally, 16 S functional prediction was linked to amino acid metabolism, implying that H-HF is associated with a disruption in amino acid metabolism. The outcome of the correlation analysis indicated a close link between gut flora and metabolites related to amino acid metabolism; for example, norvaline demonstrated a strong and negative correlation with uncultured_bacterium_f_Ruminococcaceae , Adlercreutzia , Streptococcus , Faecalibaculum , [Eubacterium]_brachy_group , Dubosiella , uncultured_bacterium_f_Atopobiaceae , Coriobacteriaceae_UCG-002 , and UBA1819 . This study revealed that SMI treatment could adjust the relevant gut microbiota and amino acid metabolism. Bile acids have been shown in earlier studies to be integral to managing metabolism and energy expenditure . Moreover, a study found that patients with HF had a higher ratio of secondary to primary bile acids in their plasma and lower levels of primary bile acids . According to the correlation analysis conducted here, SMI may enhance the gut microbiota and metabolites linked to bile acid metabolism. The interactions between differential metabolites and gut microbiota are complex and multifactorial, with each influencing the other in a dynamic manner. Our study provides evidence that gut microbiota plays a pivotal role in modulating the metabolism of key metabolites involved in energy production, amino acid metabolism, and bile acid metabolism. Moreover, the therapeutic effects of SMI seem to be partly mediated by its ability to regulate these microbiota-related metabolites, thus restoring metabolic homeostasis and improving heart function in H-HF. Microbial homeostasis is defined as the maintenance of a balanced composition of gut microbiota in a healthy state . Disruption of this equilibrium, however, can lead to the proliferation of pathogenic microorganisms, which raises serum concentrations of TMA and TMAO and increases the risk of cardiovascular diseases . TMAO is considered a risk factor for cardiovascular disease as it is found in high concentrations in the blood when the intestinal wall is disrupted . Measuring serum TMAO levels has consequently emerged as a crucial marker of cardiovascular risk . Our study revealed that the MOD group had lower fecal TMAO levels than the control group. Nevertheless, data from targeted metabolomics indicated that the MOD group had significantly higher serum TMAO and TMA concentrations than any other groups, a result that is consistent with numerous earlier studies on H-HF. A potential cause for the disparity could be the use of different sample types, such as fecal samples instead of serum samples. The reason for the decrease in TMAO in feces is thought to be related to the elevated TMAO levels in serum. According to studies by Nagatomo et al., elevated TMAO levels may cause myocardial fibrosis, LVEF reduction, multi-organ fibrosis, and an increase in BNP levels, all of which can contribute to heart failure . Moreover, it has been noted that raised serum TMAO levels are linked to an increased risk of heart failure and its associated mortality . A study found that TMAO combined with NT-proBNP were useful prognostic indicators for heart failure in patients . This study revealed that the intervention involving SMI had a slight, albeit not statistically relevant, impact on the decrease of serum TMAO and TMA levels. Correlation analysis results indicate that SMI significantly reduces gut microbiota associated with TMAO and TMA, suggesting that SMI may influence serum TMAO levels by modulating these related microbial communities. Thirteen of SMI’s active ingredients were screened based on the TCMSP and BATMAN-TCM database. Network integration analysis showed that by targeting 46 proteins, the 11 potentially active components of SMI can affect the differential expression of 8 metabolites and 24 gut microbes. Taken together, these investigations revealed that various SMI components work synergistically to exert their therapeutic function. This study has some limitations. Firstly, while 13 active components of SMI were identified through database screening, their effectiveness was not experimentally validated. Future research could verify these active ingredients through pharmacokinetic experiments. Secondly, while the gut microbiota and some metabolites showed a strong correlation in this study, the associations do not always imply causation. Furthermore, the distinct levels of TMAO observed between fecal and serum samples suggest a potential tissue-specific role of TMAO. Therefore, additional experiments are warranted to investigate the role of TMAO in different tissues or organs in heart failure. Our study offers a thorough investigation into the mechanisms through which SMI generates therapeutic benefits in heart failure. We observed that in hypertensive heart failure rats, SMI dramatically improves gut barrier function, cardiac function, and gut microbiota composition. By reestablishing homeostasis in the gut microbiota, SMI drives vital metabolic pathways such as energy metabolism, amino acid metabolism, and bile acid metabolism, as indicated by metabolomics and 16S rRNA sequencing analyses. Higher serum TMAO levels were found to be a risk factor for H-HF using TMAO-targeted metabolomics analysis, and SMI was able to downregulate these levels of TMAO-related metabolites. Network pharmacology analysis identified 13 active components of SMI targeting 46 proteins, resulting in differential expression changes in 8 metabolites and 24 gut microbes. This study highlights the effectiveness of SMI in alleviating H-HF and its potential to modulate microbial-host co-metabolism, underscoring the synergistic actions of multiple SMI components on various biological pathways implicated in heart failure. Future research should focus on validating these observations in clinical settings and elucidating the specific molecular mechanisms underlying SMI’s therapeutic benefits. Below is the link to the electronic supplementary material. Supplementary Material 1: Figure S1. (A) The chromatogram of Ginsenoside Rg1, Re, Rb1 reference solution. (B) The chromatogram of Shenmai injection test solution. Supplementary Material 2: Figure S2. (A) The PCA score of three groups and QC samples in positive-ion mode. (B) The PCA score of three groups and QC samples in negative-ion mode. (C) Chao1 index. (D) Shannon index. ## Data were represented as the mean ± SD Supplementary Material 3: Figure S3. (A) systolic blood pressure (SBP). (B) diastolic blood pressure (DBP). (C) SBP before and after treatment. (D) DBP before and after treatment. (E) Body weight prior to and after treatment. No remarkable alteration was noticed in the same group prior to and after treatment. Supplementary Material 4: Figure S4. PCA analysis of metabolites profile. CON: turquoise; MOD: red; SM: dark blue. (A) PCA score plots for positive-ion mode (R 2 X = 0.680). (B) PCA score plots for negative-ion mode (R 2 X = 0.674). (C) Heat map of the differential metabolites in positive-ion mode. (D) Heat map of the differential metabolites in negative-ion mode. Supplementary Material 5: Figure S5. OPLS-DA analysis of metabolites profile. (A) Comparison plots in positive-ion mode for CON and MOD groups (R 2 X=0.606, R 2 Y=0.998, Q 2 =0.974). (B) Permutation test for comparison between CON and MOD groups in positive-ion mode (n=200). (C) Comparison between CON and MOD groups in negative-ion mode (R 2 X=0.573, R 2 Y=0.999, Q 2 =0.974). (D) Permutation test for comparison between CON and MOD groups in positive-ion mode (n= 200). (E) Comparison between CON and MOD groups in positive-ion mode (R 2 X=0.513, R 2 Y=0.995, Q 2 =0.954). (F) Permutation test for comparison between CON and MOD groups in positive-ion mode (n= 200). (G) Comparison between MOD and SM group in negative-ion mode (R 2 X=0.532, R 2 Y=0.997, Q 2 =0.956). (H) Permutation test for comparison between MOD and SM groups in positive-ion mode (n= 200). Supplementary Material 6: Table S1: The related dataset of spearman’s correlation at the phylum level. Supplementary Material 7: Table S2: The related dataset of spearman’s correlation at the genus level. Supplementary Material 8 Supplementary Material 9 |
An Introduction to Artificial Intelligence in Developmental and Behavioral Pediatrics | cc465f95-ad2a-45f1-98ef-b31720003914 | 9907689 | Pediatrics[mh] | First coined – in 1956 by John McCarthy, artificial intelligence (AI) is an interdisciplinary field of computer science that involves the use of computers to develop systems able to perform tasks that are generally associated with intelligence in the intuitive sense. The rise of “big data,” alongside the development of increasingly complex algorithms and enhanced computational power and storage capabilities, has contributed to the recent surge in AI-based technologies. AI operates on a continuum, variously assisting, augmenting, or autonomizing task performance. Automating repetitive tasks, for example, may only require an “assisted” form of intelligence. On the other end of the human-machine continuum, however, an “autonomous” form of intelligence is required for machines to independently make decisions in adaptive intelligent systems. Within the health care context, AI is not envisioned as a technology that would supersede the need for skilled human clinicians. Rather, AI-based technologies will likely play pivotal roles in augmenting existing diagnostic and therapeutic toolkits to improve outcomes. While pediatricians may have limited familiarity with AI in a health care context, they are likely already making use of AI-powered technologies in their daily lives. Email spam filters, e-commerce platforms, and entertainment recommendation systems, for example, all rely on AI. Machine Learning: Underlying Principles Machine learning (ML) refers specifically to AI methodologies that incorporate an adaptive element wherein systems have the ability to “learn” using data to improve overall accuracy. At a basic level, all ML involves an input, a function (or some mathematical calculations), and an output. In ML models, the independent variables are termed inputs or features (e.g., age, gender, medical history, clinical symptom) and the dependent variable is referred to as the output label or target variable (e.g., diagnostic label, disease level, survival time). Both structured and unstructured data can be used to train ML models (see Fig. ). While statistical and ML techniques overlap, they are distinguishable by their underlying goals. Statistical learning is usually hypothesis-driven with the goal of inferring relationships between variables. ML is method-driven with the goal of building a model that makes actionable predictions. Machine Learning Workflow Clinicians familiar with the development and validation of existing behavioral screening and diagnostic tools will already have a general sense of product development workflow. Specific to the ML workflow, however, is an ability to learn from exposure to data, without the level of prespecified instructions or prior assumptions required in traditional product development. The likelihood of discovering new features and associations is therefore higher because workflow does not require the product designer to determine in advance which variables may be important. When building predictive ML models, the full data set must first be split into parts. Typically, the largest of these parts is the “training set” used for initial model training. A smaller portion of the data, the “validation set,” is then used to support hyperparameter tuning and model selection. Ultimately, the data sets help “train” the system to learn from similar patients, clinical features, or outcomes, which helps the algorithm become more accurate over time as data increase and more tests are conducted. At the end of this iterative process, a “test” set may be used to test the model's generalizability to data that was not previously seen during any prior aspect of the model's development. Prospective clinical validation studies may then be conducted to test the real-world performance of the model on data that were not previously available to model developers. Figure illustrates a typical ML workflow. Machine Learning Approaches Data type, structure, and number of features, along with the nature of the clinical questions being explored, all inform the type of ML approach that may be taken. Classical ML, which includes supervised learning and unsupervised learning , is generally applied to less complex data sets and clinical scenarios with a small number of features. Table provides a brief descriptive summary of these approaches. Neural Networks and Deep Learning While the ML techniques described above are suitable for many clinical problems, in cases in which nonlinear and complex relationships need to be mapped, networks and deep learning techniques may be more appropriate. An artificial neural network is a complex form of ML model designed to mimic how the neurons in the brain work. This is achieved through multiple layers of aggregation nodes known as neurons because they simulate the function of biological neurons with mathematical functions guiding when each node neuron would fire a signal to another. Features can be multiplied or added together repeatedly. The mathematical formulation of neural networks is such that complex nonlinear relationships can be modeled efficiently before an output layer that depends on the prediction task. Given the complexity and volume of health care data, this technique is becoming increasingly popular. Figure illustrates both simple neural network and deep learning neural network. Evaluating Model Performance The ability to explain the output of a model and assess its performance, both in the clinical context and in relation to its intended purpose, will increasingly become part of the future clinician's role. Key performance metrics relevant to ML are summarized in Table . We should note that these metrics do not always provide a straightforward interpretation of the performance of a model because they can be biased by the nature of the data; the end result should be determined by whether clinical value can be obtained from the model. The model's accuracy and threshold, along with the disease prevalence and the model's performance compared with existing “gold-standard” non-AI–informed approaches, should all be considered. Machine learning (ML) refers specifically to AI methodologies that incorporate an adaptive element wherein systems have the ability to “learn” using data to improve overall accuracy. At a basic level, all ML involves an input, a function (or some mathematical calculations), and an output. In ML models, the independent variables are termed inputs or features (e.g., age, gender, medical history, clinical symptom) and the dependent variable is referred to as the output label or target variable (e.g., diagnostic label, disease level, survival time). Both structured and unstructured data can be used to train ML models (see Fig. ). While statistical and ML techniques overlap, they are distinguishable by their underlying goals. Statistical learning is usually hypothesis-driven with the goal of inferring relationships between variables. ML is method-driven with the goal of building a model that makes actionable predictions. Clinicians familiar with the development and validation of existing behavioral screening and diagnostic tools will already have a general sense of product development workflow. Specific to the ML workflow, however, is an ability to learn from exposure to data, without the level of prespecified instructions or prior assumptions required in traditional product development. The likelihood of discovering new features and associations is therefore higher because workflow does not require the product designer to determine in advance which variables may be important. When building predictive ML models, the full data set must first be split into parts. Typically, the largest of these parts is the “training set” used for initial model training. A smaller portion of the data, the “validation set,” is then used to support hyperparameter tuning and model selection. Ultimately, the data sets help “train” the system to learn from similar patients, clinical features, or outcomes, which helps the algorithm become more accurate over time as data increase and more tests are conducted. At the end of this iterative process, a “test” set may be used to test the model's generalizability to data that was not previously seen during any prior aspect of the model's development. Prospective clinical validation studies may then be conducted to test the real-world performance of the model on data that were not previously available to model developers. Figure illustrates a typical ML workflow. Data type, structure, and number of features, along with the nature of the clinical questions being explored, all inform the type of ML approach that may be taken. Classical ML, which includes supervised learning and unsupervised learning , is generally applied to less complex data sets and clinical scenarios with a small number of features. Table provides a brief descriptive summary of these approaches. Neural Networks and Deep Learning While the ML techniques described above are suitable for many clinical problems, in cases in which nonlinear and complex relationships need to be mapped, networks and deep learning techniques may be more appropriate. An artificial neural network is a complex form of ML model designed to mimic how the neurons in the brain work. This is achieved through multiple layers of aggregation nodes known as neurons because they simulate the function of biological neurons with mathematical functions guiding when each node neuron would fire a signal to another. Features can be multiplied or added together repeatedly. The mathematical formulation of neural networks is such that complex nonlinear relationships can be modeled efficiently before an output layer that depends on the prediction task. Given the complexity and volume of health care data, this technique is becoming increasingly popular. Figure illustrates both simple neural network and deep learning neural network. While the ML techniques described above are suitable for many clinical problems, in cases in which nonlinear and complex relationships need to be mapped, networks and deep learning techniques may be more appropriate. An artificial neural network is a complex form of ML model designed to mimic how the neurons in the brain work. This is achieved through multiple layers of aggregation nodes known as neurons because they simulate the function of biological neurons with mathematical functions guiding when each node neuron would fire a signal to another. Features can be multiplied or added together repeatedly. The mathematical formulation of neural networks is such that complex nonlinear relationships can be modeled efficiently before an output layer that depends on the prediction task. Given the complexity and volume of health care data, this technique is becoming increasingly popular. Figure illustrates both simple neural network and deep learning neural network. The ability to explain the output of a model and assess its performance, both in the clinical context and in relation to its intended purpose, will increasingly become part of the future clinician's role. Key performance metrics relevant to ML are summarized in Table . We should note that these metrics do not always provide a straightforward interpretation of the performance of a model because they can be biased by the nature of the data; the end result should be determined by whether clinical value can be obtained from the model. The model's accuracy and threshold, along with the disease prevalence and the model's performance compared with existing “gold-standard” non-AI–informed approaches, should all be considered. Opportunities In the field of developmental and behavioral pediatrics, artificial intelligence (AI) can assist with a broad array of tasks including diagnosis, risk prediction and stratification, treatment, administration, and regulation. Core analytic tasks that may fall under the AI umbrella are depicted in Figure . Enhancing Clinical Decision-Making, Risk Prediction, and Diagnosis Massive and constantly expanding quantities of medical data including electronic medical records (EMRs), high-resolution medical images, public health data sets, genomics, and wearables have exceeded the limits of human analysis. AI offers opportunities to harness and derive clinically meaningful insights from this ever-growing volume of health care data in ways that traditional analytic techniques cannot. Neural networks, for example, trained on much larger quantities of data than any single clinician could possibly be exposed to in the course of their career, can support the identification of subtle nonlinear data patterns. Such approaches promise to significantly enhance interpretation of medical scans, pathology slides, and other imaging data that rely on pattern recognition. By observing subtle nonlinear correlations in the data, , machine learning (ML) approaches have potential to augment risk prediction and diagnostic processes and ultimately provide an enhanced quality of care. A number of studies within the field of developmental and behavioral pediatrics demonstrate the potential for AI to enhance risk prediction practices. ML techniques were used, for example, to analyze the EMRs of over a million individuals to identify risk for Fragile X based on associations with comorbid medical conditions. The resulting predictive model was able to flag Fragile X cases 5 years earlier than current practice without relying on genetic or familial data. A similar approach was taken to predict autism spectrum disorder (ASD) risk based on high-prevalence comorbidity clusters detected in EMRs. Digital biomarkers inferred from deep comorbidity patterns have also been leveraged to develop an ASD comorbid risk score with a superior predictive performance than some questionnaire-based screening tools. Automated speech analysis was combined with ML in another study to accurately predict risk for psychosis onset in clinically vulnerable youths. In this proof-of-principle study, speech features from transcripts of interviews with at-risk youth were fed into a classification algorithm to assess their predictive value for psychosis. Research has also highlighted the potential for AI to streamline diagnostic pathways for conditions such as ASD and attention-deficit/hyperactivity disorder (ADHD). Such research is promising given that streamlining diagnosis could allow for earlier treatment initiation during the critical neurodevelopmental window. For example, researchers have used ML to examine behavioral phenotypes of children with ASD with a high rate of accuracy and to shorten the time for observation-based screening and diagnosis. – In addition, ML has been used to differentiate between ASD and ADHD with high accuracy using a small number of measured behaviors. Research has also explored facial expression analysis based on dynamic deep learning and 3D behavior analysis to detect and distinguish between ADHD, ASD, and comorbid ADHD/ASD presentations. Expanding Treatment Options Along with risk prediction and diagnostics, AI holds potential to enhance treatment delivery in the field of developmental and behavioral pediatrics. AI robots have been used in a number of intervention studies to enhance social skill acquisition and spontaneous language development in children with ASD. , The potential utility of AI-enabled technologies to treat disruptive behaviors and mood and anxiety disorders in children is also being explored. In the field of ADHD, a number of emerging technologies show promise to infer behaviors that can then be used to tailor feedback to enhance self-regulation. For example, in 1 study, a neural network was used to learn behavioral intervention delivery techniques based on human demonstrations. The trained network then enabled a robot to autonomously deliver a similar intervention to children with ADHD. As health data infrastructure expands in size and sophistication, future AI technologies could potentially support increasingly personalized treatment options. Once comprehensive and integrated biologic, anatomic, physiologic, environmental, socioeconomic, behavioral, and pharmacogenomic patient data become routinely available, AI-based nearest-neighbor analysis could be used to identify “digital twins.” Patients with similar genomic and clinical features could be identified and clustered, for example, to allow for highly targeted treatments. Such approaches, while currently largely theoretical, could help predict therapeutic and adverse medication responses more accurately and also form an evidence base for personalized treatment pathways. Streamlining Care and Enhancing Workplace Efficiency Artificial intelligence algorithms have potential to automate many arduous administrative tasks, thereby streamlining care pathways and freeing clinicians to spend more time with patients. Natural language processing solutions, for example, are being developed to decrease reliance on human scribes in clinical encounters. Such approaches may be used to process and transform clinical-free text into meaningful structured outcomes, automate some documentation practices through text summarization, and scan text-based reports to support accurate and rapid diagnostic recommendations. Natural language processing solutions may prove particularly valuable in fields such as child and adolescent psychiatry and developmental and behavioral pediatrics that are extremely text-heavy. Developmental behavioral assessments often involve text-heavy tasks such as documentation of complex patient histories, results of extensive testing, and collateral history from multiple informants. Given the current national US shortage of child and adolescent psychiatrists with a median of 11 psychiatrists per 100,000 children, AI-based approaches that streamline and automate administrative tasks seem particularly promising. AI-assisted image interpretation has also shown potential to increase workplace productivity and provide considerable cost savings over current practice. , Techniques to streamline medical research and drug discovery by using natural language processing to rapidly scan biomedical literature and data mine molecular structures are also being developed. Promoting Equity and Access Artificial intelligence has potential to address several bias and access disparities apparent in existing care models. While access to developmental and behavioral specialists is extremely limited in much of the world, it is estimated that over 50% of the global population has access to a smartphone. Digital AI–based diagnostic and treatment platforms could thus potentially expand access to underserved and geographically remote populations. Thoughtful use of AI may also help to address racial, socioeconomic, and gender biases. In the field of developmental and behavioral pediatrics, for example, it has been noted that despite ASD prevalence rates being roughly equal across racial/ethnic and socioeconomic groups, human clinicians are more likely to diagnose Black, Latinx, and Asian children, as well as children from low-income families, at a later date than White children and children with a higher socioeconomic status. By integrating and training on large racial-conscious and gender-conscious data sets, AI algorithms can assess thousands of traits and features and build on the findings to assist clinicians in making more accurate, timely, and less biased ASD diagnoses. Challenges While AI presents multiple opportunities to the field of developmental and behavioral pediatrics, to date, a very few AI-based technologies have been broadly integrated into clinical practice. Challenges to the widespread deployment of AI in health care settings, along with potential solutions, are outlined below. Data Bias Any AI algorithm is only as good as the data from which it was derived. Simply put, the performance of the algorithm and the quality of its prediction are dependent on the quality of the data supplied. If the data are biased, imbalanced, or otherwise an incomplete representation of the target group, the derivative model's generalizability will be limited. A class imbalance problem can occur, for example, in cases in which the total number of 1 class of data (i.e., “girls” or “disease positive status”) is far less than the total number of another class of data (i.e., “boys” or “disease negative status”). These biases can sometimes be identified and addressed through techniques such as over- or undersampling, but at other times with “black-box” learning, it is harder to detect and fix the bias in the algorithm(s). Representative populations for most conditions, including in pediatrics, are not a homogeneous group. Thus, it is essential that the data sets used to train these models include balanced data with diverse representations of the clinical symptoms across gender, age, and race in order for the model to be generalizable. Without such safeguards, models may, in fact, perpetuate or amplify stereotypes or biases. , Data Sharing and Privacy Concerns To robustly train an algorithm, sufficient data are required, yet pediatric data sets can be limited by small sample sizes, especially when split by age group. For the case of rare pediatric diseases, extremely low prevalence rates mean the amount and type of data available to train a model on very limited. Creative use of deep learning techniques such as generative adversarial networks may be required in such cases to counteract the lack of data. Generative adversarial networks pit one neural network against another for the purpose of generating synthetic (yet realistic) data to support a variety of tasks such as image and voice generation. While such techniques can be extremely useful when appropriately applied, overreliance on synthetic data also comes with its own set of risks. Multiple data sets may also be combined to produce a sufficient volume of data for model training. However, combining data sets presents its own set of challenges including data privacy and ownership issues and difficulties integrating data with heterogeneous features. Federated learning is an emerging ML technique with potential to address some of these data sharing and privacy concerns. Federated learning allows algorithms to train across many decentralized servers or edge devices, exchanging parameters (i.e., the models' weights and biases) without explicitly exchanging the data samples themselves. This technique obviates many of the privacy issues engendered by uploading highly sensitive health data from different sources onto a single server. Algorithmic Transparency and Explainability A lack of transparency in certain types of ML algorithms such as deep neural networks has raised concern about their clinical trustworthiness. Many models used to analyze images and text, for example, include levels of complexity and multidimensionality that exceed intuitive understanding or interpretation. In cases in which a clinician is unable to understand how the algorithm produces an output, should the algorithm be relied on as part of their clinical decision-making process? Such concerns have led to calls for algorithmic deconvolution before use in health care settings. Other researchers argue, however, that current approaches to explainability disregard the reality that local explanations can be unreliable or too superficial to be meaningful and that rigorous model validation before deployment may be a more important marker of trustworthiness. Consumer and Clinician Preparedness If patients and clinicians mistrust AI-based technologies, and/or lack sufficient training to understand, in broad terms, how they function, clinical adoption may be delayed. As with all new tools, implementation matters, and discipline is required to ensure safe deployment of AI-based devices without loss of clinician skill. Overreliance on AI-based imaging at the expense of history and physical examination, for example, should be avoided. Patient reservations that will require consideration include safety, cost, choice, data bias, and data security concerns. While the American Medical Association has called for research into how AI should be addressed in medical education, current medical training lacks a consistent approach to AI education, and key licensing examinations do not test on this content. Clinicians and medical students alike have identified knowledge gaps and reservations regarding the use of AI in health care. – A number of preliminary frameworks for integrating AI curriculum into medical training have been proposed , , ; however, additional research is urgently required to develop and then integrate standardized AI content into medical training pathways. Ethical Ambiguities From an ethical standpoint, users of this technology must consider the direct impact and unintended consequences of AI implementation in general as well as specific implications within the clinical context. A number of well-publicized non–health-related cases have illustrated AI data privacy concerns, along with the ethically problematic potential for AI to amplify social, racial, and gender biases. There are also ethical concerns around the magnitude of harm that could occur if an ML algorithm, deployed clinically at scale, were to malfunction; associated impacts could far exceed the harm caused by a single clinician's malpractice. The use of AI for clinical decision-making also raises questions of accountability, such as who is liable if unintended consequences result from use of the technology (e.g., missed diagnosis), or what course of action an autonomous therapy chatbot might take if it detects speech patterns indicative of risk for self-harm. Ethical AI frameworks addressing such concerns are under development, , and researchers are calling for AI technologies to undergo robust simulation, validation, and prospective scrutiny before clinical adoption. Regulatory and Payment Barriers Notwithstanding data, provider adoption, and ethical safeguards, AI technologies face several systematic challenges to be readily implemented into clinical practice, including regulatory, interoperability with EMRs and data exchange, and payment barriers. Given that AI devices can learn from data and alter their algorithms accordingly, traditional medical device regulatory frameworks might not be sufficient. As a result, the Food and Drug Administration has developed a proposed regulatory framework that includes a potential “Predetermined Change Control Plan” for premarket submissions, including “Software Pre-Specifications” and an “Algorithm Change Protocol,” to address the iterative nature of AI/ML-based Software as a Medical Device technologies. Health care organizations and practices will also need to establish a data infrastructure and privacy policy for data that are stored across multiple servers and sources (e.g., medical records, health sensors, medical devices, etc). Development of new digital medical software and devices that use AI are likely to outpace the current health care payment structure. New billing codes associated with new treatments and procedures require formal approvals by national organizations with subsequent adoption by insurances, both public and private. This process can take many months to years. To facilitate provider and patient adoption of new AI technologies which may improve quality of care, streamlined development of billing codes for technologies using AI should be developed. In the field of developmental and behavioral pediatrics, artificial intelligence (AI) can assist with a broad array of tasks including diagnosis, risk prediction and stratification, treatment, administration, and regulation. Core analytic tasks that may fall under the AI umbrella are depicted in Figure . Enhancing Clinical Decision-Making, Risk Prediction, and Diagnosis Massive and constantly expanding quantities of medical data including electronic medical records (EMRs), high-resolution medical images, public health data sets, genomics, and wearables have exceeded the limits of human analysis. AI offers opportunities to harness and derive clinically meaningful insights from this ever-growing volume of health care data in ways that traditional analytic techniques cannot. Neural networks, for example, trained on much larger quantities of data than any single clinician could possibly be exposed to in the course of their career, can support the identification of subtle nonlinear data patterns. Such approaches promise to significantly enhance interpretation of medical scans, pathology slides, and other imaging data that rely on pattern recognition. By observing subtle nonlinear correlations in the data, , machine learning (ML) approaches have potential to augment risk prediction and diagnostic processes and ultimately provide an enhanced quality of care. A number of studies within the field of developmental and behavioral pediatrics demonstrate the potential for AI to enhance risk prediction practices. ML techniques were used, for example, to analyze the EMRs of over a million individuals to identify risk for Fragile X based on associations with comorbid medical conditions. The resulting predictive model was able to flag Fragile X cases 5 years earlier than current practice without relying on genetic or familial data. A similar approach was taken to predict autism spectrum disorder (ASD) risk based on high-prevalence comorbidity clusters detected in EMRs. Digital biomarkers inferred from deep comorbidity patterns have also been leveraged to develop an ASD comorbid risk score with a superior predictive performance than some questionnaire-based screening tools. Automated speech analysis was combined with ML in another study to accurately predict risk for psychosis onset in clinically vulnerable youths. In this proof-of-principle study, speech features from transcripts of interviews with at-risk youth were fed into a classification algorithm to assess their predictive value for psychosis. Research has also highlighted the potential for AI to streamline diagnostic pathways for conditions such as ASD and attention-deficit/hyperactivity disorder (ADHD). Such research is promising given that streamlining diagnosis could allow for earlier treatment initiation during the critical neurodevelopmental window. For example, researchers have used ML to examine behavioral phenotypes of children with ASD with a high rate of accuracy and to shorten the time for observation-based screening and diagnosis. – In addition, ML has been used to differentiate between ASD and ADHD with high accuracy using a small number of measured behaviors. Research has also explored facial expression analysis based on dynamic deep learning and 3D behavior analysis to detect and distinguish between ADHD, ASD, and comorbid ADHD/ASD presentations. Expanding Treatment Options Along with risk prediction and diagnostics, AI holds potential to enhance treatment delivery in the field of developmental and behavioral pediatrics. AI robots have been used in a number of intervention studies to enhance social skill acquisition and spontaneous language development in children with ASD. , The potential utility of AI-enabled technologies to treat disruptive behaviors and mood and anxiety disorders in children is also being explored. In the field of ADHD, a number of emerging technologies show promise to infer behaviors that can then be used to tailor feedback to enhance self-regulation. For example, in 1 study, a neural network was used to learn behavioral intervention delivery techniques based on human demonstrations. The trained network then enabled a robot to autonomously deliver a similar intervention to children with ADHD. As health data infrastructure expands in size and sophistication, future AI technologies could potentially support increasingly personalized treatment options. Once comprehensive and integrated biologic, anatomic, physiologic, environmental, socioeconomic, behavioral, and pharmacogenomic patient data become routinely available, AI-based nearest-neighbor analysis could be used to identify “digital twins.” Patients with similar genomic and clinical features could be identified and clustered, for example, to allow for highly targeted treatments. Such approaches, while currently largely theoretical, could help predict therapeutic and adverse medication responses more accurately and also form an evidence base for personalized treatment pathways. Streamlining Care and Enhancing Workplace Efficiency Artificial intelligence algorithms have potential to automate many arduous administrative tasks, thereby streamlining care pathways and freeing clinicians to spend more time with patients. Natural language processing solutions, for example, are being developed to decrease reliance on human scribes in clinical encounters. Such approaches may be used to process and transform clinical-free text into meaningful structured outcomes, automate some documentation practices through text summarization, and scan text-based reports to support accurate and rapid diagnostic recommendations. Natural language processing solutions may prove particularly valuable in fields such as child and adolescent psychiatry and developmental and behavioral pediatrics that are extremely text-heavy. Developmental behavioral assessments often involve text-heavy tasks such as documentation of complex patient histories, results of extensive testing, and collateral history from multiple informants. Given the current national US shortage of child and adolescent psychiatrists with a median of 11 psychiatrists per 100,000 children, AI-based approaches that streamline and automate administrative tasks seem particularly promising. AI-assisted image interpretation has also shown potential to increase workplace productivity and provide considerable cost savings over current practice. , Techniques to streamline medical research and drug discovery by using natural language processing to rapidly scan biomedical literature and data mine molecular structures are also being developed. Promoting Equity and Access Artificial intelligence has potential to address several bias and access disparities apparent in existing care models. While access to developmental and behavioral specialists is extremely limited in much of the world, it is estimated that over 50% of the global population has access to a smartphone. Digital AI–based diagnostic and treatment platforms could thus potentially expand access to underserved and geographically remote populations. Thoughtful use of AI may also help to address racial, socioeconomic, and gender biases. In the field of developmental and behavioral pediatrics, for example, it has been noted that despite ASD prevalence rates being roughly equal across racial/ethnic and socioeconomic groups, human clinicians are more likely to diagnose Black, Latinx, and Asian children, as well as children from low-income families, at a later date than White children and children with a higher socioeconomic status. By integrating and training on large racial-conscious and gender-conscious data sets, AI algorithms can assess thousands of traits and features and build on the findings to assist clinicians in making more accurate, timely, and less biased ASD diagnoses. Massive and constantly expanding quantities of medical data including electronic medical records (EMRs), high-resolution medical images, public health data sets, genomics, and wearables have exceeded the limits of human analysis. AI offers opportunities to harness and derive clinically meaningful insights from this ever-growing volume of health care data in ways that traditional analytic techniques cannot. Neural networks, for example, trained on much larger quantities of data than any single clinician could possibly be exposed to in the course of their career, can support the identification of subtle nonlinear data patterns. Such approaches promise to significantly enhance interpretation of medical scans, pathology slides, and other imaging data that rely on pattern recognition. By observing subtle nonlinear correlations in the data, , machine learning (ML) approaches have potential to augment risk prediction and diagnostic processes and ultimately provide an enhanced quality of care. A number of studies within the field of developmental and behavioral pediatrics demonstrate the potential for AI to enhance risk prediction practices. ML techniques were used, for example, to analyze the EMRs of over a million individuals to identify risk for Fragile X based on associations with comorbid medical conditions. The resulting predictive model was able to flag Fragile X cases 5 years earlier than current practice without relying on genetic or familial data. A similar approach was taken to predict autism spectrum disorder (ASD) risk based on high-prevalence comorbidity clusters detected in EMRs. Digital biomarkers inferred from deep comorbidity patterns have also been leveraged to develop an ASD comorbid risk score with a superior predictive performance than some questionnaire-based screening tools. Automated speech analysis was combined with ML in another study to accurately predict risk for psychosis onset in clinically vulnerable youths. In this proof-of-principle study, speech features from transcripts of interviews with at-risk youth were fed into a classification algorithm to assess their predictive value for psychosis. Research has also highlighted the potential for AI to streamline diagnostic pathways for conditions such as ASD and attention-deficit/hyperactivity disorder (ADHD). Such research is promising given that streamlining diagnosis could allow for earlier treatment initiation during the critical neurodevelopmental window. For example, researchers have used ML to examine behavioral phenotypes of children with ASD with a high rate of accuracy and to shorten the time for observation-based screening and diagnosis. – In addition, ML has been used to differentiate between ASD and ADHD with high accuracy using a small number of measured behaviors. Research has also explored facial expression analysis based on dynamic deep learning and 3D behavior analysis to detect and distinguish between ADHD, ASD, and comorbid ADHD/ASD presentations. Along with risk prediction and diagnostics, AI holds potential to enhance treatment delivery in the field of developmental and behavioral pediatrics. AI robots have been used in a number of intervention studies to enhance social skill acquisition and spontaneous language development in children with ASD. , The potential utility of AI-enabled technologies to treat disruptive behaviors and mood and anxiety disorders in children is also being explored. In the field of ADHD, a number of emerging technologies show promise to infer behaviors that can then be used to tailor feedback to enhance self-regulation. For example, in 1 study, a neural network was used to learn behavioral intervention delivery techniques based on human demonstrations. The trained network then enabled a robot to autonomously deliver a similar intervention to children with ADHD. As health data infrastructure expands in size and sophistication, future AI technologies could potentially support increasingly personalized treatment options. Once comprehensive and integrated biologic, anatomic, physiologic, environmental, socioeconomic, behavioral, and pharmacogenomic patient data become routinely available, AI-based nearest-neighbor analysis could be used to identify “digital twins.” Patients with similar genomic and clinical features could be identified and clustered, for example, to allow for highly targeted treatments. Such approaches, while currently largely theoretical, could help predict therapeutic and adverse medication responses more accurately and also form an evidence base for personalized treatment pathways. Artificial intelligence algorithms have potential to automate many arduous administrative tasks, thereby streamlining care pathways and freeing clinicians to spend more time with patients. Natural language processing solutions, for example, are being developed to decrease reliance on human scribes in clinical encounters. Such approaches may be used to process and transform clinical-free text into meaningful structured outcomes, automate some documentation practices through text summarization, and scan text-based reports to support accurate and rapid diagnostic recommendations. Natural language processing solutions may prove particularly valuable in fields such as child and adolescent psychiatry and developmental and behavioral pediatrics that are extremely text-heavy. Developmental behavioral assessments often involve text-heavy tasks such as documentation of complex patient histories, results of extensive testing, and collateral history from multiple informants. Given the current national US shortage of child and adolescent psychiatrists with a median of 11 psychiatrists per 100,000 children, AI-based approaches that streamline and automate administrative tasks seem particularly promising. AI-assisted image interpretation has also shown potential to increase workplace productivity and provide considerable cost savings over current practice. , Techniques to streamline medical research and drug discovery by using natural language processing to rapidly scan biomedical literature and data mine molecular structures are also being developed. Artificial intelligence has potential to address several bias and access disparities apparent in existing care models. While access to developmental and behavioral specialists is extremely limited in much of the world, it is estimated that over 50% of the global population has access to a smartphone. Digital AI–based diagnostic and treatment platforms could thus potentially expand access to underserved and geographically remote populations. Thoughtful use of AI may also help to address racial, socioeconomic, and gender biases. In the field of developmental and behavioral pediatrics, for example, it has been noted that despite ASD prevalence rates being roughly equal across racial/ethnic and socioeconomic groups, human clinicians are more likely to diagnose Black, Latinx, and Asian children, as well as children from low-income families, at a later date than White children and children with a higher socioeconomic status. By integrating and training on large racial-conscious and gender-conscious data sets, AI algorithms can assess thousands of traits and features and build on the findings to assist clinicians in making more accurate, timely, and less biased ASD diagnoses. While AI presents multiple opportunities to the field of developmental and behavioral pediatrics, to date, a very few AI-based technologies have been broadly integrated into clinical practice. Challenges to the widespread deployment of AI in health care settings, along with potential solutions, are outlined below. Data Bias Any AI algorithm is only as good as the data from which it was derived. Simply put, the performance of the algorithm and the quality of its prediction are dependent on the quality of the data supplied. If the data are biased, imbalanced, or otherwise an incomplete representation of the target group, the derivative model's generalizability will be limited. A class imbalance problem can occur, for example, in cases in which the total number of 1 class of data (i.e., “girls” or “disease positive status”) is far less than the total number of another class of data (i.e., “boys” or “disease negative status”). These biases can sometimes be identified and addressed through techniques such as over- or undersampling, but at other times with “black-box” learning, it is harder to detect and fix the bias in the algorithm(s). Representative populations for most conditions, including in pediatrics, are not a homogeneous group. Thus, it is essential that the data sets used to train these models include balanced data with diverse representations of the clinical symptoms across gender, age, and race in order for the model to be generalizable. Without such safeguards, models may, in fact, perpetuate or amplify stereotypes or biases. , Data Sharing and Privacy Concerns To robustly train an algorithm, sufficient data are required, yet pediatric data sets can be limited by small sample sizes, especially when split by age group. For the case of rare pediatric diseases, extremely low prevalence rates mean the amount and type of data available to train a model on very limited. Creative use of deep learning techniques such as generative adversarial networks may be required in such cases to counteract the lack of data. Generative adversarial networks pit one neural network against another for the purpose of generating synthetic (yet realistic) data to support a variety of tasks such as image and voice generation. While such techniques can be extremely useful when appropriately applied, overreliance on synthetic data also comes with its own set of risks. Multiple data sets may also be combined to produce a sufficient volume of data for model training. However, combining data sets presents its own set of challenges including data privacy and ownership issues and difficulties integrating data with heterogeneous features. Federated learning is an emerging ML technique with potential to address some of these data sharing and privacy concerns. Federated learning allows algorithms to train across many decentralized servers or edge devices, exchanging parameters (i.e., the models' weights and biases) without explicitly exchanging the data samples themselves. This technique obviates many of the privacy issues engendered by uploading highly sensitive health data from different sources onto a single server. Algorithmic Transparency and Explainability A lack of transparency in certain types of ML algorithms such as deep neural networks has raised concern about their clinical trustworthiness. Many models used to analyze images and text, for example, include levels of complexity and multidimensionality that exceed intuitive understanding or interpretation. In cases in which a clinician is unable to understand how the algorithm produces an output, should the algorithm be relied on as part of their clinical decision-making process? Such concerns have led to calls for algorithmic deconvolution before use in health care settings. Other researchers argue, however, that current approaches to explainability disregard the reality that local explanations can be unreliable or too superficial to be meaningful and that rigorous model validation before deployment may be a more important marker of trustworthiness. Consumer and Clinician Preparedness If patients and clinicians mistrust AI-based technologies, and/or lack sufficient training to understand, in broad terms, how they function, clinical adoption may be delayed. As with all new tools, implementation matters, and discipline is required to ensure safe deployment of AI-based devices without loss of clinician skill. Overreliance on AI-based imaging at the expense of history and physical examination, for example, should be avoided. Patient reservations that will require consideration include safety, cost, choice, data bias, and data security concerns. While the American Medical Association has called for research into how AI should be addressed in medical education, current medical training lacks a consistent approach to AI education, and key licensing examinations do not test on this content. Clinicians and medical students alike have identified knowledge gaps and reservations regarding the use of AI in health care. – A number of preliminary frameworks for integrating AI curriculum into medical training have been proposed , , ; however, additional research is urgently required to develop and then integrate standardized AI content into medical training pathways. Ethical Ambiguities From an ethical standpoint, users of this technology must consider the direct impact and unintended consequences of AI implementation in general as well as specific implications within the clinical context. A number of well-publicized non–health-related cases have illustrated AI data privacy concerns, along with the ethically problematic potential for AI to amplify social, racial, and gender biases. There are also ethical concerns around the magnitude of harm that could occur if an ML algorithm, deployed clinically at scale, were to malfunction; associated impacts could far exceed the harm caused by a single clinician's malpractice. The use of AI for clinical decision-making also raises questions of accountability, such as who is liable if unintended consequences result from use of the technology (e.g., missed diagnosis), or what course of action an autonomous therapy chatbot might take if it detects speech patterns indicative of risk for self-harm. Ethical AI frameworks addressing such concerns are under development, , and researchers are calling for AI technologies to undergo robust simulation, validation, and prospective scrutiny before clinical adoption. Regulatory and Payment Barriers Notwithstanding data, provider adoption, and ethical safeguards, AI technologies face several systematic challenges to be readily implemented into clinical practice, including regulatory, interoperability with EMRs and data exchange, and payment barriers. Given that AI devices can learn from data and alter their algorithms accordingly, traditional medical device regulatory frameworks might not be sufficient. As a result, the Food and Drug Administration has developed a proposed regulatory framework that includes a potential “Predetermined Change Control Plan” for premarket submissions, including “Software Pre-Specifications” and an “Algorithm Change Protocol,” to address the iterative nature of AI/ML-based Software as a Medical Device technologies. Health care organizations and practices will also need to establish a data infrastructure and privacy policy for data that are stored across multiple servers and sources (e.g., medical records, health sensors, medical devices, etc). Development of new digital medical software and devices that use AI are likely to outpace the current health care payment structure. New billing codes associated with new treatments and procedures require formal approvals by national organizations with subsequent adoption by insurances, both public and private. This process can take many months to years. To facilitate provider and patient adoption of new AI technologies which may improve quality of care, streamlined development of billing codes for technologies using AI should be developed. Any AI algorithm is only as good as the data from which it was derived. Simply put, the performance of the algorithm and the quality of its prediction are dependent on the quality of the data supplied. If the data are biased, imbalanced, or otherwise an incomplete representation of the target group, the derivative model's generalizability will be limited. A class imbalance problem can occur, for example, in cases in which the total number of 1 class of data (i.e., “girls” or “disease positive status”) is far less than the total number of another class of data (i.e., “boys” or “disease negative status”). These biases can sometimes be identified and addressed through techniques such as over- or undersampling, but at other times with “black-box” learning, it is harder to detect and fix the bias in the algorithm(s). Representative populations for most conditions, including in pediatrics, are not a homogeneous group. Thus, it is essential that the data sets used to train these models include balanced data with diverse representations of the clinical symptoms across gender, age, and race in order for the model to be generalizable. Without such safeguards, models may, in fact, perpetuate or amplify stereotypes or biases. , To robustly train an algorithm, sufficient data are required, yet pediatric data sets can be limited by small sample sizes, especially when split by age group. For the case of rare pediatric diseases, extremely low prevalence rates mean the amount and type of data available to train a model on very limited. Creative use of deep learning techniques such as generative adversarial networks may be required in such cases to counteract the lack of data. Generative adversarial networks pit one neural network against another for the purpose of generating synthetic (yet realistic) data to support a variety of tasks such as image and voice generation. While such techniques can be extremely useful when appropriately applied, overreliance on synthetic data also comes with its own set of risks. Multiple data sets may also be combined to produce a sufficient volume of data for model training. However, combining data sets presents its own set of challenges including data privacy and ownership issues and difficulties integrating data with heterogeneous features. Federated learning is an emerging ML technique with potential to address some of these data sharing and privacy concerns. Federated learning allows algorithms to train across many decentralized servers or edge devices, exchanging parameters (i.e., the models' weights and biases) without explicitly exchanging the data samples themselves. This technique obviates many of the privacy issues engendered by uploading highly sensitive health data from different sources onto a single server. A lack of transparency in certain types of ML algorithms such as deep neural networks has raised concern about their clinical trustworthiness. Many models used to analyze images and text, for example, include levels of complexity and multidimensionality that exceed intuitive understanding or interpretation. In cases in which a clinician is unable to understand how the algorithm produces an output, should the algorithm be relied on as part of their clinical decision-making process? Such concerns have led to calls for algorithmic deconvolution before use in health care settings. Other researchers argue, however, that current approaches to explainability disregard the reality that local explanations can be unreliable or too superficial to be meaningful and that rigorous model validation before deployment may be a more important marker of trustworthiness. If patients and clinicians mistrust AI-based technologies, and/or lack sufficient training to understand, in broad terms, how they function, clinical adoption may be delayed. As with all new tools, implementation matters, and discipline is required to ensure safe deployment of AI-based devices without loss of clinician skill. Overreliance on AI-based imaging at the expense of history and physical examination, for example, should be avoided. Patient reservations that will require consideration include safety, cost, choice, data bias, and data security concerns. While the American Medical Association has called for research into how AI should be addressed in medical education, current medical training lacks a consistent approach to AI education, and key licensing examinations do not test on this content. Clinicians and medical students alike have identified knowledge gaps and reservations regarding the use of AI in health care. – A number of preliminary frameworks for integrating AI curriculum into medical training have been proposed , , ; however, additional research is urgently required to develop and then integrate standardized AI content into medical training pathways. From an ethical standpoint, users of this technology must consider the direct impact and unintended consequences of AI implementation in general as well as specific implications within the clinical context. A number of well-publicized non–health-related cases have illustrated AI data privacy concerns, along with the ethically problematic potential for AI to amplify social, racial, and gender biases. There are also ethical concerns around the magnitude of harm that could occur if an ML algorithm, deployed clinically at scale, were to malfunction; associated impacts could far exceed the harm caused by a single clinician's malpractice. The use of AI for clinical decision-making also raises questions of accountability, such as who is liable if unintended consequences result from use of the technology (e.g., missed diagnosis), or what course of action an autonomous therapy chatbot might take if it detects speech patterns indicative of risk for self-harm. Ethical AI frameworks addressing such concerns are under development, , and researchers are calling for AI technologies to undergo robust simulation, validation, and prospective scrutiny before clinical adoption. Notwithstanding data, provider adoption, and ethical safeguards, AI technologies face several systematic challenges to be readily implemented into clinical practice, including regulatory, interoperability with EMRs and data exchange, and payment barriers. Given that AI devices can learn from data and alter their algorithms accordingly, traditional medical device regulatory frameworks might not be sufficient. As a result, the Food and Drug Administration has developed a proposed regulatory framework that includes a potential “Predetermined Change Control Plan” for premarket submissions, including “Software Pre-Specifications” and an “Algorithm Change Protocol,” to address the iterative nature of AI/ML-based Software as a Medical Device technologies. Health care organizations and practices will also need to establish a data infrastructure and privacy policy for data that are stored across multiple servers and sources (e.g., medical records, health sensors, medical devices, etc). Development of new digital medical software and devices that use AI are likely to outpace the current health care payment structure. New billing codes associated with new treatments and procedures require formal approvals by national organizations with subsequent adoption by insurances, both public and private. This process can take many months to years. To facilitate provider and patient adoption of new AI technologies which may improve quality of care, streamlined development of billing codes for technologies using AI should be developed. Artificial intelligence (AI) in health care is not just a futuristic premise, and adoption has shifted from the “early adopter” fringe to a mainstream concept. The convergence of enhanced computational power and cloud storage solutions, increasingly sophisticated machine learning (ML) approaches and rapidly expanding volumes of digitized health care data, has ushered in this new wave of AI-based technologies. Strong economic investment in the AI health care sector, together with the growing number of AI-driven devices being granted regulatory approval, underscores the increasing role of AI in the future health care landscape. In the field of developmental and behavioral pediatrics, we are at an inflection point at which AI-driven technologies show potential to augment clinical decision-making, risk prediction, diagnostics, and treatment delivery. In addition, AI may be leveraged to automate certain time-intensive and arduous clinical tasks and to streamline workflows. Future research is still needed to address impediments to widespread clinical adoption. These include data bias, privacy, ownership and integration issues, disquietude over a perceived lack of algorithmic transparency, regulatory and payer bottlenecks, ethical ambiguities, and lack of rigorous and standardized AI-focused clinician training. AI technologies are not meant to replace the practicing physician or his/her clinical judgment, nor will they serve as a panacea to all the shortcomings of modern health care. However, we are optimistic about the future of AI in health care, including developmental and behavioral pediatrics. By enhancing the efficiency and impact of health care processes, AI approaches promise to reduce barriers to care and maximize the time clinicians are able to spend with their patients. |
Pharmacogenetic profiling via genome sequencing in children with medical complexity | 15b74f82-b867-4ef1-9424-f9b10db44ffb | 10033400 | Pharmacology[mh] | Children with medical complexity (CMC) are a well-studied, clinically defined group in pediatrics. – They typically have at least one severe chronic condition, technology dependence, multiple subspecialist involvement, and extensive care coordination needs. Polypharmacy is common in CMC – and was recently identified as a high-priority research area by clinicians and families. Adverse drug reactions (ADRs) and drug therapeutic failure are both a cause and a consequence of polypharmacy in children. , For many medications, dose requirements, efficacy, and risk for ADRs are partially determined by an individual’s genetic profile. , Genotype-guided prescribing is an innovative care model in pediatric medicine that has not been explored in CMC. Medications prescribed to children often have established, clinically actionable drug-gene interactions that afford opportunities for genotype-guided prescribing. A major barrier to the wider adoption of pharmacogenetic (PGx) testing in routine clinical practice is that results are rarely already available at the point of prescription. To address this in CMC, it may be possible to utilize the data already generated from the high rate of uptake of diagnostics-focused genetic testing. , Exome sequencing and genome sequencing (GS) are increasingly considered first- or second-tier tests for genetically heterogeneous pediatric presentations, , , which includes most CMC. GS data can be repurposed to identify PGx variation and corresponding phenotypes, in a process that we term GS-PGx profiling. The utility of GS-PGx profiling in the overall CMC population is unknown. In this study, we characterized the landscape of polypharmacy in a large cohort of CMC, including annotating medications for known drug-gene associations. We then organized GS-PGx profiling for a subgroup with existing GS data. We hypothesized that a majority of CMC would be prescribed medications with established PGx associations detectable by GS-PGx profiling. Defining the study population CMC were considered for this study if they were followed by the Complex Care Program at The Hospital for Sick Children (Toronto, Canada) at any point between January 1, 2010, and November 1, 2020. Polypharmacy is not a formal criterion for acceptance into this Complex Care Program. Of the 837 potentially eligible CMC, 35 were excluded because the family: (i) declined Complex Care services after referral or were not followed long enough to have a comprehensive care plan, and/or (ii) requested a closed chart and declined data sharing. For each of the remaining 802 CMC, phenotype, medication, and genetic testing data were extracted from their electronic medical records and stored in a REDCap database. This retrospective chart review with an accompanying patient consent waiver was approved by the Research Ethics Board at The Hospital for Sick Children. A subgroup of CMC and their family members had existing GS data and subsequently underwent GS-PGx profiling (see below). Additional recruitment details and phenotype data for this subgroup were published previously; one additional proband and his two parents were sequenced after this publication, for a total of n = 50 CMC probands and n = 89 parents. Annotating medications with PGx associations Current medications were those listed in each child’s most recent comprehensive care plan. Medications were categorized by target system(s) and pharmacologic indication(s) using pharmacology indexing databases including Micromedex® (micromedexsolutions.com), and then annotated for PGx associations with “pharmacogenes”. These drug-gene interactions may prompt clinical action to alter medication plans according to Clinical Pharmacogenetics Implementation Consortium (CPIC®) Dosing Guidelines. We consulted either guidelines specific to pediatric populations, or guidelines applicable to both adult and pediatric populations. We included drug-gene associations with confirmed CPIC® levels of significance A or B, and/or those with an “Actionable PGx” label as denoted by the Food and Drug Administration. Natural health products, topical agents, as-needed or PRN medications, and select other compounds were a priori excluded from medication counts (Supplementary Table ) for the following reasons: (i) precedent set by prior PGx studies, , , , and (ii) suspected high rate of use and inconsistent reporting in comprehensive care plans. GS-PGx profiling We performed GS using our established methods , at The Centre for Applied Genomics (Toronto, Canada). Briefly, we completed short-read GS with the HiSeq X Platform (Illumina Inc) using blood-derived DNA from 50 CMC and their family members. Stargazer (version 1.0.8) was used to call genetic polymorphisms with known PGx associations. Stargazer detects single nucleotide, indel, and structural variants to output PGx diplotypes of 51 possible pharmacogenes. We selected and obtained results for 16 pharmacogenes with clinically significant associations: CACNA1S, CFTR, CYP2B6, CYP2C19, CYP2C9, CYP2D6, CYP3A5, DPYD, G6PD, NAT2, NUDT15, RYR1, SLCO1B1, TPMT, UGT1A1 , and VKORC1 . Quality control measures included using the family data to ensure Mendelian segregation of specific alleles. We called known CYP2D6 structural variants (e.g., CYP2D6*5) but not novel structural variants, because of the complexity of the region and consequent technical limitations of Stargazer. Variants that do not follow conventional PGx nomenclature were named with an “S” prefix, as per the naming convention within Stargazer. PGx diplotypes were then analyzed to determine their corresponding phenotypes (where known). Phenotype categories included a metabolizer status of normal, intermediate, poor, rapid, or ultrarapid, as well as a gene function status of normal, increased, or decreased function. We use the term “PGx variant(s)” in this study to refer to all non-normal metabolizer and gene function statuses. Statistical methods Standard descriptive statistics and graphs were generated using R statistical software, version 4.1.0 (R Foundation for Statistical Computing). Statistical significance was defined as a two-tailed p value of <0.05. CMC were considered for this study if they were followed by the Complex Care Program at The Hospital for Sick Children (Toronto, Canada) at any point between January 1, 2010, and November 1, 2020. Polypharmacy is not a formal criterion for acceptance into this Complex Care Program. Of the 837 potentially eligible CMC, 35 were excluded because the family: (i) declined Complex Care services after referral or were not followed long enough to have a comprehensive care plan, and/or (ii) requested a closed chart and declined data sharing. For each of the remaining 802 CMC, phenotype, medication, and genetic testing data were extracted from their electronic medical records and stored in a REDCap database. This retrospective chart review with an accompanying patient consent waiver was approved by the Research Ethics Board at The Hospital for Sick Children. A subgroup of CMC and their family members had existing GS data and subsequently underwent GS-PGx profiling (see below). Additional recruitment details and phenotype data for this subgroup were published previously; one additional proband and his two parents were sequenced after this publication, for a total of n = 50 CMC probands and n = 89 parents. Current medications were those listed in each child’s most recent comprehensive care plan. Medications were categorized by target system(s) and pharmacologic indication(s) using pharmacology indexing databases including Micromedex® (micromedexsolutions.com), and then annotated for PGx associations with “pharmacogenes”. These drug-gene interactions may prompt clinical action to alter medication plans according to Clinical Pharmacogenetics Implementation Consortium (CPIC®) Dosing Guidelines. We consulted either guidelines specific to pediatric populations, or guidelines applicable to both adult and pediatric populations. We included drug-gene associations with confirmed CPIC® levels of significance A or B, and/or those with an “Actionable PGx” label as denoted by the Food and Drug Administration. Natural health products, topical agents, as-needed or PRN medications, and select other compounds were a priori excluded from medication counts (Supplementary Table ) for the following reasons: (i) precedent set by prior PGx studies, , , , and (ii) suspected high rate of use and inconsistent reporting in comprehensive care plans. We performed GS using our established methods , at The Centre for Applied Genomics (Toronto, Canada). Briefly, we completed short-read GS with the HiSeq X Platform (Illumina Inc) using blood-derived DNA from 50 CMC and their family members. Stargazer (version 1.0.8) was used to call genetic polymorphisms with known PGx associations. Stargazer detects single nucleotide, indel, and structural variants to output PGx diplotypes of 51 possible pharmacogenes. We selected and obtained results for 16 pharmacogenes with clinically significant associations: CACNA1S, CFTR, CYP2B6, CYP2C19, CYP2C9, CYP2D6, CYP3A5, DPYD, G6PD, NAT2, NUDT15, RYR1, SLCO1B1, TPMT, UGT1A1 , and VKORC1 . Quality control measures included using the family data to ensure Mendelian segregation of specific alleles. We called known CYP2D6 structural variants (e.g., CYP2D6*5) but not novel structural variants, because of the complexity of the region and consequent technical limitations of Stargazer. Variants that do not follow conventional PGx nomenclature were named with an “S” prefix, as per the naming convention within Stargazer. PGx diplotypes were then analyzed to determine their corresponding phenotypes (where known). Phenotype categories included a metabolizer status of normal, intermediate, poor, rapid, or ultrarapid, as well as a gene function status of normal, increased, or decreased function. We use the term “PGx variant(s)” in this study to refer to all non-normal metabolizer and gene function statuses. Standard descriptive statistics and graphs were generated using R statistical software, version 4.1.0 (R Foundation for Statistical Computing). Statistical significance was defined as a two-tailed p value of <0.05. Genetic test utilization and polypharmacy were both common in CMC In the cohort of 802 CMC, 447 were males (56%), the median year of birth was 2013 (range, 1999–2020), and the diversity in reported ancestry was reflective of the general population in our region (Supplementary Table ). Over 88% ( n = 706) had undergone at least one clinical genetic test. This included 314 CMC (39%) who had genome-wide testing (chromosomal microarray analysis, exome sequencing, and/or GS) before 1 year of age. The median number of current medications per child was 3 (range, 0–13) after relevant exclusions (Supplementary Table ), and 558 CMC (70%) were prescribed at least two medications. The most common classes of drugs were gastrointestinal (GI) agents ( n = 493, 61%) and central nervous system (CNS) agents ( n = 405, 50%) (Supplementary Table ). The most common medication sub-categories were gastric acid reducers ( n = 467, 58%), anticonvulsants ( n = 346, 43%), antiemetics ( n = 224, 28%), and asthma ( n = 205, 26%) (Supplementary Table ). CMC were often prescribed medications with PGx associations Overall, 546 (68%) of 802 CMC were currently prescribed at least one medication with an established PGx association (Fig. ). This included 450 CMC (56%) for one or more GI agents (e.g., 347 CMC were prescribed omeprazole, which interacts with CYP2C19 ) and 217 (27%) for one or more CNS agents (e.g., 117 CMC were prescribed clobazam, which interacts with CYP2C19 ) (Fig. ). The proportions of CMC prescribed medications with PGx associations, by drug category and sub-category, are listed in Supplementary Tables and , respectively. Results were similar in the subgroup that underwent GS-PGx profiling (Fig. ), with 39 of 50 (78%) currently prescribed at least one medication with an established PGx association. GI, CNS, and respiratory agents with PGx associations were all in use by ten or more of these CMC (Supplementary Table ). The two medication sub-categories with the highest PGx relevance were gastric acid reducers (specifically, the proton-pump inhibitors (PPIs) omeprazole, lansoprazole, and pantoprazole; currently prescribed to a total of 31 CMC) and anticonvulsants (specifically, carbamazepine, clobazam, lamotrigine, oxcarbazepine, and valproic acid; currently prescribed to a total of 16 CMC). Half (8 of 16 CMC) were prescribed two or more of these anticonvulsants. GS-PGx profiling identified findings in CMC relevant to their current medications The median number of PGx variants per CMC was 5 (range, 2–8). GS-PGx findings by pharmacogene are summarized in Fig. . For example, 32 (64%) of the 50 CMC had CYP2C19 diplotypes that could impact dosing for some of the most prescribed medications in CMC (i.e., PPIs): 13 were intermediate metabolizers, 13 were rapid metabolizers, 3 were ultrarapid metabolizers, and 3 were poor metabolizers (Fig. ). The burden of PGx variants amongst the parents of CMC was similar, with a median of 5 (range, 1–8) per parent (Supplementary Table ). Cross-referencing GS-PGx profiling results with current medication lists identified 48% of CMC (24 of 50) with at least one applicable drug-gene association (Fig. and Supplementary Table ). This included 5 CMC (10%) who were prescribed two or more different medications with each impacted by that child’s variation in a different pharmacogene. A major contributor to these findings was the association between CYP2C19 diplotypes and metabolism of PPIs (Fig. and Supplementary Table ). There were 18 CMC with metabolizer statuses currently affecting a prescribed medication: 9 intermediate, 7 rapid, 1 ultrarapid, and 1 poor. Eight additional CMC were not currently prescribed a PPI but had CYP2C19 diplotypes indicating a rapid or ultrarapid metabolizer status. Review of lifetime medication histories revealed that at least five of eight had trialed a PPI in the past, suggesting a missed opportunity for genotype-guided prescribing. Figure depicts a representative case vignette. In the cohort of 802 CMC, 447 were males (56%), the median year of birth was 2013 (range, 1999–2020), and the diversity in reported ancestry was reflective of the general population in our region (Supplementary Table ). Over 88% ( n = 706) had undergone at least one clinical genetic test. This included 314 CMC (39%) who had genome-wide testing (chromosomal microarray analysis, exome sequencing, and/or GS) before 1 year of age. The median number of current medications per child was 3 (range, 0–13) after relevant exclusions (Supplementary Table ), and 558 CMC (70%) were prescribed at least two medications. The most common classes of drugs were gastrointestinal (GI) agents ( n = 493, 61%) and central nervous system (CNS) agents ( n = 405, 50%) (Supplementary Table ). The most common medication sub-categories were gastric acid reducers ( n = 467, 58%), anticonvulsants ( n = 346, 43%), antiemetics ( n = 224, 28%), and asthma ( n = 205, 26%) (Supplementary Table ). Overall, 546 (68%) of 802 CMC were currently prescribed at least one medication with an established PGx association (Fig. ). This included 450 CMC (56%) for one or more GI agents (e.g., 347 CMC were prescribed omeprazole, which interacts with CYP2C19 ) and 217 (27%) for one or more CNS agents (e.g., 117 CMC were prescribed clobazam, which interacts with CYP2C19 ) (Fig. ). The proportions of CMC prescribed medications with PGx associations, by drug category and sub-category, are listed in Supplementary Tables and , respectively. Results were similar in the subgroup that underwent GS-PGx profiling (Fig. ), with 39 of 50 (78%) currently prescribed at least one medication with an established PGx association. GI, CNS, and respiratory agents with PGx associations were all in use by ten or more of these CMC (Supplementary Table ). The two medication sub-categories with the highest PGx relevance were gastric acid reducers (specifically, the proton-pump inhibitors (PPIs) omeprazole, lansoprazole, and pantoprazole; currently prescribed to a total of 31 CMC) and anticonvulsants (specifically, carbamazepine, clobazam, lamotrigine, oxcarbazepine, and valproic acid; currently prescribed to a total of 16 CMC). Half (8 of 16 CMC) were prescribed two or more of these anticonvulsants. The median number of PGx variants per CMC was 5 (range, 2–8). GS-PGx findings by pharmacogene are summarized in Fig. . For example, 32 (64%) of the 50 CMC had CYP2C19 diplotypes that could impact dosing for some of the most prescribed medications in CMC (i.e., PPIs): 13 were intermediate metabolizers, 13 were rapid metabolizers, 3 were ultrarapid metabolizers, and 3 were poor metabolizers (Fig. ). The burden of PGx variants amongst the parents of CMC was similar, with a median of 5 (range, 1–8) per parent (Supplementary Table ). Cross-referencing GS-PGx profiling results with current medication lists identified 48% of CMC (24 of 50) with at least one applicable drug-gene association (Fig. and Supplementary Table ). This included 5 CMC (10%) who were prescribed two or more different medications with each impacted by that child’s variation in a different pharmacogene. A major contributor to these findings was the association between CYP2C19 diplotypes and metabolism of PPIs (Fig. and Supplementary Table ). There were 18 CMC with metabolizer statuses currently affecting a prescribed medication: 9 intermediate, 7 rapid, 1 ultrarapid, and 1 poor. Eight additional CMC were not currently prescribed a PPI but had CYP2C19 diplotypes indicating a rapid or ultrarapid metabolizer status. Review of lifetime medication histories revealed that at least five of eight had trialed a PPI in the past, suggesting a missed opportunity for genotype-guided prescribing. Figure depicts a representative case vignette. These results indicate that CMC are often prescribed medications with established PGx associations and dosing guidelines. PGx diplotypes can be reliably extracted from GS data. Genetic test utilization is already high in CMC, and exome sequencing and chromosomal microarray analysis are expected to be replaced by GS in the coming years. , , , Genotype-guided prescribing can have the greatest impact when initiated at a child’s first point of contact with the healthcare system, with the caveat that some findings may not be applicable until after the neonatal period or infancy. , GS-PGx profiling at the time of initial etiologic-based testing therefore warrants strong consideration in CMC (Fig. ). CMC are a priority population for trialing genotype-guided prescribing in pediatrics Unique characteristics of CMC provide the rationale for positioning them at the leading-edge of broad PGx testing amongst children and adolescents. Neurological impairment, multi-organ system disease, and multiple subspecialist prescribers are all common, and these factors can complicate clinical assessment of treatment response/failure and side effects. Medication use patterns are shared across CMC because of the development of similar comorbidities over time, particularly in those with severe neurological impairment. Medication dosing that is unsuited to the individual’s genetic profile may place additional stress on patients and their families. , PGx data can also provide insight into drug-drug interactions, a common concern in polypharmacy. The prevalence of polypharmacy in this study cohort was comparable to adults with psychiatric illness and the elderly, populations where PGx profiling is most common and best established. As expected, PGx variants were as common in CMC as they are in the general population. , These observations suggest a strong potential for GS-PGx profiling to alter medication choices and dosages for CMC, particularly with PPI selection and dosing for rapid and ultrarapid metabolizers in accordance with published CPIC® guidelines. We propose to integrate GS-PGx as a secondary use of GS data being generated for diagnostic purposes. Efforts to clinically validate this approach are in progress at our center and others. Automated reporting will facilitate its application. With GS-PGx being a low-cost adjunct analysis to an already planned GS experiment, there is the potential for cost-effectiveness. Important barriers and knowledge gaps remain, however. There is a relative paucity of data in the pediatric age range. Certain “established” PGx associations cannot be extrapolated to neonates because of key physiological differences (e.g., immature enzyme expression). Clinical implementation of GS-PGx will need to be accompanied by continuing professional education and other initiatives to ensure appropriate interpretation of findings at the bedside. Advantages and limitations We used a cross-sectional design that captured current medication use at a single point of time. As illustrated by our post hoc review of lifetime medical records for those with CYP2C19 rapid and ultrarapid metabolizer statuses, we have likely underestimated both the scope of polypharmacy and the potential role of PGx. Our a priori exclusion criteria with respect to medication counts were also conservative; many as-needed or PRN medications have well-established drug-gene associations (e.g., ibuprofen and CYP2C9 ). We were unable to determine conclusively whether current or past medication use was influenced by ADRs. We were conservative in considering only drug-gene associations at CPIC® levels of significance A and B only. Many drugs remain under review for clinical significance and have not yet been assigned a CPIC® significance level (resulting in “Provisional” status). Provisional drug-gene pairs like valproic acid and POLG , or fluticasone propionate and CRHR1 , could become particularly relevant to CMC given the high rate of associated medication use. Compared with targeted genotyping approaches, GS-PGx profiling was able to identify uncommon PGx alleles in this ethnically diverse cohort (e.g., CYP2C9*3, CYP2D6*20; Supplementary Table ). However, interpretation of ultra-rare and novel genetic variants in pharmacogenes, which can be detected by GS, remains challenging. HLA genotyping remains beyond the analytical scope of GS-PGx for now because of the complexity of that genomic region. Last, we acknowledge the ongoing technical limitations of Stargazer. There continue to be challenges in predicting rare alleles and resolving star alleles in instances of heavy sequence noise and complex structural variation. For example, UGT1A1 *28 is a short tandem repeat in a non-coding region that cannot be reliably detected because of the complexity of regional structural variation. At present, Stargazer is the bioinformatics tool that is most readily available and widely used to perform GS-PGx profiling. Improvements in both GS and PGx profiling are expected over time. Unique characteristics of CMC provide the rationale for positioning them at the leading-edge of broad PGx testing amongst children and adolescents. Neurological impairment, multi-organ system disease, and multiple subspecialist prescribers are all common, and these factors can complicate clinical assessment of treatment response/failure and side effects. Medication use patterns are shared across CMC because of the development of similar comorbidities over time, particularly in those with severe neurological impairment. Medication dosing that is unsuited to the individual’s genetic profile may place additional stress on patients and their families. , PGx data can also provide insight into drug-drug interactions, a common concern in polypharmacy. The prevalence of polypharmacy in this study cohort was comparable to adults with psychiatric illness and the elderly, populations where PGx profiling is most common and best established. As expected, PGx variants were as common in CMC as they are in the general population. , These observations suggest a strong potential for GS-PGx profiling to alter medication choices and dosages for CMC, particularly with PPI selection and dosing for rapid and ultrarapid metabolizers in accordance with published CPIC® guidelines. We propose to integrate GS-PGx as a secondary use of GS data being generated for diagnostic purposes. Efforts to clinically validate this approach are in progress at our center and others. Automated reporting will facilitate its application. With GS-PGx being a low-cost adjunct analysis to an already planned GS experiment, there is the potential for cost-effectiveness. Important barriers and knowledge gaps remain, however. There is a relative paucity of data in the pediatric age range. Certain “established” PGx associations cannot be extrapolated to neonates because of key physiological differences (e.g., immature enzyme expression). Clinical implementation of GS-PGx will need to be accompanied by continuing professional education and other initiatives to ensure appropriate interpretation of findings at the bedside. We used a cross-sectional design that captured current medication use at a single point of time. As illustrated by our post hoc review of lifetime medical records for those with CYP2C19 rapid and ultrarapid metabolizer statuses, we have likely underestimated both the scope of polypharmacy and the potential role of PGx. Our a priori exclusion criteria with respect to medication counts were also conservative; many as-needed or PRN medications have well-established drug-gene associations (e.g., ibuprofen and CYP2C9 ). We were unable to determine conclusively whether current or past medication use was influenced by ADRs. We were conservative in considering only drug-gene associations at CPIC® levels of significance A and B only. Many drugs remain under review for clinical significance and have not yet been assigned a CPIC® significance level (resulting in “Provisional” status). Provisional drug-gene pairs like valproic acid and POLG , or fluticasone propionate and CRHR1 , could become particularly relevant to CMC given the high rate of associated medication use. Compared with targeted genotyping approaches, GS-PGx profiling was able to identify uncommon PGx alleles in this ethnically diverse cohort (e.g., CYP2C9*3, CYP2D6*20; Supplementary Table ). However, interpretation of ultra-rare and novel genetic variants in pharmacogenes, which can be detected by GS, remains challenging. HLA genotyping remains beyond the analytical scope of GS-PGx for now because of the complexity of that genomic region. Last, we acknowledge the ongoing technical limitations of Stargazer. There continue to be challenges in predicting rare alleles and resolving star alleles in instances of heavy sequence noise and complex structural variation. For example, UGT1A1 *28 is a short tandem repeat in a non-coding region that cannot be reliably detected because of the complexity of regional structural variation. At present, Stargazer is the bioinformatics tool that is most readily available and widely used to perform GS-PGx profiling. Improvements in both GS and PGx profiling are expected over time. GS-PGx profiling at the time of diagnostics-focused genetic testing could be an efficient way to incorporate precision prescribing practices into the lifelong care of CMC. These data provide the impetus for further study of GS-PGx, to determine therapeutic, patient outcome, and societal efficacies in clinical practice. Supplementary File Table S3 |
SEOM–GEICO clinical guideline on epithelial ovarian cancer (2023) | 2dc9c336-b3ef-41d5-8f84-dd4cbd771606 | 11467069 | Internal Medicine[mh] | EOC represents a heterogeneous disease with clinically, pathologically, and clinically different tumours. Histological subtype, stage at diagnosis, molecular biomarkers, and access to appropriate surgery and systemic therapy in specialized centres are crucial factors that will impact outcomes. Cytoreductive surgery with no macroscopic residual disease and combination of platinum–taxane chemotherapy (ChT) remain the mainstay of therapy. Maintenance treatment with antiangiogenics and/or poly (adenosine diphosphate-ribose) polymerase (PARP) inhibitors has proven to exert an important impact on clinical outcomes . The incorporation of molecular biology to identify predictive biomarkers into standard practice will enable selection of those patients who would benefit most from targeted agents. Therapeutic options in the setting of recurrence remain limited, which highlights a substantial unmet need . This SEOM–GEICO guideline provides updated evidence-based recommendations for the current treatment of EOC, primary peritoneal, and fallopian tube cancer, globally considered as EOC throughout this guideline. This guideline is based on a systematic review of relevant published studies and with the consensus of ten oncologists who are experts in the treatment of these neoplasms from GEICO (Grupo Español de Investigación en Cáncer Ginecológico) and SEOM (Sociedad Española de Oncología Médica), as well as an external review panel of two experts designated by SEOM. The Infectious Diseases Society of America–US Public Health Service Grading System for Ranking Recommendations in Clinical Guidelines has been used to assign levels of evidence and grades of recommendation. Final recommendations on each chapter are based solely on those drugs approved by the EMA and/or FDA. Incidence and epidemiology EOC is the second most deadly gynaecological cancer worldwide and the first in developed countries, responsible for some 200,000 deaths annually. In Spain, 1979 women died of OC in 2021, and approximately 3584 new cases were diagnosed during 2023 . Median age at the time of diagnosis is approximately 63 years. Nulliparity, obesity, and treatment with oestrogen therapy are known risk factors for EOC. Genital use of talc powder has been suggested as a potential risk factor. In contrast, higher parity, oral contraceptive use, and breastfeeding have a protective role. Tobacco have been studied without conclusive results . Moreover, high grade serous (HGS) carcinomas are strongly associated with family history and hereditary syndromes. Mutations in the BRCA1/2 genes, which are detected in 10–15% of patients with OC and cause Hereditary Breast and Ovarian Cancer syndrome, are associated with a 15–65% risk of EOC, especially HGS histology. Mutations in the MLH1 , MSH2 , MSH6 , and PMS2 genes (diagnostic of Lynch syndrome), also correlate with a 12% risk of developing OC, mainly endometrioid or clear cell. Mutations in the ATM , BRIP1 , PALB2 , RAD51C and RAD51D genes are associated with a moderate risk of developing OC . Diagnosis and staging More than two-thirds of all cases are diagnosed at an advanced stage, given that the symptoms of early stage EOC are not specific. Two large prospective trials, the UKTCTOCS and PLCO trials that enrolled 202,562 and 78,216 women, respectively, found that a screening program for OC had no clear impact on mortality . Common symptoms of EOC include abdominal/pelvic pain, constipation, urinary frequency, abdominal distension, shortness of breath, and fatigue. The initial evaluation includes a physical examination, laboratory testing including CA 125, and pelvic ultrasound. Elevated HE4 levels identify malignancy with a sensitivity similar to that of CA 125, albeit with greater specificity. Algorithms such as the International Ovarian Tumour Analysis (IOTA) Simple Rules risk model or IOTA Assessment of Different NEoplasias in the adneXa (ADNEX) model can be useful in distinguishing benign from malignant pelvic tumours. Computed tomography (CT) imaging of the thorax, abdomen, and pelvis defines the extent of the disease and informs treatment planning. If available, magnetic resonance imaging (MRI) and positron emission tomography (PET)–CT can enhance assessment accuracy in advanced disease. Initial laparoscopy is a mainstay for histopathological diagnosis and to evaluate the likelihood of complete cytoreduction . Peritoneal extension scores, such as the peritoneal cancer index or the Fagotti score could be useful in this context. An adequate amount of tissue is essential to establish a pathological diagnosis and analyse biomarkers to guide the treatment plan. Cytological testing of ascites and pleural fluid, if present, is required to complete staging. FIGO 2014 is the staging system currently recommended (Table ). Recommendations Screening for EOC in average risk women is not recommended (I, E). The initial evaluation of a patient with suspicion of EOC should include physical examination, laboratory testing with CA-125, and pelvic ultrasound [II, A]. CT of the thorax, abdomen, and pelvis is recommended in the diagnostic workup to assess the extent of the disease (II, A). Laparoscopic surgery is recommended in advanced EOC to evaluate cytoreduction, obtain material for pathologic diagnosis and predictive biomarkers (II, B). Pathology and molecular biology EOC is the most common ovarian cancer histology (~ 90) and can be classified into five main subtypes according to the 2020 World Health Organization (WHO) classification: high-grade serous (HGS, 70%), low grade serous (LGS, 10%), endometrioid carcinoma (EC, 10%), clear cell carcinoma (CCC, 5%), and mucinous carcinoma (MC, 3%). Other less frequent histologies (2–3%) include undifferentiated carcinoma, malignant Brenner tumour, mesonephric-like carcinoma, mixed carcinomas, or carcinosarcoma . Diagnosis should be made by an expert gynaecological pathologist by immunohistochemistry (IHC) to avoid misdiagnoses. Each EOC subtype depends on a different molecular background (Table ) that can address different pathogenesis, clinical features, response to treatments, and prognosis . BRCA1/2 mutational status should be determined at primary diagnosis in every non-mucinous EOC regardless of age at diagnosis or family history of cancer. Although less common, Lynch syndrome can be associated with mucinous and non-mucinous EOC. Mismatch repair ( MMR ) genes testing is highly recommended depending on the family history of cancer. More extensive germline next generation sequencing (NGS) panels should be offered, depending on clinical suspicion . Homologous recombination deficiency (HRD) provides prognostic information and can be used as a predictive biomarker of the magnitude of response to PARP inhibitors (PARPi). Commercially available NGS tests detect genetic scars and somatic BRCA1/2 mutations as subrogates of HRD and are the most common way to test homologous recombination status in clinics. Different cutoffs can be observed depending on the test. HRD status should be assessed at diagnosis in at least every high-grade EOC patient (i.e., serous and endometroid subtype) . Recommendations Initial IHC-based diagnosis should be made by a gynaecological pathology expert [IV, A]. All patients with non-mucinous EOC should be tested for somatic and/or germline BRCA1/2 mutation at primary diagnosis to discard a hereditary syndrome, to obtain prognostic information, and to select a biomarker of response to PARPi [I, A]. Determination of HRD with a clinically validated test is strongly recommended at initial diagnosis in every high-grade serous or endometrioid EOC to provide prognostic and predictive biomarker information [I, A].MMR testing and/or more extensive germline NGS panels are recommended depending on clinical suspicion and family history of cancer [II, A]. Management of early stage disease (FIGO stages I–II) Surgery Treatment of early EOC requires surgery including hysterectomy, bilateral salpingo-oophorectomy, systematic pelvic and para-aortic lymphadenectomy, omentectomy, appendectomy in MC, random peritoneal biopsies of all surfaces, peritoneal washings with cytological examination, and complete exploration of the peritoneal cavity (Fig. ). In FIGO stage I, low-grade endometroid and expansile mucinous due to a rate of lymph node involvement < 1% systematic pelvic and para-aortic lymphadenectomy is not recommended. The standard approach is by supra- and infra-umbilical median laparotomy. The laparoscopic approach is being investigated, but we currently lack prospective data demonstrating its equivalence to open surgery . The aim of the surgery is to remove the entire tumour and stage the disease so as to establish the indication for subsequent adjuvant treatment. Fertility-preserving surgery could be considered in stage IA or IC disease with unilateral ovarian involvement and low histologic grade . Chemotherapy Adjuvant platinum-based ChT after complete surgery will be offered to all patients diagnosed with a high-grade tumour or stage II disease, based on long-term follow-up results from the ICON1-ACTION studies. Adjuvant ChT is not recommended in completely staged patients with LGS stage IA, low-grade EC stage IA, or expansile MC stage IA–IB. The benefit of adjuvant ChT is uncertain and could be regarded as optional in LGS stage IB–IC, CCC stage IA–IC1, low-grade EC stage IB–IC, expansile MC stage IC, and infiltrative MC stage IA . Standard adjuvant treatment consists of six cycles of platinum-based ChT. However, there is no consensus regarding treatment duration or optimal schedule. The GOG 157 study compared three vs. six cycles of carboplatin and paclitaxel. The study revealed no difference in recurrence rate, although a subsequent analysis did demonstrate a benefit in HGS in favour of the longer treatment; nevertheless, this group represented only 23% of the study population . Although the most commonly used schedule is the platinum and taxane combination, there is no evidence that adding paclitaxel is of benefit in adjuvant treatment of early ovarian cancer. Thus, carboplatin monotherapy may be an appropriate option in this setting. The magnitude of benefit as well as short- and long-term toxicity should be discussed with the patient prior to deciding on the adjuvant scheme to be used . Recommendations Surgery with complete tumour resection and complete surgical staging is the mainstay for both treatment and to establish the extent of the disease that informs the subsequent indication for systemic treatment (I, A). Adjuvant platinum-based ChT is recommended in all cases of high grade or stage II (IA). The most widely used option is the combination of carboplatin and paclitaxel, although carboplatin monotherapy may be acceptable (II, A). Treatment duration should be at least three cycles for all subtypes, albeit six cycles are recommended in HGS (II, A). Management of advanced-stage disease (AOC) (FIGO stages III–IV) Cytoreductive surgery The recommended strategy in individuals with stage III–IV disease is primary debulking surgery (PDS), followed by systemic treatment . Cytoreduction surgery in advanced EOC (FIGO stages III–IV) has a proven therapeutic purpose. Its goal must be complete excision (R0) of any visible tumour without leaving macroscopic residual disease, given that post-surgical volume will impact the risk of recurrence and patient survival . In light of the results of the LION and the CARACO trials, lymphadenectomy in primary fully resected AOC with clinically negative lymph nodes is no longer recommended nor in upfront nor in interval cytoreductive surgery . Surgical expertise and specialist training are known to result in better rates of complete cytoreduction. Hence, subjects with advanced disease are advised to undergo surgery in specialized centres with suitable infrastructure and trained teams . Chemotherapy Conventional treatment is based on a combination of carboplatin (AUC 5–6) and paclitaxel (175 mg/m 2 ) every 3 weeks for 6 cycles . Schedules with weekly ChT without bevacizumab improve neither progression-free survival (PFS) nor overall survival (OS) in patient populations from Western countries . Weekly carboplatin–paclitaxel improved QoL vs. 3-weekly suggesting a role in elder patients. However in a study in vulnerable elderly patients single-agent carboplatin or weekly ChT might have worse outcomes; consequently, the 3-week regimen remains the standard for all cases of AOC including the elderly . Prophylaxis of venous thromboembolism in advanced EOC patients should be discussed with each patient receiving systemic therapy considering their specific risk factors . Neoadjuvant chemotherapy The EORTC55971 trial and CHORUS trial found similar PFS and OS rates for patients with stage IIIC or IV disease receiving neoadjuvant chemotherapy (NACT) and interval debulking surgery (IDS) compared with PDS. Despite these non-inferiority results, the aforementioned trials have been criticized for their short median OS, mean operative time, and low optimal cytoreduction rates. Due to the limitations of these trials, it has yet to be determined whether NACT and IDS might be an option for people for whom complete resection at PDS seems feasible. Therefore, both approaches (PDS or NACT followed by IDS) may be deemed valid, although PDS is still the preferred primary treatment option when complete cytoreduction is feasible and patient is operable . Intraperitoneal chemotherapy Despite the fact that three large, randomized studies (GOG 104, GOG 114, and GOG 172) and one meta-analysis have found clinically significant improvements in PFS and OS with intraperitoneal (IP) ChT , the results of the GOG 252 trial revealed no benefit of IP therapy when bevacizumab was incorporated in all arms. Therefore, IP chemotherapy is not considered a standard of care . A randomized phase III trial evaluating hyperthermic intraperitoneal chemotherapy (HIPEC) after IDS evidenced better PFS and OS for the HIPEC arm. Nevertheless, this trial received notable methodological criticisms, in particular, the lack of stratification for known prognostic molecular factors. Therefore, HIPEC cannot be regarded as a standard treatment and should not be offered outside of the context of clinical trials . Maintenance treatment with bevacizumab Two large, randomized studies (GOG 218 and ICON 7) have reported that bevacizumab (15 mg/kg or 7.5 mg/kg every 3 weeks) added to adjuvant ChT after PDS and followed by maintenance therapy with bevacizumab for a maximum of 15 months improves PFS compared to standard adjuvant ChT alone. Post-hoc subgroup analyses indicated statistically significant OS benefit only in patients with stage IV disease in GOG 218 and patients at high risk of progression (defined as FIGO stage III with > 1 cm residual disease following PDS or stage IV) in the ICON7 trial . In the ENGOT Ov-15 study, prolonging the duration of bevacizumab administration (30 vs. 15 months) failed to improve PFS . Two small, prospective trials revealed that bevacizumab added to platinum-containing ChT in the neoadjuvant setting was safe, although it has no impact on complete resection rate or PFS . Maintenance treatment with PARP inhibitors Several phase III, randomized clinical trials (SOLO 1, PRIMA, PAOLA, PRIME, and ATHENA–MONO) have demonstrated that maintenance therapy with PARPi (olaparib, niraparib, olaparib plus bevacizumab, and rucaparib) after response to front-line platinum-containing regimens significantly increased median PFS in HGSOC . All trials have manifested a remarkable, unprecedented benefit inBRCA1/2 mutated individuals. Moreover, olaparib–bevacizumab, niraparib, and rucaparib also displayed a significant benefit in the HRD population. Finally, niraparib and rucaparib exhibited a benefit in the HR proficient (HRP) and HRD-unknown subgroups, albeit of lesser magnitude. The benefit observed with PARPi has been sustained throughout follow-up as demonstrated by their impact on PFS2, as well as by the results of long-term overall survival in the SOLO-1 and PAOLA-1 trials, with survival rates of 67% at 7 years and 65.5% at 5 years in the experimental arm vs. 46.5% and 48.4% in the control arm, respectively. The benefit was observed despite the fact that 40% of the subjects in the control group received subsequent PARP therapy . ATHENA-MONO trial also showed that rucaparib imrpoved PFS in all populations regardless of BRCA or HRD status having received a recent EMA approval for the indication of maintenance in first line. Based on these results, olaparib (with or without bevacizumab), rucaparib or niraparib after partial or complete response to first-line, platinum-based ChT are highly effective in BRCA1/2-mutated patients and are strongly recommended. According to the outcomes of the PAOLA-1 and PRIMA and ATHENA trials, niraparib, rucaparib or olaparib–bevacizumab are also highly recommended for women with HRD tumours. In the HRP subgroup, maintenance with niraparib, or rucaparib can also be considered, although bevacizumab remains a reasonable alternative. The choice of maintenance treatment should be based on: (1) molecular biomarkers ( BRCA1/2 and HRD status), (2) disease-related factors (stage at diagnosis, post-surgical residual disease, response to ChT, chemotherapy response score (CRS), CA-125 ELIMination Rate Constant K (KELIM)), and (3) patient characteristics (comorbidities, concomitant medication). Advanced, non-high grade serous ovarian cancer Paclitaxel–carboplatin ± bevacizumab is the standard systemic ChT used in non-HGSC . Multiple retrospective studies, however, demonstrated lower response rates in these histologic subtypes compared with HGSC . Due to the lower chemosensitivity, PDS is strongly recommended in uncommon histologies. Most LGSCs have high expression of oestrogen (ER) and progesterone receptors (PgR). Retrospective studies suggest a possible therapeutic value of hormone therapy in the maintenance of newly diagnosed advanced LGSC [IV, B]. Uncommon non-HGS histologies are an unmet need and inclusion of these patients in clinical trials is fervently encouraged. Recommendations Cytoreductive surgery aimed at achieving complete cytoreduction (absence of all visible residual disease) is the backbone of treatment for advanced EOC and should be performed in specialised centres [II, A]. Lymphadenectomy in primary, completely debulked AOC with clinically negative lymph nodes is not recommended [I, A]. Standard treatment is based on a combination of carboplatin (AUC 5–6) and paclitaxel (175 mg/m 2 ) every 3 weeks for 6 cycles [I, A]. If complete cytoreductive surgery is not feasible, NACT for three cycles followed by ICS and three cycles of ChT is recommended [I, A]. IP ChT and HIPEC are not considered standard of care [I, D]. The addition of bevacizumab to ChT should be contemplated, especially in patients with stage III and residual disease or stage IV [I, A]. Post-ChT maintenance treatment is recommended after in HG-EOC. PARPi are recommended only after partial or complete response or no evidence of disease after first-line platinum-based ChT: o In BRCA1/2-mutated subjects: single agent with olaparib, rucaparib or niraparib or combination with olaparib plus bevacizumab [I, A] o HRD tumours: niraparib, rucaparib or olaparib–bevacizumab [I, A] o HRP tumours: Niraparib, rucaparib [I, B] or bevacizumab [I, A] Paclitaxel–carboplatin ± bevacizumab is the standard systemic ChT used in non-HGSC [I, B]. In newly diagnosed advanced LGSC maintenance treatment after Cht with hormone therapy can be considered [IV, B]. Management of recurrent disease Approximately 85% of all individuals diagnosed with advanced EOC experience disease recurrence within 10 years. Choosing the optimal strategy for recurrent ovarian cancer (ROC) demands that several critical factors be evaluated. Treatment-free interval (TFI) following last-platinum (TFIp) therapy remains a pivotal prognostic factor. Nevertheless, TFI should not be the sole consideration when making clinical decisions, since other clinical and molecular characteristics can impact treatment response and must be pondered . Factors to consider in treatment assessment are histological subtype, BRCA1/2 status, extension of the disease and symptoms, feasibility and outcome of a potential second surgery, prior treatment and response, TFI from the last treatment, residual toxicity from previous therapeutic interventions, patient condition and comorbidities, and patient preferences . Surgery for relapse of ovarian cancer The role of secondary cytoreduction (SC) in patients with first relapse more than 6 month TFIp has been examined in the DESKTOP III trial . SC followed by ChT in cases selected on the basis of a favourable AGO score defined as a good performance status (ECOG 0), no residual disease after PDS, and the absence of ascites (< 500 ml) demonstrated an improvement in PFS and OS compared to ChT alone. The SOC-1 trial also assessed the role of SC in patients selected by an iModel algorithm. This trial evidenced a benefit in PFS from surgery, but OS data are still immature . In contrast, the GOG-0213 trial failed to prove superiority of SC. The absence of well-defined selection criteria for surgical intervention questions the negative outcomes of the study . Patient selection appears to be crucial to identify those who will benefit from this strategy. Systemic treatment when platinum is the best option Chemotherapy Platinum is considered the best treatment option for individuals who do not exhibit progression during previous treatment with platinum, do not present early symptomatic relapse, or who have no contraindication to platinum. Carboplatin doublets have an increased PFS and OS compared to carboplatin monotherapy . Likewise, association with paclitaxel, gemcitabine, or pegylated liposomal doxorubicin (PLD) have demonstrated similar efficacy, with the choice of companion agent being based on the patient's previous toxicity profile and preferences. For individuals with previous hypersensitivity to platinum, there are validated desensitization protocols that can be used under supervision if platinum is deemed the best option . The INOVATYON study compared a non-platinum doublet (trabectedin–PLD) to a platinum doublet (carboplatin–PLD) in patients with a TFIp of 6–12 months and failed to prove an increase in OS. Therefore, a platinum doublet is the preferred option in the first relapse. Maintenance treatment Antiangiogenic treatment Bevacizumab, in combination with a platinum doublet with gemcitabine or paclitaxel and then as a maintenance treatment, has demonstrated increased percentages of objective responses and PFS compared to ChT and placebo. Carboplatin–PLD–bevacizumab increased PFS and OS vs. carboplatin–gemcitabine–bevacizumab, making it the preferred option for patients who have not received prior antiangiogenic treatment . Although not authorized in Europe, in individuals with TFIp > 6 months who have previously received bevacizumab, retreatment with bevacizumab increased PFS with respect to ChT alone . PARP inhibitors Three PARPi (olaparib, niraparib, and rucaparib) have been approved as maintenance therapy for subjects with ROC and response to platinum. Olaparib significantly increased PFS vs. placebo in those with BRCA1/2 mutations in the SOLO2 study . In the NOVA study , niraparib increased PFS in the BRCA1/2 mutated population, BRCA non-mutated patients with HRD, as well as in the overall non- BRCA cohort. In the ARIEL3 study , rucaparib demonstrated better PFS in BRCA1/2 mutated cases, the HRD population, and in the intention-to-treat population. The recent communication of results of longer OS follow-up has generated controversy as to whether maintenance with PARPi could be harmful in the long-term in non-germline BRCA (g BRCA ) mutation carriers, and affect response to subsequent retreatment with platinum, thereby generating resistance . Nonetheless, OS was not a primary objective in these studies nor were they powered to do so. Consequently, the European Medicines Agency (EMA) continues to endorse niraparib and rucaparib as maintenance treatments in non-g BRCA mutation carriers. Nevertheless, the risks and benefits should be discussed with patients. Retreatment with iPARP has been explored in the OREO study which found a modest PFS benefit in selected patients; albeit it is not currently approved. Systemic treatment when platinum is not the best option In patients with ROC that develop progressive disease while on platinum-based ChT or shortly thereafter, platinum rechallenge may not be an option. Patients with a short TFIp (< 6 months) have often been treated with multiple prior lines of therapy and may be symptomatic . Priority should be given to improve symptom control. Consequently, clinicians must discuss potential risks and treatment-emergent adverse events in addition to potential benefit from therapy, and early palliative care referral should be considered . In women with a good performance status (PS), participation in clinical trials is highly recommended. Response rates to single-agent ChT, such as weekly paclitaxel, PLD, topotecan, gemcitabine, or metronomic cyclophosphamide are low, averaging 5–15% and median OS of 12 months . Studies in this setting have been designed until progression or unacceptable toxicity. There is no robust data comparing these regimens and the choice may be guided by toxicity profile and patient preferences. Yet, patients with poor PS should be considered for best supportive care only. Trabectedin–PLD is an option for patients with TFIp > 6 months unable to receive further platinum-based ChT . The AURELIA (NCT00976911) phase III trial assessed the role of bevacizumab in combination with single-agent ChT (weekly paclitaxel, PLD, or topotecan). The addition of bevacizumab resulted in improved PFS and quality of life, with no statistically significant differences in OS. Nevertheless, the study was not powered to detect these differences and 40% of patients cross-overed to bevacizumab. Weekly paclitaxel and bevacizumab demonstrated better outcomes that the other two arms . Novel therapeutics, including antibody drug conjugates have shown promising results. The MIRASOL (NCT04209855) trial demonstrated that mirvetuximab soravtansine improved PFS and OS in high FRα expression platinum-resistant ovarian cancer, when compared to single agent ChT . Approval from European and Spanish healthcare authorities is pending. Recommendations Secondary cytoreduction should be contemplated in selected patients [I, A]. For cases in which platinum is the best option, (1) bevacizumab and a platinum doublet (if no prior Bev) or (2) platinum doublet followed by PARPi maintenance (if response to platinum and no prior PARPi) regardless of BRCA and/or HRD status, can be considered [I, A]. For subjects with a highly symptomatic relapse or requiring a rapid response, combination with bevacizumab is the preferred approach [III, B]. When platinum is not an option, single-agent non-platinum alternatives can be considered [I, B]. The addition of bevacizumab should be recommended in those without contraindications (IA). Early referral to palliative care should be considered [II, A]. Follow-up, long-term implications, and survivorship There is no consensus regarding the optimal follow-up strategy for ovarian cancer survivors. Table can be used as a general guideline. Given that some patients can suffer from late relapses, extended follow-up beyond 5 years can be considered for some patients. For BRCA1/2 mutation carriers, high-risk breast cancer screening guidelines should be followed, although some retrospective studies have reported a low rate of metachronous breast cancer in EOC patients . With the prolonged survivals achieved in the era of PARPi, long-term toxicity, such as long-lasting neuropathy, sexual impairment, or myelodysplastic syndromes associated with ChT and PARPi treatments, should be carefully monitored throughout follow-up (Table ) . Recommendations As a minimum, follow-up of EOC survivors includes reviewing symptoms, physical and pelvic examination, and Ca125 until 5 years after the end of treatment [IV, B]. Long-term follow-up beyond 5 years after the end of treatment should be considered for some patients [III, B]. Long-term toxicity should be monitored throughout follow-up [V, C]. EOC is the second most deadly gynaecological cancer worldwide and the first in developed countries, responsible for some 200,000 deaths annually. In Spain, 1979 women died of OC in 2021, and approximately 3584 new cases were diagnosed during 2023 . Median age at the time of diagnosis is approximately 63 years. Nulliparity, obesity, and treatment with oestrogen therapy are known risk factors for EOC. Genital use of talc powder has been suggested as a potential risk factor. In contrast, higher parity, oral contraceptive use, and breastfeeding have a protective role. Tobacco have been studied without conclusive results . Moreover, high grade serous (HGS) carcinomas are strongly associated with family history and hereditary syndromes. Mutations in the BRCA1/2 genes, which are detected in 10–15% of patients with OC and cause Hereditary Breast and Ovarian Cancer syndrome, are associated with a 15–65% risk of EOC, especially HGS histology. Mutations in the MLH1 , MSH2 , MSH6 , and PMS2 genes (diagnostic of Lynch syndrome), also correlate with a 12% risk of developing OC, mainly endometrioid or clear cell. Mutations in the ATM , BRIP1 , PALB2 , RAD51C and RAD51D genes are associated with a moderate risk of developing OC . More than two-thirds of all cases are diagnosed at an advanced stage, given that the symptoms of early stage EOC are not specific. Two large prospective trials, the UKTCTOCS and PLCO trials that enrolled 202,562 and 78,216 women, respectively, found that a screening program for OC had no clear impact on mortality . Common symptoms of EOC include abdominal/pelvic pain, constipation, urinary frequency, abdominal distension, shortness of breath, and fatigue. The initial evaluation includes a physical examination, laboratory testing including CA 125, and pelvic ultrasound. Elevated HE4 levels identify malignancy with a sensitivity similar to that of CA 125, albeit with greater specificity. Algorithms such as the International Ovarian Tumour Analysis (IOTA) Simple Rules risk model or IOTA Assessment of Different NEoplasias in the adneXa (ADNEX) model can be useful in distinguishing benign from malignant pelvic tumours. Computed tomography (CT) imaging of the thorax, abdomen, and pelvis defines the extent of the disease and informs treatment planning. If available, magnetic resonance imaging (MRI) and positron emission tomography (PET)–CT can enhance assessment accuracy in advanced disease. Initial laparoscopy is a mainstay for histopathological diagnosis and to evaluate the likelihood of complete cytoreduction . Peritoneal extension scores, such as the peritoneal cancer index or the Fagotti score could be useful in this context. An adequate amount of tissue is essential to establish a pathological diagnosis and analyse biomarkers to guide the treatment plan. Cytological testing of ascites and pleural fluid, if present, is required to complete staging. FIGO 2014 is the staging system currently recommended (Table ). Screening for EOC in average risk women is not recommended (I, E). The initial evaluation of a patient with suspicion of EOC should include physical examination, laboratory testing with CA-125, and pelvic ultrasound [II, A]. CT of the thorax, abdomen, and pelvis is recommended in the diagnostic workup to assess the extent of the disease (II, A). Laparoscopic surgery is recommended in advanced EOC to evaluate cytoreduction, obtain material for pathologic diagnosis and predictive biomarkers (II, B). EOC is the most common ovarian cancer histology (~ 90) and can be classified into five main subtypes according to the 2020 World Health Organization (WHO) classification: high-grade serous (HGS, 70%), low grade serous (LGS, 10%), endometrioid carcinoma (EC, 10%), clear cell carcinoma (CCC, 5%), and mucinous carcinoma (MC, 3%). Other less frequent histologies (2–3%) include undifferentiated carcinoma, malignant Brenner tumour, mesonephric-like carcinoma, mixed carcinomas, or carcinosarcoma . Diagnosis should be made by an expert gynaecological pathologist by immunohistochemistry (IHC) to avoid misdiagnoses. Each EOC subtype depends on a different molecular background (Table ) that can address different pathogenesis, clinical features, response to treatments, and prognosis . BRCA1/2 mutational status should be determined at primary diagnosis in every non-mucinous EOC regardless of age at diagnosis or family history of cancer. Although less common, Lynch syndrome can be associated with mucinous and non-mucinous EOC. Mismatch repair ( MMR ) genes testing is highly recommended depending on the family history of cancer. More extensive germline next generation sequencing (NGS) panels should be offered, depending on clinical suspicion . Homologous recombination deficiency (HRD) provides prognostic information and can be used as a predictive biomarker of the magnitude of response to PARP inhibitors (PARPi). Commercially available NGS tests detect genetic scars and somatic BRCA1/2 mutations as subrogates of HRD and are the most common way to test homologous recombination status in clinics. Different cutoffs can be observed depending on the test. HRD status should be assessed at diagnosis in at least every high-grade EOC patient (i.e., serous and endometroid subtype) . Initial IHC-based diagnosis should be made by a gynaecological pathology expert [IV, A]. All patients with non-mucinous EOC should be tested for somatic and/or germline BRCA1/2 mutation at primary diagnosis to discard a hereditary syndrome, to obtain prognostic information, and to select a biomarker of response to PARPi [I, A]. Determination of HRD with a clinically validated test is strongly recommended at initial diagnosis in every high-grade serous or endometrioid EOC to provide prognostic and predictive biomarker information [I, A].MMR testing and/or more extensive germline NGS panels are recommended depending on clinical suspicion and family history of cancer [II, A]. Surgery Treatment of early EOC requires surgery including hysterectomy, bilateral salpingo-oophorectomy, systematic pelvic and para-aortic lymphadenectomy, omentectomy, appendectomy in MC, random peritoneal biopsies of all surfaces, peritoneal washings with cytological examination, and complete exploration of the peritoneal cavity (Fig. ). In FIGO stage I, low-grade endometroid and expansile mucinous due to a rate of lymph node involvement < 1% systematic pelvic and para-aortic lymphadenectomy is not recommended. The standard approach is by supra- and infra-umbilical median laparotomy. The laparoscopic approach is being investigated, but we currently lack prospective data demonstrating its equivalence to open surgery . The aim of the surgery is to remove the entire tumour and stage the disease so as to establish the indication for subsequent adjuvant treatment. Fertility-preserving surgery could be considered in stage IA or IC disease with unilateral ovarian involvement and low histologic grade . Treatment of early EOC requires surgery including hysterectomy, bilateral salpingo-oophorectomy, systematic pelvic and para-aortic lymphadenectomy, omentectomy, appendectomy in MC, random peritoneal biopsies of all surfaces, peritoneal washings with cytological examination, and complete exploration of the peritoneal cavity (Fig. ). In FIGO stage I, low-grade endometroid and expansile mucinous due to a rate of lymph node involvement < 1% systematic pelvic and para-aortic lymphadenectomy is not recommended. The standard approach is by supra- and infra-umbilical median laparotomy. The laparoscopic approach is being investigated, but we currently lack prospective data demonstrating its equivalence to open surgery . The aim of the surgery is to remove the entire tumour and stage the disease so as to establish the indication for subsequent adjuvant treatment. Fertility-preserving surgery could be considered in stage IA or IC disease with unilateral ovarian involvement and low histologic grade . Adjuvant platinum-based ChT after complete surgery will be offered to all patients diagnosed with a high-grade tumour or stage II disease, based on long-term follow-up results from the ICON1-ACTION studies. Adjuvant ChT is not recommended in completely staged patients with LGS stage IA, low-grade EC stage IA, or expansile MC stage IA–IB. The benefit of adjuvant ChT is uncertain and could be regarded as optional in LGS stage IB–IC, CCC stage IA–IC1, low-grade EC stage IB–IC, expansile MC stage IC, and infiltrative MC stage IA . Standard adjuvant treatment consists of six cycles of platinum-based ChT. However, there is no consensus regarding treatment duration or optimal schedule. The GOG 157 study compared three vs. six cycles of carboplatin and paclitaxel. The study revealed no difference in recurrence rate, although a subsequent analysis did demonstrate a benefit in HGS in favour of the longer treatment; nevertheless, this group represented only 23% of the study population . Although the most commonly used schedule is the platinum and taxane combination, there is no evidence that adding paclitaxel is of benefit in adjuvant treatment of early ovarian cancer. Thus, carboplatin monotherapy may be an appropriate option in this setting. The magnitude of benefit as well as short- and long-term toxicity should be discussed with the patient prior to deciding on the adjuvant scheme to be used . Surgery with complete tumour resection and complete surgical staging is the mainstay for both treatment and to establish the extent of the disease that informs the subsequent indication for systemic treatment (I, A). Adjuvant platinum-based ChT is recommended in all cases of high grade or stage II (IA). The most widely used option is the combination of carboplatin and paclitaxel, although carboplatin monotherapy may be acceptable (II, A). Treatment duration should be at least three cycles for all subtypes, albeit six cycles are recommended in HGS (II, A). Cytoreductive surgery The recommended strategy in individuals with stage III–IV disease is primary debulking surgery (PDS), followed by systemic treatment . Cytoreduction surgery in advanced EOC (FIGO stages III–IV) has a proven therapeutic purpose. Its goal must be complete excision (R0) of any visible tumour without leaving macroscopic residual disease, given that post-surgical volume will impact the risk of recurrence and patient survival . In light of the results of the LION and the CARACO trials, lymphadenectomy in primary fully resected AOC with clinically negative lymph nodes is no longer recommended nor in upfront nor in interval cytoreductive surgery . Surgical expertise and specialist training are known to result in better rates of complete cytoreduction. Hence, subjects with advanced disease are advised to undergo surgery in specialized centres with suitable infrastructure and trained teams . The recommended strategy in individuals with stage III–IV disease is primary debulking surgery (PDS), followed by systemic treatment . Cytoreduction surgery in advanced EOC (FIGO stages III–IV) has a proven therapeutic purpose. Its goal must be complete excision (R0) of any visible tumour without leaving macroscopic residual disease, given that post-surgical volume will impact the risk of recurrence and patient survival . In light of the results of the LION and the CARACO trials, lymphadenectomy in primary fully resected AOC with clinically negative lymph nodes is no longer recommended nor in upfront nor in interval cytoreductive surgery . Surgical expertise and specialist training are known to result in better rates of complete cytoreduction. Hence, subjects with advanced disease are advised to undergo surgery in specialized centres with suitable infrastructure and trained teams . Conventional treatment is based on a combination of carboplatin (AUC 5–6) and paclitaxel (175 mg/m 2 ) every 3 weeks for 6 cycles . Schedules with weekly ChT without bevacizumab improve neither progression-free survival (PFS) nor overall survival (OS) in patient populations from Western countries . Weekly carboplatin–paclitaxel improved QoL vs. 3-weekly suggesting a role in elder patients. However in a study in vulnerable elderly patients single-agent carboplatin or weekly ChT might have worse outcomes; consequently, the 3-week regimen remains the standard for all cases of AOC including the elderly . Prophylaxis of venous thromboembolism in advanced EOC patients should be discussed with each patient receiving systemic therapy considering their specific risk factors . The EORTC55971 trial and CHORUS trial found similar PFS and OS rates for patients with stage IIIC or IV disease receiving neoadjuvant chemotherapy (NACT) and interval debulking surgery (IDS) compared with PDS. Despite these non-inferiority results, the aforementioned trials have been criticized for their short median OS, mean operative time, and low optimal cytoreduction rates. Due to the limitations of these trials, it has yet to be determined whether NACT and IDS might be an option for people for whom complete resection at PDS seems feasible. Therefore, both approaches (PDS or NACT followed by IDS) may be deemed valid, although PDS is still the preferred primary treatment option when complete cytoreduction is feasible and patient is operable . Despite the fact that three large, randomized studies (GOG 104, GOG 114, and GOG 172) and one meta-analysis have found clinically significant improvements in PFS and OS with intraperitoneal (IP) ChT , the results of the GOG 252 trial revealed no benefit of IP therapy when bevacizumab was incorporated in all arms. Therefore, IP chemotherapy is not considered a standard of care . A randomized phase III trial evaluating hyperthermic intraperitoneal chemotherapy (HIPEC) after IDS evidenced better PFS and OS for the HIPEC arm. Nevertheless, this trial received notable methodological criticisms, in particular, the lack of stratification for known prognostic molecular factors. Therefore, HIPEC cannot be regarded as a standard treatment and should not be offered outside of the context of clinical trials . Two large, randomized studies (GOG 218 and ICON 7) have reported that bevacizumab (15 mg/kg or 7.5 mg/kg every 3 weeks) added to adjuvant ChT after PDS and followed by maintenance therapy with bevacizumab for a maximum of 15 months improves PFS compared to standard adjuvant ChT alone. Post-hoc subgroup analyses indicated statistically significant OS benefit only in patients with stage IV disease in GOG 218 and patients at high risk of progression (defined as FIGO stage III with > 1 cm residual disease following PDS or stage IV) in the ICON7 trial . In the ENGOT Ov-15 study, prolonging the duration of bevacizumab administration (30 vs. 15 months) failed to improve PFS . Two small, prospective trials revealed that bevacizumab added to platinum-containing ChT in the neoadjuvant setting was safe, although it has no impact on complete resection rate or PFS . Several phase III, randomized clinical trials (SOLO 1, PRIMA, PAOLA, PRIME, and ATHENA–MONO) have demonstrated that maintenance therapy with PARPi (olaparib, niraparib, olaparib plus bevacizumab, and rucaparib) after response to front-line platinum-containing regimens significantly increased median PFS in HGSOC . All trials have manifested a remarkable, unprecedented benefit inBRCA1/2 mutated individuals. Moreover, olaparib–bevacizumab, niraparib, and rucaparib also displayed a significant benefit in the HRD population. Finally, niraparib and rucaparib exhibited a benefit in the HR proficient (HRP) and HRD-unknown subgroups, albeit of lesser magnitude. The benefit observed with PARPi has been sustained throughout follow-up as demonstrated by their impact on PFS2, as well as by the results of long-term overall survival in the SOLO-1 and PAOLA-1 trials, with survival rates of 67% at 7 years and 65.5% at 5 years in the experimental arm vs. 46.5% and 48.4% in the control arm, respectively. The benefit was observed despite the fact that 40% of the subjects in the control group received subsequent PARP therapy . ATHENA-MONO trial also showed that rucaparib imrpoved PFS in all populations regardless of BRCA or HRD status having received a recent EMA approval for the indication of maintenance in first line. Based on these results, olaparib (with or without bevacizumab), rucaparib or niraparib after partial or complete response to first-line, platinum-based ChT are highly effective in BRCA1/2-mutated patients and are strongly recommended. According to the outcomes of the PAOLA-1 and PRIMA and ATHENA trials, niraparib, rucaparib or olaparib–bevacizumab are also highly recommended for women with HRD tumours. In the HRP subgroup, maintenance with niraparib, or rucaparib can also be considered, although bevacizumab remains a reasonable alternative. The choice of maintenance treatment should be based on: (1) molecular biomarkers ( BRCA1/2 and HRD status), (2) disease-related factors (stage at diagnosis, post-surgical residual disease, response to ChT, chemotherapy response score (CRS), CA-125 ELIMination Rate Constant K (KELIM)), and (3) patient characteristics (comorbidities, concomitant medication). Paclitaxel–carboplatin ± bevacizumab is the standard systemic ChT used in non-HGSC . Multiple retrospective studies, however, demonstrated lower response rates in these histologic subtypes compared with HGSC . Due to the lower chemosensitivity, PDS is strongly recommended in uncommon histologies. Most LGSCs have high expression of oestrogen (ER) and progesterone receptors (PgR). Retrospective studies suggest a possible therapeutic value of hormone therapy in the maintenance of newly diagnosed advanced LGSC [IV, B]. Uncommon non-HGS histologies are an unmet need and inclusion of these patients in clinical trials is fervently encouraged. Cytoreductive surgery aimed at achieving complete cytoreduction (absence of all visible residual disease) is the backbone of treatment for advanced EOC and should be performed in specialised centres [II, A]. Lymphadenectomy in primary, completely debulked AOC with clinically negative lymph nodes is not recommended [I, A]. Standard treatment is based on a combination of carboplatin (AUC 5–6) and paclitaxel (175 mg/m 2 ) every 3 weeks for 6 cycles [I, A]. If complete cytoreductive surgery is not feasible, NACT for three cycles followed by ICS and three cycles of ChT is recommended [I, A]. IP ChT and HIPEC are not considered standard of care [I, D]. The addition of bevacizumab to ChT should be contemplated, especially in patients with stage III and residual disease or stage IV [I, A]. Post-ChT maintenance treatment is recommended after in HG-EOC. PARPi are recommended only after partial or complete response or no evidence of disease after first-line platinum-based ChT: o In BRCA1/2-mutated subjects: single agent with olaparib, rucaparib or niraparib or combination with olaparib plus bevacizumab [I, A] o HRD tumours: niraparib, rucaparib or olaparib–bevacizumab [I, A] o HRP tumours: Niraparib, rucaparib [I, B] or bevacizumab [I, A] Paclitaxel–carboplatin ± bevacizumab is the standard systemic ChT used in non-HGSC [I, B]. In newly diagnosed advanced LGSC maintenance treatment after Cht with hormone therapy can be considered [IV, B]. Approximately 85% of all individuals diagnosed with advanced EOC experience disease recurrence within 10 years. Choosing the optimal strategy for recurrent ovarian cancer (ROC) demands that several critical factors be evaluated. Treatment-free interval (TFI) following last-platinum (TFIp) therapy remains a pivotal prognostic factor. Nevertheless, TFI should not be the sole consideration when making clinical decisions, since other clinical and molecular characteristics can impact treatment response and must be pondered . Factors to consider in treatment assessment are histological subtype, BRCA1/2 status, extension of the disease and symptoms, feasibility and outcome of a potential second surgery, prior treatment and response, TFI from the last treatment, residual toxicity from previous therapeutic interventions, patient condition and comorbidities, and patient preferences . The role of secondary cytoreduction (SC) in patients with first relapse more than 6 month TFIp has been examined in the DESKTOP III trial . SC followed by ChT in cases selected on the basis of a favourable AGO score defined as a good performance status (ECOG 0), no residual disease after PDS, and the absence of ascites (< 500 ml) demonstrated an improvement in PFS and OS compared to ChT alone. The SOC-1 trial also assessed the role of SC in patients selected by an iModel algorithm. This trial evidenced a benefit in PFS from surgery, but OS data are still immature . In contrast, the GOG-0213 trial failed to prove superiority of SC. The absence of well-defined selection criteria for surgical intervention questions the negative outcomes of the study . Patient selection appears to be crucial to identify those who will benefit from this strategy. Chemotherapy Platinum is considered the best treatment option for individuals who do not exhibit progression during previous treatment with platinum, do not present early symptomatic relapse, or who have no contraindication to platinum. Carboplatin doublets have an increased PFS and OS compared to carboplatin monotherapy . Likewise, association with paclitaxel, gemcitabine, or pegylated liposomal doxorubicin (PLD) have demonstrated similar efficacy, with the choice of companion agent being based on the patient's previous toxicity profile and preferences. For individuals with previous hypersensitivity to platinum, there are validated desensitization protocols that can be used under supervision if platinum is deemed the best option . The INOVATYON study compared a non-platinum doublet (trabectedin–PLD) to a platinum doublet (carboplatin–PLD) in patients with a TFIp of 6–12 months and failed to prove an increase in OS. Therefore, a platinum doublet is the preferred option in the first relapse. Platinum is considered the best treatment option for individuals who do not exhibit progression during previous treatment with platinum, do not present early symptomatic relapse, or who have no contraindication to platinum. Carboplatin doublets have an increased PFS and OS compared to carboplatin monotherapy . Likewise, association with paclitaxel, gemcitabine, or pegylated liposomal doxorubicin (PLD) have demonstrated similar efficacy, with the choice of companion agent being based on the patient's previous toxicity profile and preferences. For individuals with previous hypersensitivity to platinum, there are validated desensitization protocols that can be used under supervision if platinum is deemed the best option . The INOVATYON study compared a non-platinum doublet (trabectedin–PLD) to a platinum doublet (carboplatin–PLD) in patients with a TFIp of 6–12 months and failed to prove an increase in OS. Therefore, a platinum doublet is the preferred option in the first relapse. Antiangiogenic treatment Bevacizumab, in combination with a platinum doublet with gemcitabine or paclitaxel and then as a maintenance treatment, has demonstrated increased percentages of objective responses and PFS compared to ChT and placebo. Carboplatin–PLD–bevacizumab increased PFS and OS vs. carboplatin–gemcitabine–bevacizumab, making it the preferred option for patients who have not received prior antiangiogenic treatment . Although not authorized in Europe, in individuals with TFIp > 6 months who have previously received bevacizumab, retreatment with bevacizumab increased PFS with respect to ChT alone . PARP inhibitors Three PARPi (olaparib, niraparib, and rucaparib) have been approved as maintenance therapy for subjects with ROC and response to platinum. Olaparib significantly increased PFS vs. placebo in those with BRCA1/2 mutations in the SOLO2 study . In the NOVA study , niraparib increased PFS in the BRCA1/2 mutated population, BRCA non-mutated patients with HRD, as well as in the overall non- BRCA cohort. In the ARIEL3 study , rucaparib demonstrated better PFS in BRCA1/2 mutated cases, the HRD population, and in the intention-to-treat population. The recent communication of results of longer OS follow-up has generated controversy as to whether maintenance with PARPi could be harmful in the long-term in non-germline BRCA (g BRCA ) mutation carriers, and affect response to subsequent retreatment with platinum, thereby generating resistance . Nonetheless, OS was not a primary objective in these studies nor were they powered to do so. Consequently, the European Medicines Agency (EMA) continues to endorse niraparib and rucaparib as maintenance treatments in non-g BRCA mutation carriers. Nevertheless, the risks and benefits should be discussed with patients. Retreatment with iPARP has been explored in the OREO study which found a modest PFS benefit in selected patients; albeit it is not currently approved. In patients with ROC that develop progressive disease while on platinum-based ChT or shortly thereafter, platinum rechallenge may not be an option. Patients with a short TFIp (< 6 months) have often been treated with multiple prior lines of therapy and may be symptomatic . Priority should be given to improve symptom control. Consequently, clinicians must discuss potential risks and treatment-emergent adverse events in addition to potential benefit from therapy, and early palliative care referral should be considered . In women with a good performance status (PS), participation in clinical trials is highly recommended. Response rates to single-agent ChT, such as weekly paclitaxel, PLD, topotecan, gemcitabine, or metronomic cyclophosphamide are low, averaging 5–15% and median OS of 12 months . Studies in this setting have been designed until progression or unacceptable toxicity. There is no robust data comparing these regimens and the choice may be guided by toxicity profile and patient preferences. Yet, patients with poor PS should be considered for best supportive care only. Trabectedin–PLD is an option for patients with TFIp > 6 months unable to receive further platinum-based ChT . The AURELIA (NCT00976911) phase III trial assessed the role of bevacizumab in combination with single-agent ChT (weekly paclitaxel, PLD, or topotecan). The addition of bevacizumab resulted in improved PFS and quality of life, with no statistically significant differences in OS. Nevertheless, the study was not powered to detect these differences and 40% of patients cross-overed to bevacizumab. Weekly paclitaxel and bevacizumab demonstrated better outcomes that the other two arms . Novel therapeutics, including antibody drug conjugates have shown promising results. The MIRASOL (NCT04209855) trial demonstrated that mirvetuximab soravtansine improved PFS and OS in high FRα expression platinum-resistant ovarian cancer, when compared to single agent ChT . Approval from European and Spanish healthcare authorities is pending. Secondary cytoreduction should be contemplated in selected patients [I, A]. For cases in which platinum is the best option, (1) bevacizumab and a platinum doublet (if no prior Bev) or (2) platinum doublet followed by PARPi maintenance (if response to platinum and no prior PARPi) regardless of BRCA and/or HRD status, can be considered [I, A]. For subjects with a highly symptomatic relapse or requiring a rapid response, combination with bevacizumab is the preferred approach [III, B]. When platinum is not an option, single-agent non-platinum alternatives can be considered [I, B]. The addition of bevacizumab should be recommended in those without contraindications (IA). Early referral to palliative care should be considered [II, A]. There is no consensus regarding the optimal follow-up strategy for ovarian cancer survivors. Table can be used as a general guideline. Given that some patients can suffer from late relapses, extended follow-up beyond 5 years can be considered for some patients. For BRCA1/2 mutation carriers, high-risk breast cancer screening guidelines should be followed, although some retrospective studies have reported a low rate of metachronous breast cancer in EOC patients . With the prolonged survivals achieved in the era of PARPi, long-term toxicity, such as long-lasting neuropathy, sexual impairment, or myelodysplastic syndromes associated with ChT and PARPi treatments, should be carefully monitored throughout follow-up (Table ) . As a minimum, follow-up of EOC survivors includes reviewing symptoms, physical and pelvic examination, and Ca125 until 5 years after the end of treatment [IV, B]. Long-term follow-up beyond 5 years after the end of treatment should be considered for some patients [III, B]. Long-term toxicity should be monitored throughout follow-up [V, C]. |
Prognostic implications of immunohistochemistry in patients with endometrial cancer | 60f32d08-fea5-4135-908c-0f01647d3255 | 11384858 | Anatomy[mh] | Endometrial carcinoma (EC) stands as the prevailing gynecological malignancy afflicting American women, and in 2022, it is estimated to have contributed to approximately 39 300 new cases and 6600 fatalities . Many women with early-stage EC experience recurrence and eventually pass away from the illness, despite receiving definitive surgical treatment. Factors such as specific histological cell types, high histological grade, extensive myometrial invasion, and the involvement of lymphovascular pathways are established risk determinants for disease progression . Obesity, unopposed estrogen states (such as polycystic ovarian syndrome), early menarche or late menopause and genetic cancer syndromes such as Lynch syndrome (LS) and Cowden syndrome are all risk factors for EC . Conversely, protective factors against EC include experiencing multiple pregnancies (multiparity) and use of oral contraceptives. Of particular significance is LS, an inherited condition stemming from germline mutations within genes responsible for deoxyribonucleic acid (DNA) mismatch repair (MMR). This syndrome accounts for approximately 3% of all instances of endometrial malignancies . Individuals harboring mutations in genes such as MutL homolog 1 (MLH1), MutS homolog 2 (MSH2), MutS homolog 6 (MSH6), or postmeiotic segregation increased 2 (PMS2) face an escalated risk for both endometrial and colorectal cancers, with a lifetime risk ranging from 40% to 60% . Furthermore, these individuals carry a lifetime susceptibility of 9% to 12% for ovarian cancer . In recent years, there has been a notable upswing in the integration of molecular analysis and tailored molecular-based therapies in the context of patient-centered care for EC . This reflects the growing emphasis on leveraging molecular insights to inform treatment strategies, optimizing the management of this complex malignancy. Aim The aim of this study was to investigate potential correlations between specific immunohistochemical (IHC) characteristics and post-surgical prognosis in patients diagnosed with EC. By analyzing the expression patterns of these IHC markers, we aimed to elucidate their potential prognostic significance in predicting disease outcomes following surgical intervention. Through comprehensive examination and statistical analysis of patient samples, we seek to identify any associations between the expression levels of these markers and key clinical parameters, such as disease recurrence, overall survival, and response to treatment. Ultimately, our goal was to contribute valuable insights into the development of more accurate prognostic tools and personalized treatment strategies for patients undergoing surgical management of EC. The aim of this study was to investigate potential correlations between specific immunohistochemical (IHC) characteristics and post-surgical prognosis in patients diagnosed with EC. By analyzing the expression patterns of these IHC markers, we aimed to elucidate their potential prognostic significance in predicting disease outcomes following surgical intervention. Through comprehensive examination and statistical analysis of patient samples, we seek to identify any associations between the expression levels of these markers and key clinical parameters, such as disease recurrence, overall survival, and response to treatment. Ultimately, our goal was to contribute valuable insights into the development of more accurate prognostic tools and personalized treatment strategies for patients undergoing surgical management of EC. We identified a total of 58 patients who had received a pathological diagnosis of EC through a retrospective analysis. Surgical staging procedures were carried out on all patients at the Prof. Dr. Panait Sîrbu Clinical Hospital of Obstetrics and Gynecology, Bucharest, Romania, between 2020 and 2022. Approval for this research inquiry was obtained from the Institutional Review Boards of the Prof. Dr. Panait Sîrbu Clinical Hospital of Obstetrics and Gynecology. A prospective study was conducted to collect socio-demographic, lifestyle, and medical data. Clinical information was extracted from medical and pathology records, including details about histological subtypes (endometrioid, serous, clear cell, carcinosarcoma), tumor stage (I, II, III, and IV), tumor grade (1, 2, and 3), presence of lymphovascular space invasion (LVSI; yes, no/unknown), and adjuvant therapy (brachytherapy, chemotherapy, radiotherapy; whether any of these treatments were administered or not). The staging of cases was performed according to the criteria established by the International Federation of Gynecology and Obstetrics (Fédération Internationale de Gynécologie et d’Obstétrique – FIGO) in 2009. Each slide was individually assessed to confirm the diagnosis, histological subtype, and grade. Further in our study, we performed IHC staining on 4 μm thick whole tissue sections using the following antibodies: anti-estrogen receptor (ER, clone 6F11), anti-progesterone receptor (PR, clone 1A6), anti-p53 (clone DO7), anti-Ki67 (clone MIB1), anti-MLH1 (clone M1), anti-MSH2 (clone G219-1129), anti-MSH6 (clone SP93), and anti-PMS2 (clone A16-4), following the manufacturer’s guidelines. For p53 immunostaining, we categorized our interpretation into two groups: wild-type or mutated, encompassing overexpression or null patterns. Regarding the IHC expression of MLH1, MSH2, MSH6, and PMS2, adjacent normal tissue and surrounding tissue lymphocytes were utilized as internal positive controls for each case. The classification of MMR immunohistochemistry was also divided into two categories: intact or deficient, based on the nuclear staining of the tumor compared to the corresponding internal control. To quantify the expressions of ER, PR, and Ki67, we calculated the percentage of positive tumoral nuclei. In cases with suboptimal specimen fixation, as occasionally observed in hysterectomy specimens, staining patterns displayed shifts from well-fixed to less well-fixed areas. In these cases, the interpretation of staining predominantly depended on the well-fixed regions. A total of 58 patients underwent evaluation, with only 21 of them undergoing immunohistochemistry tests on the specimens following surgical intervention. The purpose of these tests was to determine whether there is a correlation between the expression of specific IHC characteristics and the prognosis of the disease after surgery. Patients’ characteristics, surgical interventions, follow-up Comprehensive follow-up data was available for all the patients. Out of the total sample size of 58 patients, a subset of 21 individuals had a satisfactory immunohistochemistry analysis, while the remaining patients did not undergo this specific diagnostic test. The mean age at the diagnostic of the disease for the entire sample was 62.1±9.53 years (Table ). The average body mass index (BMI) in this group was 31.91 kg/m2, ranging from 20.7 kg/m2 to 44.4 kg/m2, which typically falls into the category of degree I obesity. At the initial diagnosis, only 10 individuals in the patient cohort were identified as smokers. The primary presenting symptom prompting medical consultation was post-menopausal metrorrhagia, which occurred in around 86.2% of the cases. Two individuals were diagnosed with breast cancer and subsequently had surgical intervention. Additionally, they were receiving Tamoxifen medication as part of their treatment regimen at the moment of presentation. Table offers detailed information in relation to the patients’ personal medical backgrounds. The major surgical technique that occurred most often was laparotomy for total hysterectomy with bilateral salpingo-oophorectomy, accounting for 58.62% of cases (Figure ). Laparoscopic total hysterectomy with bilateral salpingo-oophorectomy was performed in 37.93% of the cases. Two cases had class B1 radical hysterectomy. In a total of eight cases, the sentinel lymph node was solely identified and examined by laparoscopic procedures, employing either Indocyanine Green or Methylene Blue as the contrasting agent. A total of 30 individuals had bilateral pelvic lymphadenectomy, which was done using either laparoscopic or laparotomy techniques. All cases were conducted without any difficulties during the surgical procedure, and the immediate postoperative progress was favorable. The macroscopic examination of hysterectomy specimens revealed that the tumor was diffuse throughout the uterine cavity in 87.7% of cases and developed in a polyp in 12% of cases (Figure , ). According to histological analysis, endometrioid adenocarcinoma was the predominant histological type, accounting for 87.93% of cases. Non-endometrioid carcinomas comprised 10 (10.34%) cases, including four cases of serous carcinoma and two cases of clear cell carcinoma (Figure , , , ). Myometrial invasion was observed in all cases, with it exceeding 50% of the myometrial thickness in 42 (72.41%) cases. Seven cases showed lymphovascular invasion, and five cases showed perineural invasion was observed in five of these cases. In the majority (96.5%) of cases, the characteristic myometrial infiltration pattern was classical. However, one case exhibited an atypical infiltration pattern known as the MELF pattern, characterized by microcystic, elongated, and fragmented glands. This particular case involved a grade 2 endometrioid carcinoma. Furthermore, in two (13%) cases, we observed the colonization of adenomyosis foci. The tumors were classified based on the World Health Organization (WHO) criteria. Among the endometrioid malignancies, 26 were categorized as grade 1 (G1), 22 were classified as grade 2 (G2), and 10 were designated as grade 3 (G3). The assignment of FIGO stages to all patients was determined by evaluating intraoperative and pathological evidence. In our series, tumor staging revealed 40 cases at stage IA, 16 at stage IB, two at stage II at the time of the diagnostic. In the conducted investigation, it was shown that none of the patients needed neoadjuvant therapy. In two cases, radiation as a standalone treatment modality sufficed, whereas in one case brachytherapy alone was considered sufficient. In 10 cases, the administration of radiotherapy was required, either in conjunction with chemotherapy in eight cases or with brachytherapy in eight cases. The follow-up duration for our study ranged from one to four years. At the point of last communication, there were 54 individuals who were alive without any discernible signs of illness, while two individuals were still alive but had experienced a documented relapse (vaginal recurrence). Additionally, one individual had passed away due to the condition, while another individual had succumbed to circumstances unrelated to the sickness [coronavirus disease 2019 (COVID-19) infection]. Immunohistochemical analysis As previously mentioned, immunohistochemistry was conducted in 21 cases. A comprehensive assessment was conducted on a set of markers, including ER (Figure ), PR, Ki67, vimentin, MLH1, MSH2, MSH6, PMS2, p53. The maximum count seen in each specimen was recorded: 71.42% had positive staining for ER, 57.14% had positive staining for PR, 8.62% displayed mutated pattern for p53 (Figure , ) and Ki67 (Figure , ) varied between 15% and 80%. The deficiency of MMR status was determined when the IHC analysis revealed a total absence of nuclear expression in carcinoma cells for one or more MMR proteins, including MLH1, MSH2, MSH6, and PMS2, indicating microsatellite instability (MSI): seven cases showed loss of both MLH1 and PMS2, three cases showed loss of both MSH2 and MSH6 and one case showed isolated loss of PMS2. Conversely, in the rest of the cases all four MMR proteins showed nuclear expression on IHC, suggesting a microsatellite stable (MSS) status (Figure ). A study using the χ2 (chi-squared) test (Pearson’s correlation) was conducted to examine the relationship between each IHC marker and clinical outcome, specifically in relation to treatment with radiation, chemotherapy, or brachytherapy, as well as the occurrence of recurrence or mortality. The analysis included variables such as FIGO stage of disease, tumor grade and each individual IHC marker. Upon doing individual analyses of each marker, it was determined that there exists no statistically significant association (p>0.05) between the ER, PR, p53 immunomarkers and the likelihood of receiving postoperative treatment including radiation, chemotherapy, or brachytherapy and also there was no statistically significant correlations between these immunomarkers and the probability of disease recurrence. No statistically significant associations were identified between the presence of MLH1 or PMS2 and the aforementioned variables. No statistically significant link was identified between the FIGO stage or tumor grade and the same factors, as per the investigation conducted to explore this relationship. In contrast, concerning the Ki67 immunomarker, our examination using the χ2 test unveiled a statistically significant relationship between Ki67 immunoexpression and the requirement for brachytherapy as part of the treatment regimen (p=0.027). However, no statistically significant associations were detected between Ki67 immunoexpression and the administration of chemotherapy or radiotherapy, nor were any significant links established between Ki67 immunoexpression and the occurrence of disease recurrences. After analysis of the findings related to MSH2, it becomes evident that a statistically significant association exists between the presence of this marker and the administration of chemotherapy (p=0.017) and brachytherapy (p=0.033). However, no statistically significant correlation is seen between MSH2 and chemotherapy or the likelihood of recurrence. Similar findings of statistical significance were seen in cases with MSH6 positive – a statistically significant association between expression of the marker and chemotherapy (p=0.017) and brachytherapy (p=0.033). As previously mentioned, immunohistochemistry was conducted in 21 cases. A comprehensive assessment was conducted on a set of markers, including ER (Figure ), PR, Ki67, vimentin, MLH1, MSH2, MSH6, PMS2, p53. The maximum count seen in each specimen was recorded: 71.42% had positive staining for ER, 57.14% had positive staining for PR, 8.62% displayed mutated pattern for p53 (Figure , ) and Ki67 (Figure , ) varied between 15% and 80%. The deficiency of MMR status was determined when the IHC analysis revealed a total absence of nuclear expression in carcinoma cells for one or more MMR proteins, including MLH1, MSH2, MSH6, and PMS2, indicating microsatellite instability (MSI): seven cases showed loss of both MLH1 and PMS2, three cases showed loss of both MSH2 and MSH6 and one case showed isolated loss of PMS2. Conversely, in the rest of the cases all four MMR proteins showed nuclear expression on IHC, suggesting a microsatellite stable (MSS) status (Figure ). A study using the χ2 (chi-squared) test (Pearson’s correlation) was conducted to examine the relationship between each IHC marker and clinical outcome, specifically in relation to treatment with radiation, chemotherapy, or brachytherapy, as well as the occurrence of recurrence or mortality. The analysis included variables such as FIGO stage of disease, tumor grade and each individual IHC marker. Upon doing individual analyses of each marker, it was determined that there exists no statistically significant association (p>0.05) between the ER, PR, p53 immunomarkers and the likelihood of receiving postoperative treatment including radiation, chemotherapy, or brachytherapy and also there was no statistically significant correlations between these immunomarkers and the probability of disease recurrence. No statistically significant associations were identified between the presence of MLH1 or PMS2 and the aforementioned variables. No statistically significant link was identified between the FIGO stage or tumor grade and the same factors, as per the investigation conducted to explore this relationship. In contrast, concerning the Ki67 immunomarker, our examination using the χ2 test unveiled a statistically significant relationship between Ki67 immunoexpression and the requirement for brachytherapy as part of the treatment regimen (p=0.027). However, no statistically significant associations were detected between Ki67 immunoexpression and the administration of chemotherapy or radiotherapy, nor were any significant links established between Ki67 immunoexpression and the occurrence of disease recurrences. After analysis of the findings related to MSH2, it becomes evident that a statistically significant association exists between the presence of this marker and the administration of chemotherapy (p=0.017) and brachytherapy (p=0.033). However, no statistically significant correlation is seen between MSH2 and chemotherapy or the likelihood of recurrence. Similar findings of statistical significance were seen in cases with MSH6 positive – a statistically significant association between expression of the marker and chemotherapy (p=0.017) and brachytherapy (p=0.033). The joint effort on EC conducted by The Cancer Genome Atlas (TCGA) has successfully discovered four unique prognostic subtypes of EC . These subtypes were determined based on genetic anomalies, holding considerable implications for the potential tailoring of adjuvant therapy with greater precision. The molecular categorization demonstrates a significant association with patient prognosis and has the potential to enhance the identification of early-stage patients who might potentially derive benefits from adjuvant therapy. Nevertheless, the implementation of genomic technologies, such as genome sequencing, can incur significant costs and present technical challenges, particularly in the context of extracting DNA from tissue samples . These four EC subgroups are defined as follows: (1) DNA polymerase epsilon (POLE) ultra-mutated: characterized by an exceptionally high mutation burden in the exonuclease domain of the POLE gene, resulting in the inactivation of POLE and the failure of DNA proofreading during replication. (2) MSI hypermutated: exhibiting a heightened level of MSI, which contributes to a high mutation rate. (3) Copy-number low (no specific molecular profile – NSMP): this subgroup lacks a distinct molecular profile and is identified by the preservation of p53 and MMR IHC expression. (4) Copy-number high (p53 abnormal – p53abn): characterized by abnormal p53 IHC expression, including complete loss and/or overexpression of the p53 protein . The mismatch repair deficient (MMRd) tumor phenotype accounts for around 17–33% of all cases of endometrial malignancies . Initially, the molecular profile in question was seen in individuals diagnosed with LS, a hereditary condition associated with a 60% lifetime probability of developing EC . Moreover, the presence of a particular molecular genetic modification, arising from a malfunction in the human (h)MLH1, hMSH2, hMSH6 or hPMS2 MMR genes, has been seen in sporadic tumors referred to as MMRd or Lynch-like syndrome tumors. Tumors that do not exhibit any abnormalities in the MMR genes are categorized as mismatch repair proficient (MMRp) . Recent research has established that employing aberrant p53 immunohistochemistry is a reliable approach for identifying cases with tumor protein p53 (TP53) gene mutations in endometrial malignancies . These mutations are found in around 25% of all EC cases . The use of abnormal p53 IHC in biopsies has shown a specificity of 94% and a sensitivity of 91% . The prevalence of POLE mutations in endometrial malignancies is 8.59%, with a majority of these changes being observed in early stages (I–II) at a rate of 89.51% . Additionally, a significant proportion of these mutations are found in tumors of the highest grade (3), accounting for 51.53% of cases . Several IHC markers have been investigated in previous studies, including ER, PR, human epidermal growth factor receptor 2 (HER2) and Ki67. However, none of these markers have demonstrated consistent enough results to be routinely used for the subclassification of EC . Over the course of the last decade, there has been notable progress in both the study and practical use of immunohistochemistry markers across a range of malignant and non-malignant medical conditions. Consequently, it is plausible that each of the markers under investigation has the potential to indicate a higher likelihood of malignancy if its expression deviates from normal . Moreover, these markers might potentially be targeted in the future for therapy that is specifically customized at the molecular level (Table ). During the early 1990s the primary emphasis in the study of hereditary nonpolyposis colorectal cancer (HNPCC) progression was directed towards investigating mutations occurring in the MLH1 and MSH2 genes . Initially, it was hypothesized that the four MMR proteins only operated as a heterodimer complex . In the event that one of the two proteins (MLH1 with PMS2 and MSH2 with MSH6) was absent, the complex would cease to function, and the expression of the other protein would be inhibited [ . Therefore, the use of four proteins in immunohistochemistry enhances the detection of instances with MMR deficiency. Furthermore, Goodfellow et al. conducted a study with a cohort of 1002 patients diagnosed with EC . Their findings revealed that the most prevalent problem in the DNA MMR system was the loss of MLH1, followed by combined losses of MSH2 and MSH6, and then isolated loss of MSH6 . Multiple papers have shown a higher occurrence of MSH6 mutations in patients with EC compared to patients with HNPCC tumors . Therefore, it is essential to do IHC screening on individuals diagnosed with EC using the four specified proteins . The correlation between MMR deficient tumors and outcomes in women diagnosed with EC remains incompletely established. Several research has shown a significant increase in survival rates among women with MMR deficient tumors . However, other studies have reported worse outcomes or no discernible changes . The findings of subsequent studies with varying designs assessing survival using multivariate analysis reported inconsistent results regarding these associations . On the other hand, one study, which analyzed 109 patients with endometrioid and non-endometrioid subtypes of EC, found no association between tumor MMR-deficiency and survival . In contrast, another study involving 191 patients reported improved survival for MMR-deficient EC based on immunohistochemistry results . Furthermore, no association between tumor MMR class and outcome was observed in a study of 1024 patients with epigenetic MMR defective EC or “probable MMR mutation” . This latest investigation involving 466 women revealed that endometrioid MLH1-methylated MMR-deficient EC cases exhibited a significantly lower recurrence-free survival rate according to the univariate analysis (p<0.001) . Previous studies have shown that MSH6 may serve as a possible prognostic indicator in EC. Specifically, elevated levels of MSH6 in hysterectomy tissue have been linked to unfavorable outcomes and the presence of non-endometrioid subtypes . Emerging evidence suggests that MSH6 serves as an independent prognostic factor for survival outcomes in a subset of EC patients. Specifically, individuals falling within this category may face increased risks of disease recurrence despite their low-grade histological profile. Several investigations have demonstrated that the presence of MSH6 in endometrioid low-grade histology serves as an independent prognostic factor for unfavorable survival outcomes . Recognizing the prognostic significance of MSH6 expression in endometrioid low-grade histology has clinical implications. Despite their histological classification, patients with MSH6 expression may warrant more vigilant follow-up and a reconsideration of treatment strategies. This subgroup represents a minority, approximately 7% of the population , but their increased risk of recurrence underscores the need for individualized care. These findings emphasize the importance of personalized treatment strategies and closer surveillance for individuals falling into this category. In summary, a thorough evaluation of MSH6 may hold the potential to enhance the prognostic assessment of pre-operative EC patients, ultimately aiding in the stratification of patients for invasive surgical procedures and additional therapeutic interventions . While our current investigation did not establish a definitive correlation between MMR status, specifically MSH2 or MSH6, and the overall prognosis of the disease, we did identify a notable association between these factors and the likelihood of requiring chemotherapy or brachytherapy. In relation to their correlation with other clinico-pathological factors, it was observed that ER exhibited a significant association with grade 1–2 tumors . Conversely, positive PR expression demonstrated a significant association with various favorable prognostic factors, such as reduced myometrial invasion, endometrioid histology, grade 1–2 tumors and the absence of lymph node involvement . Numerous studies have shown the correlation between ER and PR expression and several favorable clinicopathological characteristics, such as endometrioid histology, well-differentiated tumor, and limited myometrial invasion . Nevertheless, our investigation was unable to establish a correlation between the expressions of ER, PR, and other prognostic variables. The presence of a p53 mutation is a significant prognostic indicator for both serous and endometrioid tumors . The determination of TP53 gene mutation status was conducted using immunohistochemistry to assess p53 expression, categorizing it as either wild-type or mutant. Previous research has highlighted a noteworthy correlation between p53 expression and high-grade tumors, disease stage, cervical involvement, and adnexal involvement. Additionally, it has been observed that endometrioid carcinoma with wild type p53 exhibited a two-year disease-free survival rate of 100%, whereas the p53 mutant variety had a rate of 86.2% . In the current study, approximately 15.9% of patients were reclassified into the high-risk category, necessitating the consideration of chemotherapy and radiation treatment. This reclassification was based on the assessment of their p53 mutation status, in accordance with the European Society of Gynecologic Oncology (ESGO)–European Society of Radiation Therapy and Oncology (ESTRO) 2021 Consensus classification . However, in the present investigation, we did not identify any statistically significant correlation between p53 expression and prognostic variables. Study limitations The extent of our research is limited due to the small number of patients included in our study, and an even smaller subset underwent IHC analysis. As a result, the applicability of our findings to a more extensive range of patients is restricted. To validate our results effectively, it is imperative to conduct randomized trials involving larger and more diverse patient cohorts. While our study of EC and IHC analysis provides valuable insights, it is important to acknowledge its limitations. Primarily, the study was conducted at a single center, potentially limiting the generalizability of our findings to broader populations. The small number of subjects included in the study further restricts the statistical power and may introduce biases that could impact the robustness of our conclusions. Additionally, the retrospective nature of the study design poses inherent limitations, such as the reliance on existing medical records and potential inconsistencies in data collection. Furthermore, variations in laboratory techniques and interpretations of IHC staining across different pathologists may introduce variability in the results. Despite these limitations, our study serves as a foundation for future research endeavors aimed at validating our findings in larger, multicenter cohorts and exploring additional factors that may influence the prognostic significance of IHC markers in EC. The extent of our research is limited due to the small number of patients included in our study, and an even smaller subset underwent IHC analysis. As a result, the applicability of our findings to a more extensive range of patients is restricted. To validate our results effectively, it is imperative to conduct randomized trials involving larger and more diverse patient cohorts. While our study of EC and IHC analysis provides valuable insights, it is important to acknowledge its limitations. Primarily, the study was conducted at a single center, potentially limiting the generalizability of our findings to broader populations. The small number of subjects included in the study further restricts the statistical power and may introduce biases that could impact the robustness of our conclusions. Additionally, the retrospective nature of the study design poses inherent limitations, such as the reliance on existing medical records and potential inconsistencies in data collection. Furthermore, variations in laboratory techniques and interpretations of IHC staining across different pathologists may introduce variability in the results. Despite these limitations, our study serves as a foundation for future research endeavors aimed at validating our findings in larger, multicenter cohorts and exploring additional factors that may influence the prognostic significance of IHC markers in EC. Although the current study could not definitively uncover an IHC marker that may effectively differentiate prognostic implications, our research efforts are ongoing in order to identify molecular markers that may better fulfill this objective. The use of tumor banking and collaborative efforts among different institutions should enhance the meaningfulness and statistical power of IHC assessment in a substantial percentage of early-stage endometrial malignancies. It is our belief that the molecular profiling of tumors in individual patients has the potential to enhance prognostication and customize therapy, hence enabling the development of more targeted and less harmful treatment approaches. The research suggests that immunohistochemistry may be used to evaluate somatic p53 mutations in endometrial samples, regardless of their histological characteristics. This finding may contribute to the identification of aggressive tumors, thereby facilitating the customization of surgical procedures, the development of appropriate adjuvant treatments and the establishment of effective follow-up protocols. Immunohistochemistry including the expression of all four proteins, together with methylation of the MLH1 promoter, continues to be the preferred diagnostic method due to its reproducibility, cost-effectiveness, and suitability for regular clinical use. The use of molecular categorization, such as MMRd tumors, has played a crucial role in the determination of prognosis. The authors declare no conflict of interests. This research received no external funding. |
Update Thoraxpathologie 2021 – Bericht der Arbeitsgemeinschaft | 99aaabe5-c791-4943-8cb3-afeee3d0fb52 | 8490854 | Pathology[mh] | Das Frühjahrstreffen fand am 5. Februar von 15 bis 19 Uhr in virtueller Form statt und wurde von Sabina Berezowska (Lausanne, Schweiz) und Verena Tischler (Bonn) organisiert. Die Veranstaltung war auch für Nicht-DGP-Mitglieder offen und sehr gut besucht, mit insgesamt 96 angemeldeten Teilnehmern aus Deutschland, Österreich, der Türkei, der Schweiz und den USA. Die erste Session beschäftigte sich mit einem Update zu COVID-19. Zunächst berichtete Alexandar Tzankov (Basel, Schweiz) darüber, was In-situ-Untersuchungen am Autopsiegewebe zur Untersuchung des Pathomechanismus von letaler COVID-19 beitragen können und bereits bisher beigetragen haben . Im Anschluss berichtete Christopher Werlein (AG Lungenforschung, Institut für Pathologie MHH) ebenfalls zu COVID-19 aus der Perspektive eines Pathologen und fasste die die hochkarätigen Publikationen und aktuelle Untersuchungen aus Hannover zusammen . Abschließend gab Viktor Kölzer (Zürich, Schweiz) eine Zusammenfassung der Studie, in der das Team zwischen Liestal (Kirsten Mertz), Zürich (Viktor Koelzer) und Trento (Francesca Demichelis) zwei distinkte immunpathologische Profile in Autopsielungen von COVID-19-Verstorbenen identifiziert haben . Durch die Sitzung führte Sabina Berezowska. Wie zu erwarten gab es viel Diskussionsbedarf, der die anschließende Pause verkürzte. Die zweite Session zum Thema „Update Molekularpathologie und translationale Lungenpathologie“ wurde durch Verena Tischler geleitet. Sabine Merkelbach-Bruse (Köln) berichtete über die Fortschritte in der molekularen Diagnostik des Verbundprojekts Nationales Netzwerk genomische Medizin (nNGM) und hier insbesondere über innovative genomische Panels, welche insbesondere den Patient:innen neben zielgerichteten Standardtherapien früh und strukturiert Zugang zu klinischen Studien ermöglichen. Außerdem legte Sabine Merkelbach-Bruse die Harmonisierung der molekularen Diagnostik an den nNGM-Netzwerkzentren und die besonderen qualitätssichernden Maßnahmen anschaulich dar . Weiter ging es mit Sonja Loges (Hamburg und Mannheim), die die Studie ihrer Arbeitsgruppe zu den typischen und atypischen EGFR -Mutationen vorstellte. In dieser Studie hat Sonja Loges mit ihrem Team eine beeindruckende präklinische Pipeline zur funktionellen Charakterisierung von klinisch unklaren EGFR- Varianten der innerhalb des Netzwerks Genomische Medizin und des nNGMs aufgefunden Mutationsspektren zur Therapieentscheidung dargelegt. Im Anschluss sprach Jürgen Wolf (Köln) über die klinische Seite der MET-Inhibitoren beim nichtkleinzelligen Lungenkarzinom (NSCLC) und ging ausführlich auf die molekularen MET -Alterationen und die Bedeutung ihrer präzisen molekularpathologischen Bestimmung für die klinische Therapie ein. Martin Sebastian (Frankfurt) sprach über die bahnbrechenden Therapieerfolge bei NSCLC-Patient:innen mit KRAS -Mutationen und betonte die Wichtigkeit der präzisen Variantenbestimmung und des exakten Reportings im molekularpathologischen Befund für die beste Therapiewahl. Zum Abschluss des wissenschaftlichen Programms gab Alexander Marx (Mannheim) ein Update zu Mediastinaltumoren – bereits als Vorgriff auf die später im Jahr erschienene WHO-Klassifikation und unter Einbeziehung ganz aktueller, noch nicht publizierter Daten. Sabina Berezowska schloss das Programm ab mit einem 10-minütigen Update zur vom August 2020 verschobenen und dann letztendlich eine Woche vor dem Frühjahrstreffen nachgeholten „World Conference on Lung Cancer 2020“ der International Association for the Study of Lung Cancer (IASLC), die, anstatt in Singapur, ebenfalls virtuell durchgeführt hat werden müssen. Die Mitgliederversammlung wurde aufgrund des virtuellen Formates stark verkürzt. In der angesetzten Viertelstunde wurde Verena Tischler zur stellvertretenden Sprecherin der AG gewählt. Zudem wurde die Anzahl der Beiratsmitglieder diskutiert und eine aktuelle Anzahl von 9 Mitgliedern inklusive eines kooptierten Mitglieds als konform mit den aktuellen Statuten der AG verifiziert. Der Vorschlag, dass zukünftig ein Ergebnisprotokoll der Mitgliedsversammlungen durch Sprecher:in oder Stellvertreter:in geführt wird, wurde von den Mitgliedern angenommen. Den Abschluss der Mitgliederversammlung bildete eine Diskussion zu den Vorschlägen der eingeladenen Referenten für die „Virtuellen Pathologietage 2021“ und eine Diskussion zum Thema nNGM. AG-Sitzung Im Pandemiejahr 2021 wurde die reguläre Jahrestagung der DGP ausgesetzt und durch die „Virtuellen Pathologietage 2021“ ersetzt. Die Sitzung der AG Thoraxpathologie fand am ersten Kongresstag, dem 8. Juni, von 16 bis 18 Uhr als Livesitzung statt. Die Resonanz war hervorragend, mit einem Maximum von zeitweise bis zu 99 eingeloggten Teilnehmern. Während der Onlinepräsentationen konnten über die Chatfunktion fortlaufend Fragen gestellt werden, die während der anschließenden Diskussion mündlich oder aus Zeitgründen anschließend daran ebenfalls im Chat beantwortet wurden. Dies ermöglichte trotz des virtuellen Formates eine angeregte Diskussion. Die Sitzung wurde durch die beiden hochkarätigen Gastvorträge eröffnet. Zunächst berichtet Katrien Grunberg aus Nijmegen (Niederlande) zur Gefäßpathologie der Lunge, die aktuell vor allem durch die Corona-Pandemie ins Scheinwerferlicht gerückt ist. Ihr Vortrag „A practical approach to vasculopathy in pulmonary hypertension“ schloss auch Kommentare zur vaskulären Pathologie bei COVID-19 mit ein. Anschließend referierte Philipp Ströbel aus Göttingen die aktuellen Erkenntnisse zu neuroendokrinen Neoplasien des Thymus und der Lunge mit der spezifischen Frage, ob es sich um den gleichen Tumor in 2 Organen handelt . Die Unterschiede der Neoplasien abhängig von der Lokalisation waren frappierend. Der zweite Teil der Sitzung war für eingereichte Beiträge reserviert. Florian Länger (Hannover) gab einen sehr umfassenden Überblick zu interstitiellen Lungenerkrankungen im Säuglings- und Kindesalter, der sicher einen längeren Slot verdient hätte . Ralf Marienfeld (Ulm) trug die Ergebnisse zur räumlichen Verteilung der Immuncheckpoint-Proteine in unterschiedlichen morphologischen Subtypen von Adenokarzinomen der Lunge vor . Philip Bischoff (Berlin) berichtete zur Charakterisierung des Tumormikromilieus in Adenokarzinomen der Lunge mittels Einzelzell-RNA-Sequenzierung. Sylvia Lohfink-Schumm (Bonn) sprach zur Mutationslandschaft von nichtkleinzelligen Lungenkarzinomen mit MET -Genalterationen. Zuletzt fasste Sabina Berezowska in einem 5‑minütigen Vortrag die als Poster präsentierten Einreichungen aus der AG Thoraxpathologie zusammen. Die Posterpräsentationen waren als „besprochene Powerpoint-Datei“ abrufbar und technisch und inhaltlich von hervorragender Qualität. Aufgrund der fortgeschrittenen Zeit und technischer Schwierigkeiten mit einem zwar pünktlichen, aber unangenehm abrupten Abschluss der Sitzung für die zugeschalteten Kongressteilnehmer konnte die Mitgliederversammlung nicht adäquat durchgeführt werden. Wir freuen uns umso mehr auf das nächste Treffen im Rahmen der Frühjahrstagung 2020. COVID-19-Sitzung Bereits bei der Organisation der Jahrestagung 2020 kam aus der AG Thoraxpathologie der anschließend umgesetzte Vorschlag, eine zusätzliche Sitzung zu dem Thema SARS-CoV‑2 mit dem Schwerpunkt Diagnostik und histopathologische Veränderungen zu organisieren. Dies wurde auch dieses Jahr beibehalten. Das Hauptprogramm der „Virtuellen Pathologietage 2021“ am Freitag den 11. Juni 2021 wurde somit mit dem Thema COVID-19 eröffnet. Nach der Keynote Lecture „Genomic Epidemiology of SARS-CoV-2: From Variants of Concern to Understanding Population Transmission Chains“ durch Alexander Dilthey (Düsseldorf) folgte ein Update zu COVID-19 vonseiten der Pathologie. Durch die Sitzung führten Danny Jonigk (Hannover) und Peter Boor (Aachen). Eröffnet wurde die Sitzung von Peter Boor, der das – immer noch international einzigartige – deutsche Register für COVID-19-Obduktionen und das deutsche Netzwerk für Autopsien bei Pandemien (Defeat Pandemics) vorstellte, erste Ergebnisse der Datenauswertung präsentierte und wiederum für das wichtige kooperative Miteinander aller Beteiligten warb. Darauf folgte Tobias Welte (Hannover), der eine beeindruckende Synopse der Entwicklung der COVID-19-Pandemie aus epidemiologischer, virologischer und klinischer Sicht und – immer gestützt auf aktuelle Studien – einen Ausblick für die wahrscheinliche weitere Entwicklung von COVID-19 in Deutschland und der Welt gab. Danach führte Leif Erik Sander (Berlin) äußerst kompetent und umfassend durch die Aufschlüsselungen der beeinträchtigten Immunantwort bei schweren (systemischen) COVID-19-Verläufen. In der Folge stellte Danny Jonigk erste umfassende Daten zu kompartimentenspezifischen Alterationen bei COVID-19-Beteiligungen von Herz und Lunge vor und arbeitete insbesondere die angiozentrische Pathogenese der Erkrankung heraus. Im Anschluss referierte Michael H. Roehrl (New York, USA), wie eine SARS-CoV-2-Infektion die Abwehrreaktion des Wirtes signifikant verändern und autoimmunologische Mechanismen aktivieren kann. Darüber hinaus stellte er erste Verbindungen der vorbeschriebenen Mechanismen zum „Long-COVID-Syndrom“ her. Abgerundet wurde die Sitzung von Konrad Steinestel (Bundeswehrkrankenhaus Ulm), der sich umfassend mit den klinischen, bildgebenden, serologischen, histopathologischen Veränderungen des (pulmonalen) „Long-COVID-Syndroms“ befasste und wie auch sein Vorredner die wahrscheinliche Rolle fehlaktivierter (auto-)immunologischer Mechanismen kompetent herausarbeitete. Während der „Virtuellen Pathologietage“ wurde zum Thema Thoraxpathologie auch in anderen Sitzungen referiert. Die Beiträge werden hier gelistet. Vortrag im Rahmen der Sitzung der AG Kopf-Hals-Pathologie High-resolution mass spectrometry identifies new immunohistochemical markers to differentiate primary squamous cell carcinomas of the lung and metastasis of squamous cell carcinomas of the head and neck (AG05.02, A. Richter et al.) Vortrag im Rahmen der Sitzung der AG Molekularpathologie Clonal hematopoiesis in lung cancer patients—detection of false-positive findings in liquid biopsies by parallel sequencing. (AG14.01, J. Fassunke et al.) Vortrag im Rahmen der Bewerbung um den DGP-Promotionspreis Solution to the challenges of automated immune cell detection in the lung—anthracosis (DGP01.10, P. Zens) Vorträge im Rahmen der Bewerbung um den DGP-Forschungspreis SARS-CoV‑2 infects and replicates in cells of the human endocrine and exocrine pancreas (DGP07.01, T. FE Barth) Deciphering the proteome of COVID-19—a multi-organ proteomic profiling of COVID-19 autopsies (DGP07.02, L. Schweizer) Um auch alle als Poster präsentierten Beiträge zum Thema Thoraxpathologie und COVID-19 zu würdigen, sind sie im Nachfolgenden aufgeführt. Nähere Informationen können dem Abstractband entnommen oder direkt von den Autoren angefordert werden. Potential role of Tenascin C (TNC) in human non-small cell lung cancer progression (P02.01, M. Schlensog et al.) Cancer-associated fibroblasts regulate kinase activity in MPM cell lines and thereby dictate cell fate and behavior (P02.02, A. Mathilakathu et al.) Influence of DNA Damage Repair, TP53-Network and Metallothionein on the Susceptibility to Cisplatin in Malignant Pleural Mesothelioma (P02.03, S. Borchert et al.) Comprehensive molecular profiling of malignant pleural mesothelioma: a proof-of-concept study (P02.04, H. Goldschmid et al.) Cancer associated fibroblasts and malignant pleural mesothelioma: revealing the role of the tumor microenvironment on tumor progression (P02.05, A. Nath et al.) Digital gene expression analysis reveals differences in immunogenicity based on subtypes of malignant pleural mesothelioma (P02.06, M. Wessolly et al.) Mitogen signal associated pathways energy metabolism regulation as well as mediation of tumor immunogenicity play important roles in cellular response of pleural mesotheliomas to cisplatin-based treatment (P02.07, A. Mathilakathu et al.) Cancer-Associated Fibroblasts and their Influence on Cancer Progression and Motilityin Malignant Pleural Mesothelioma (P02.08, A. Mathilakathu et al.) Autopsy findings after long-term treatment of COVID-19 patients with microbiological correlation (PCov.01, K. Evert et al.) All-Body-Cavity (ABC)-Scopy as a New Approach for Minimal Invasive Post Mortem Examination (PCov.02, L. Rentschler et al.) COVID-19 in a patient with active miliary tuberculosis: clinical outcome and histopathology of lung parenchyma (PCov.03, S. Opitz et al.) How COVID-19 pathology data contribute to research and patient treatment (PCov.04, L. Domke et al.) A sensitive protocol for SARS-CoV‑2 RNA screening in diagnostic archives (PCov.05, S. Villwock et al.) COVID-19: Preliminary results from a Swab-Test-Study to detect SARS-CoV2 on personal protective equipment of autopsy staff at four German centers (PCov.06, S. Dintner et al.) Results from the Augsburg-COVID-19-autopsy series (PCov.07, T. Schaller et al.) Präsentiert im Rahmen der Poster der Kinder- und Fetalpathologie: ABCA3 deficiency – Aggravating factor for neonatal respiratory distress syndrome (RDS) in sepsis? (P03.01, C. Jayasinghe et al.) Präsentiert im Rahmen der Poster der Uropathologie: Gene expression analysis of prostate cancer lung metastases (P11.10, K. Hempel et al.) Präsentiert im Rahmen der Poster der Molekularpathologie: Immune cell biomarkers in lung adenocarcinoma: inter- and intra-tumor heterogeneity (P14.02, J. Budczies et al.) Detecting EGFR T790M resistance mutation in plasma at progression on TKI therapy using ultrasensitive digital PCR (P14.15, S. Isaksson et al.) Comparison of two different targeted NGS panels for molecular diagnosis of non-small cell lung cancer: ArcherDx versus ThermoFisher Scientific. (P14.16, M. Wetz et al.) Somit kann die AG Thoraxpathologie auch im zweiten Corona-Jahr trotz aller Einschränkungen auf ein ereignisreiches und erfolgreiches Jahr zurückblicken. Die nächste Sitzung der AG soll während des Frühjahrstreffens 2022 in Bonn als Hybridveranstaltung stattfinden und steht unter der Leitung von PD Dr. Verena Tischler. Wir freuen uns schon sehr darauf! Im Pandemiejahr 2021 wurde die reguläre Jahrestagung der DGP ausgesetzt und durch die „Virtuellen Pathologietage 2021“ ersetzt. Die Sitzung der AG Thoraxpathologie fand am ersten Kongresstag, dem 8. Juni, von 16 bis 18 Uhr als Livesitzung statt. Die Resonanz war hervorragend, mit einem Maximum von zeitweise bis zu 99 eingeloggten Teilnehmern. Während der Onlinepräsentationen konnten über die Chatfunktion fortlaufend Fragen gestellt werden, die während der anschließenden Diskussion mündlich oder aus Zeitgründen anschließend daran ebenfalls im Chat beantwortet wurden. Dies ermöglichte trotz des virtuellen Formates eine angeregte Diskussion. Die Sitzung wurde durch die beiden hochkarätigen Gastvorträge eröffnet. Zunächst berichtet Katrien Grunberg aus Nijmegen (Niederlande) zur Gefäßpathologie der Lunge, die aktuell vor allem durch die Corona-Pandemie ins Scheinwerferlicht gerückt ist. Ihr Vortrag „A practical approach to vasculopathy in pulmonary hypertension“ schloss auch Kommentare zur vaskulären Pathologie bei COVID-19 mit ein. Anschließend referierte Philipp Ströbel aus Göttingen die aktuellen Erkenntnisse zu neuroendokrinen Neoplasien des Thymus und der Lunge mit der spezifischen Frage, ob es sich um den gleichen Tumor in 2 Organen handelt . Die Unterschiede der Neoplasien abhängig von der Lokalisation waren frappierend. Der zweite Teil der Sitzung war für eingereichte Beiträge reserviert. Florian Länger (Hannover) gab einen sehr umfassenden Überblick zu interstitiellen Lungenerkrankungen im Säuglings- und Kindesalter, der sicher einen längeren Slot verdient hätte . Ralf Marienfeld (Ulm) trug die Ergebnisse zur räumlichen Verteilung der Immuncheckpoint-Proteine in unterschiedlichen morphologischen Subtypen von Adenokarzinomen der Lunge vor . Philip Bischoff (Berlin) berichtete zur Charakterisierung des Tumormikromilieus in Adenokarzinomen der Lunge mittels Einzelzell-RNA-Sequenzierung. Sylvia Lohfink-Schumm (Bonn) sprach zur Mutationslandschaft von nichtkleinzelligen Lungenkarzinomen mit MET -Genalterationen. Zuletzt fasste Sabina Berezowska in einem 5‑minütigen Vortrag die als Poster präsentierten Einreichungen aus der AG Thoraxpathologie zusammen. Die Posterpräsentationen waren als „besprochene Powerpoint-Datei“ abrufbar und technisch und inhaltlich von hervorragender Qualität. Aufgrund der fortgeschrittenen Zeit und technischer Schwierigkeiten mit einem zwar pünktlichen, aber unangenehm abrupten Abschluss der Sitzung für die zugeschalteten Kongressteilnehmer konnte die Mitgliederversammlung nicht adäquat durchgeführt werden. Wir freuen uns umso mehr auf das nächste Treffen im Rahmen der Frühjahrstagung 2020. Bereits bei der Organisation der Jahrestagung 2020 kam aus der AG Thoraxpathologie der anschließend umgesetzte Vorschlag, eine zusätzliche Sitzung zu dem Thema SARS-CoV‑2 mit dem Schwerpunkt Diagnostik und histopathologische Veränderungen zu organisieren. Dies wurde auch dieses Jahr beibehalten. Das Hauptprogramm der „Virtuellen Pathologietage 2021“ am Freitag den 11. Juni 2021 wurde somit mit dem Thema COVID-19 eröffnet. Nach der Keynote Lecture „Genomic Epidemiology of SARS-CoV-2: From Variants of Concern to Understanding Population Transmission Chains“ durch Alexander Dilthey (Düsseldorf) folgte ein Update zu COVID-19 vonseiten der Pathologie. Durch die Sitzung führten Danny Jonigk (Hannover) und Peter Boor (Aachen). Eröffnet wurde die Sitzung von Peter Boor, der das – immer noch international einzigartige – deutsche Register für COVID-19-Obduktionen und das deutsche Netzwerk für Autopsien bei Pandemien (Defeat Pandemics) vorstellte, erste Ergebnisse der Datenauswertung präsentierte und wiederum für das wichtige kooperative Miteinander aller Beteiligten warb. Darauf folgte Tobias Welte (Hannover), der eine beeindruckende Synopse der Entwicklung der COVID-19-Pandemie aus epidemiologischer, virologischer und klinischer Sicht und – immer gestützt auf aktuelle Studien – einen Ausblick für die wahrscheinliche weitere Entwicklung von COVID-19 in Deutschland und der Welt gab. Danach führte Leif Erik Sander (Berlin) äußerst kompetent und umfassend durch die Aufschlüsselungen der beeinträchtigten Immunantwort bei schweren (systemischen) COVID-19-Verläufen. In der Folge stellte Danny Jonigk erste umfassende Daten zu kompartimentenspezifischen Alterationen bei COVID-19-Beteiligungen von Herz und Lunge vor und arbeitete insbesondere die angiozentrische Pathogenese der Erkrankung heraus. Im Anschluss referierte Michael H. Roehrl (New York, USA), wie eine SARS-CoV-2-Infektion die Abwehrreaktion des Wirtes signifikant verändern und autoimmunologische Mechanismen aktivieren kann. Darüber hinaus stellte er erste Verbindungen der vorbeschriebenen Mechanismen zum „Long-COVID-Syndrom“ her. Abgerundet wurde die Sitzung von Konrad Steinestel (Bundeswehrkrankenhaus Ulm), der sich umfassend mit den klinischen, bildgebenden, serologischen, histopathologischen Veränderungen des (pulmonalen) „Long-COVID-Syndroms“ befasste und wie auch sein Vorredner die wahrscheinliche Rolle fehlaktivierter (auto-)immunologischer Mechanismen kompetent herausarbeitete. Während der „Virtuellen Pathologietage“ wurde zum Thema Thoraxpathologie auch in anderen Sitzungen referiert. Die Beiträge werden hier gelistet. Vortrag im Rahmen der Sitzung der AG Kopf-Hals-Pathologie High-resolution mass spectrometry identifies new immunohistochemical markers to differentiate primary squamous cell carcinomas of the lung and metastasis of squamous cell carcinomas of the head and neck (AG05.02, A. Richter et al.) Vortrag im Rahmen der Sitzung der AG Molekularpathologie Clonal hematopoiesis in lung cancer patients—detection of false-positive findings in liquid biopsies by parallel sequencing. (AG14.01, J. Fassunke et al.) Vortrag im Rahmen der Bewerbung um den DGP-Promotionspreis Solution to the challenges of automated immune cell detection in the lung—anthracosis (DGP01.10, P. Zens) Vorträge im Rahmen der Bewerbung um den DGP-Forschungspreis SARS-CoV‑2 infects and replicates in cells of the human endocrine and exocrine pancreas (DGP07.01, T. FE Barth) Deciphering the proteome of COVID-19—a multi-organ proteomic profiling of COVID-19 autopsies (DGP07.02, L. Schweizer) Um auch alle als Poster präsentierten Beiträge zum Thema Thoraxpathologie und COVID-19 zu würdigen, sind sie im Nachfolgenden aufgeführt. Nähere Informationen können dem Abstractband entnommen oder direkt von den Autoren angefordert werden. Potential role of Tenascin C (TNC) in human non-small cell lung cancer progression (P02.01, M. Schlensog et al.) Cancer-associated fibroblasts regulate kinase activity in MPM cell lines and thereby dictate cell fate and behavior (P02.02, A. Mathilakathu et al.) Influence of DNA Damage Repair, TP53-Network and Metallothionein on the Susceptibility to Cisplatin in Malignant Pleural Mesothelioma (P02.03, S. Borchert et al.) Comprehensive molecular profiling of malignant pleural mesothelioma: a proof-of-concept study (P02.04, H. Goldschmid et al.) Cancer associated fibroblasts and malignant pleural mesothelioma: revealing the role of the tumor microenvironment on tumor progression (P02.05, A. Nath et al.) Digital gene expression analysis reveals differences in immunogenicity based on subtypes of malignant pleural mesothelioma (P02.06, M. Wessolly et al.) Mitogen signal associated pathways energy metabolism regulation as well as mediation of tumor immunogenicity play important roles in cellular response of pleural mesotheliomas to cisplatin-based treatment (P02.07, A. Mathilakathu et al.) Cancer-Associated Fibroblasts and their Influence on Cancer Progression and Motilityin Malignant Pleural Mesothelioma (P02.08, A. Mathilakathu et al.) Autopsy findings after long-term treatment of COVID-19 patients with microbiological correlation (PCov.01, K. Evert et al.) All-Body-Cavity (ABC)-Scopy as a New Approach for Minimal Invasive Post Mortem Examination (PCov.02, L. Rentschler et al.) COVID-19 in a patient with active miliary tuberculosis: clinical outcome and histopathology of lung parenchyma (PCov.03, S. Opitz et al.) How COVID-19 pathology data contribute to research and patient treatment (PCov.04, L. Domke et al.) A sensitive protocol for SARS-CoV‑2 RNA screening in diagnostic archives (PCov.05, S. Villwock et al.) COVID-19: Preliminary results from a Swab-Test-Study to detect SARS-CoV2 on personal protective equipment of autopsy staff at four German centers (PCov.06, S. Dintner et al.) Results from the Augsburg-COVID-19-autopsy series (PCov.07, T. Schaller et al.) Präsentiert im Rahmen der Poster der Kinder- und Fetalpathologie: ABCA3 deficiency – Aggravating factor for neonatal respiratory distress syndrome (RDS) in sepsis? (P03.01, C. Jayasinghe et al.) Präsentiert im Rahmen der Poster der Uropathologie: Gene expression analysis of prostate cancer lung metastases (P11.10, K. Hempel et al.) Präsentiert im Rahmen der Poster der Molekularpathologie: Immune cell biomarkers in lung adenocarcinoma: inter- and intra-tumor heterogeneity (P14.02, J. Budczies et al.) Detecting EGFR T790M resistance mutation in plasma at progression on TKI therapy using ultrasensitive digital PCR (P14.15, S. Isaksson et al.) Comparison of two different targeted NGS panels for molecular diagnosis of non-small cell lung cancer: ArcherDx versus ThermoFisher Scientific. (P14.16, M. Wetz et al.) Somit kann die AG Thoraxpathologie auch im zweiten Corona-Jahr trotz aller Einschränkungen auf ein ereignisreiches und erfolgreiches Jahr zurückblicken. Die nächste Sitzung der AG soll während des Frühjahrstreffens 2022 in Bonn als Hybridveranstaltung stattfinden und steht unter der Leitung von PD Dr. Verena Tischler. Wir freuen uns schon sehr darauf! |
Development and implementation of medication-related clinical rules for obstetrics, gynaecology, and paediatric outpatients | 99d087e9-ebee-4f17-99c2-13ca72901472 | 10895191 | Gynaecology[mh] | Medication errors may lead to patient injury, disability, or even death, and increase medical care costs and wastage of medical care resources. Women, especially pregnant women, and children are highly susceptible to adverse drug events caused by prescription errors. Prescription errors are associated with medication errors, which can lead to serious adverse drug events and even death. The use of information technologies, such as computerised provider order entry (CPOE) and clinical decision support systems (CDSSs), is an effective means of reducing prescription errors. CPOE ensures standardised prescriptions and prevents transcription errors, but without embedding in a CDSS, other patient characteristics, such as indications, biochemical tests, and demographic data, cannot be obtained. CPOE embedded in a CDSS can match patient-specific data with a knowledge base to achieve basic functions, such as limited drug–drug interaction (DDI) checking, basic dosage guidance, and drug–allergy interaction checking. The CDSSs of some medical institutions have more inbuilt complex clinical rules, which can achieve more advanced functions, such as DDI checking, dosing support for renal insufficiency and geriatric patients, guidance for medication-related laboratory testing, drug–pregnancy checking, and drug–disease contraindication checking. There have been few studies on large data volume and real-time clinical rules concerning indications, usage and dosage, combination medication, and typographical errors in prescriptions for high-risk groups, such as obstetrics, gynaecology, and paediatric patients. In most areas of China, the number of outpatient prescriptions is high. Thus, there is a great risk of prescription errors and serious adverse outcomes. Health management policies require pharmacists to review prescriptions before prescription pricing, and medication dispensing. This requires perfect digital communication, effective data integration, frequent updates, and a high degree of localised customisation. Before 2019, prescription information was subjected to manual review by a pharmacist, without considering other patient characteristics. After 2019, we developed a real-time prescription review system used by pharmacists, with more than 15 000 medication-related clinical rules and visual rule adjustment functions. Clinical rules present prescription error alerts directly to the prescriber during order entry and provide advice for modification. Considering the lack of information regarding the application of real-time clinical rules in obstetrics, gynaecology, and paediatric outpatients, in this study, we aimed to evaluate the effects of localised, real-time clinical rules on alert rates and acceptance rates compared with manual prescription review. Design This retrospective cohort (before and after) study was approved by the Ethics Council of Human Research in Xiamen Maternal and Child Healthcare Hospital (No. KY-2020–085). Setting and population This study was conducted in a grade III tertiary hospital for maternal and child health located in Xiamen, China. The hospital has 1100 employees, 700 beds, and 1.4 million outpatient visits per year. We used data from the hospital information system (HIS) to compare the alert and recommendation acceptance rates before and after clinical rule implementation, including all outpatient prescriptions for 2 years, from 1 January 2018 to 31 December 2019. Clinical rules were introduced in July 2017, and doctors and pharmacists were trained to use the rules which included 15 000 rules when launched. Classification and definition of prescription errors We formed prescription review committees, including doctors, pharmacists, nurses, and administrators. Medication prescribing errors were defined as deviations from drug labels and did not include low risk off-label use approved by the prescription review committee in accordance with clinical practice guidelines. According to the Standards for Prescription Reviewing in Medical Institutions issued by the National Health Commission of the People’s Republic of China, prescription errors included indication errors (contraindications, wrong diagnosis, and mismatch between diagnosis and medicine), dosage errors (improper prescription involves improper dosage (tolerability over 20%), wrong frequency, wrong administration route, and wrong dosage form), entry errors, DDI errors, and combination medication errors (unreasonable simultaneous or sequential use of two or more drugs for therapeutic purposes, such as treatment of vaginitis with moxifloxacin and metronidazole to addresss anaerobic bacteria). Alerts were divided into five categories based on the above mentioned prescription errors. Manual review of prescription The prescribing system requires doctors to enter all content present in the prescription, as well as at least one indication and one drug. The prescribing system only provides basic prescribing services, without mandatory default dosage, frequency, route, and automatic error checking. In a pharmacist team including 20 members, approximately 10 people were on duty every day to review prescriptions. All pharmacists were examined and qualified after a uniform training for error review. Before medication dispensing, prescriptions were reviewed by a pharmacist after drug orders were sent by physicians. Questionable prescriptions were verified and recorded by another pharmacist, who called the prescribing doctor to provide advice or communicate to reach an agreement . To prevent review fatigue, the pharmacists took turns to review the prescriptions every hour. The details of prescription reviewing comprised patient conditions (age, sex, and diagnosis) and therapy regimens (medication selection, dose, frequency, route of administration, and DDI). System review of prescription The prescription review system was developed based on the HIS and data integration platforms, including CPOE, laboratory information system, electronic medical records, and other systems . In addition to medication-related information, the system integrated patient characteristics and laboratory values that were included in the algorithms to generate alerts. After a physician sent the prescription order through CPOE, the system determined whether the patient’s medication matched all available clinical and demographic information by examining the available structured data in each system database through clinical rules. In the clinical rule service, the pharmacist does not have to check all alerts. The clinical rule directly alerts the doctor on the prescription interface and provides relevant advice if the prescription does not match the clinical rules, implying prescription errors. A doctor who does not accept the recommendation is required to state the reason and communicate with the pharmacist to reach an agreement . For example, for an 8-year-old child of 25 kg weight with Helicobacter pylori infection, if a doctor prescribes 400 mg amoxicillin, the rules alert the doctor on the low dose and recommend 625 mg amoxicillin, which is the dosage standard of 50 mg/kg/day for H. pylori infection treatment. Clinical rules are established and revised by prescription review committees, according to labels, professional books, relevant evidence-based guidelines, the latest literature, health management policy requirements, and nearly 30 000 pharmacist review data entries in our hospital over the last 10 years. For each clinical rule, a ‘Yes’ or ‘No’ response was displayed. A user-friendly standardised flowchart or decision tree for pharmacists was drawn, and the rules were adjusted as needed . Data analysis The main outcome indicators included the alert rate and the recommendation acceptance rate of all prescription errors in paediatrics and obstetrics and gynaecology. Secondary outcome indicators included the alert rate of different types of prescription errors in paediatrics and obstetrics and gynaecology. Relative risk (RR) was used to analyse the differences in alert and acceptance rates between the groups. The recommendation acceptance rate was defined as the ratio of the number of prescriptions modified by the prescriber to the number of all recommendations. For a prescription in which the doctor does not accept the recommendation of the rule and uses a manual review, the prescription is included in the number of recommendations during calculation of acceptance rates depending on whether the pharmacist agrees with the reason offered by the doctor or not. To avoid confusion and bias, our study excluded rotation prescriptions. Results with a two-sided p value <0.05 were considered statistically significant. All statistical analyses were performed using R (version 3.6.1). This retrospective cohort (before and after) study was approved by the Ethics Council of Human Research in Xiamen Maternal and Child Healthcare Hospital (No. KY-2020–085). This study was conducted in a grade III tertiary hospital for maternal and child health located in Xiamen, China. The hospital has 1100 employees, 700 beds, and 1.4 million outpatient visits per year. We used data from the hospital information system (HIS) to compare the alert and recommendation acceptance rates before and after clinical rule implementation, including all outpatient prescriptions for 2 years, from 1 January 2018 to 31 December 2019. Clinical rules were introduced in July 2017, and doctors and pharmacists were trained to use the rules which included 15 000 rules when launched. We formed prescription review committees, including doctors, pharmacists, nurses, and administrators. Medication prescribing errors were defined as deviations from drug labels and did not include low risk off-label use approved by the prescription review committee in accordance with clinical practice guidelines. According to the Standards for Prescription Reviewing in Medical Institutions issued by the National Health Commission of the People’s Republic of China, prescription errors included indication errors (contraindications, wrong diagnosis, and mismatch between diagnosis and medicine), dosage errors (improper prescription involves improper dosage (tolerability over 20%), wrong frequency, wrong administration route, and wrong dosage form), entry errors, DDI errors, and combination medication errors (unreasonable simultaneous or sequential use of two or more drugs for therapeutic purposes, such as treatment of vaginitis with moxifloxacin and metronidazole to addresss anaerobic bacteria). Alerts were divided into five categories based on the above mentioned prescription errors. The prescribing system requires doctors to enter all content present in the prescription, as well as at least one indication and one drug. The prescribing system only provides basic prescribing services, without mandatory default dosage, frequency, route, and automatic error checking. In a pharmacist team including 20 members, approximately 10 people were on duty every day to review prescriptions. All pharmacists were examined and qualified after a uniform training for error review. Before medication dispensing, prescriptions were reviewed by a pharmacist after drug orders were sent by physicians. Questionable prescriptions were verified and recorded by another pharmacist, who called the prescribing doctor to provide advice or communicate to reach an agreement . To prevent review fatigue, the pharmacists took turns to review the prescriptions every hour. The details of prescription reviewing comprised patient conditions (age, sex, and diagnosis) and therapy regimens (medication selection, dose, frequency, route of administration, and DDI). The prescription review system was developed based on the HIS and data integration platforms, including CPOE, laboratory information system, electronic medical records, and other systems . In addition to medication-related information, the system integrated patient characteristics and laboratory values that were included in the algorithms to generate alerts. After a physician sent the prescription order through CPOE, the system determined whether the patient’s medication matched all available clinical and demographic information by examining the available structured data in each system database through clinical rules. In the clinical rule service, the pharmacist does not have to check all alerts. The clinical rule directly alerts the doctor on the prescription interface and provides relevant advice if the prescription does not match the clinical rules, implying prescription errors. A doctor who does not accept the recommendation is required to state the reason and communicate with the pharmacist to reach an agreement . For example, for an 8-year-old child of 25 kg weight with Helicobacter pylori infection, if a doctor prescribes 400 mg amoxicillin, the rules alert the doctor on the low dose and recommend 625 mg amoxicillin, which is the dosage standard of 50 mg/kg/day for H. pylori infection treatment. Clinical rules are established and revised by prescription review committees, according to labels, professional books, relevant evidence-based guidelines, the latest literature, health management policy requirements, and nearly 30 000 pharmacist review data entries in our hospital over the last 10 years. For each clinical rule, a ‘Yes’ or ‘No’ response was displayed. A user-friendly standardised flowchart or decision tree for pharmacists was drawn, and the rules were adjusted as needed . The main outcome indicators included the alert rate and the recommendation acceptance rate of all prescription errors in paediatrics and obstetrics and gynaecology. Secondary outcome indicators included the alert rate of different types of prescription errors in paediatrics and obstetrics and gynaecology. Relative risk (RR) was used to analyse the differences in alert and acceptance rates between the groups. The recommendation acceptance rate was defined as the ratio of the number of prescriptions modified by the prescriber to the number of all recommendations. For a prescription in which the doctor does not accept the recommendation of the rule and uses a manual review, the prescription is included in the number of recommendations during calculation of acceptance rates depending on whether the pharmacist agrees with the reason offered by the doctor or not. To avoid confusion and bias, our study excluded rotation prescriptions. Results with a two-sided p value <0.05 were considered statistically significant. All statistical analyses were performed using R (version 3.6.1). A total of 1 830 131 prescriptions over 2 years were included in the study. Of these, 735 798 prescriptions were reviewed manually, 3106 alerts were sent, and 2846 recommendations were accepted by doctors (91.6%). The clinical rules reviewed 1 094 333 prescriptions, sent 41 524 alerts, and doctors accepted 38 145 recommendations (91.9%). The on-duty pharmacists for the system review were adjusted from the original 10 to two persons . The alert rate of the system review in obstetrics and gynaecology was higher than that of the manual review (RR 3.97, 95% CI 3.75 to 4.20). The alert rates of DDI errors and combination medication errors in obstetrics and gynaecology significantly increased with the system review (RR 26.10, 95% CI 20.58 to 33.10, and RR 26.54, 95% CI 17.38 to 40.54, respectively) compared with those with the manual review. In paediatrics, all types of prescription error alert rates of the system were higher than those of the manual review (RR 11.26, 95% CI 10.73 to 11.82). The alert rates of DDI, dosage, and combination medication errors significantly increased with the system review (RR 35.49, 95% CI 28.33 to 44.45, RR 36.55, 95% CI 31.61 to 41.49, and RR 18.89, 95% CI 11.30 to 31.58, respectively) compared with those with the manual review. Although the number of alerts was significantly increased, there was no difference in the acceptance rate between the manual reviewer and the system reviewer in obstetrics, gynaecology, and paediatrics . In our study, the alert rates for clinical rules review were significantly higher than that for manual review in obstetrics and gynaecology and paediatric outpatients. This may be because clinical rules could obtain structured characteristic data and test values, and could identify many prescription errors that were not detected by the manual review. As expected, although different types of prescription error alert rates have different changes, most prescription error alert rates during the clinical rules review were significantly higher than those during the manual review. The most obvious changes were interactions and combination medications. The built-in knowledge base of drug information in clinical rules makes up for the limitation of pharmacists’ manual knowledge. Complicated prescription errors might be ignored by pharmacists during the manual review. For example, supplementing oestrogen and progesterone in menopausal women taking sleeping pills will weaken the effect of hormones. Our rules also solve the problem of a higher possibility of errors when patients have multiple prescriptions, as described by Usha et al . Furthermore, historical prescription information can be obtained based on interaction with other systems in the hospital. For example, if a woman visits department A to receive anticoagulant therapy and is prescribed warfarin, and then visits department B and is prescribed oestrogen and progesterone, the rule will alert doctors that oestrogen and progesterone will weaken the anticoagulant effect of warfarin. If a patient visits department A and is prescribed amoxicillin granules, and then visits department B, which prescribes amoxicillin and clavulanate potassium, the rule will raise a combination medication error alert. The results show that the alert rate for paediatric dosage and indication errors is higher with the clinical rules review than with the manual review, but it has a negligible effect on obstetrics and gynaecology. Oral and topical drugs in obstetrics and gynaecology are mostly administered at fixed doses; however, the dosage for children is mostly based on weight and age. Doctors sometimes adjust dosage according to disease severity. Although we set a range of ±20% for some usages based on evidence-based data, most drug dosages are based on labels. The clinical rules require a high degree of matching, without which false positives are prone to occur. For example, when children are prescribed ibuprofen granules, the dosage for a child weighing 21 kg is 210 mg, according to the label. The commercially available packaging specification is 0.2 g. Doctors will usually consider compliance when safety and effectiveness are controllable and prescribe a whole package; however, the rule will raise an alert in this case. Obstetrics and gynaecology outpatient diseases, such as vaginitis, pelvic inflammatory disease, and metrorrhagia, are diagnosed as common diseases. Many medications do not involve detection values, and the clinical rules have a negligible effect on the alert rate. When manually reviewing paediatric prescriptions, only prescription information can be obtained, and more characteristic information cannot be obtained. Thus, it is difficult to comprehensively judge a variety of physiological information in a short time. The present clinical rules can obtain the patient’s characteristic information and test values to review. For example, if methyldopa is prescribed to treat hypertension in pregnancy and the patient’s liver function is insufficient, the rule will raise an alert and suggest that labetalol be used instead. There was no difference in the acceptance rates between the two groups, which were over 90% in both groups. This may be due to the high acceptance rate in the manual review setting and a ceiling effect. Clinical rules do not change the display interface and the operating habits of doctors and display about 13 times as many alerts as that of the manual review. The average processing time for a prescription error was about 1 min for the manual review. The clinical rules automatically screen and obtain multiple data sources on the integrated platform, integrate the specific characteristics of patients and details of drug treatment, and provide personalised recommendations while alerting. It takes an average of 5 s for the clinical rules to deal with a prescription error. In our study, there may have been few doctors who performed the recommended changes without reviewing the alerts properly to save time, which is also the case with the manual review. The actual acceptance rate may be lower. It is necessary to consider false positive alerts; furthermore, alert fatigue caused by frequently occurring alerts that are not clinically relevant tends to be ignored. Eppenga et al demonstrated that when additional patient-related features are included, the clinical relevance of alerts is improved but is still not optimal. The localised system is highly customised to the condition of the implementing agency; thus, it is more likely to have a positive effect on safety and treatment quality. Clinical rules are established or revised by prescription review committees, and we used historical prescriptions for testing clinical rules to minimise false positives. Several studies have described the implementation and evaluation of CPOE and basic and advanced CPOE/CDSS, including look-alike/sound-alike and outlier detection. However, few studies have described services such as outpatient prescription review systems, especially where pharmacists review outpatient prescriptions in real-time and intervene in cases of incorrect prescriptions. Our clinical rules are similar to the Check of Medication Appropriateness (CMA) system developed by Charlotte et al . The CMA system mainly serves inpatients. It includes a list of comparable clinical rules classified by risk. The generated alerts are sent to the pharmacist instead of directly sending them to the prescribing doctor to prevent the doctor from feeling alarm fatigue. Our rules target outpatient prescriptions for special populations, such as women and children. Most alerts and advice are sent directly to doctors, while pharmacists are responsible for reviewing prescriptions when a doctor does not accept recommendations from clinical rules. This avoids a manual review of errors and review omissions and ensures real-time review efficiency in a large number of outpatient prescription environments. This study had some limitations. As a retrospective single-centre study, we did not assess the actual injury caused by prescription errors and the related cost-effectiveness. False positives alerts can cause frustration and alert fatigue to doctors. Besides, the number of clinically relevant alerts that might have been ignored by the prescribers due to alert fatigue remains unknown, although this scenario may rarely happen. We plan to promote this service in other healthcare institutions in the region and perform a multicentre study to assess the accessibility of the system. In the subsequent studies, patients’ actual injury and doctors’ satisfaction will be evaluated, including their general experience with the service, the overall reasons for agreeing or disagreeing with the medication recommendations, and their specific wishes or opinions for future expansion. Although doctors accept most medication-related recommendations, we plan to use more data, frequent updates, and advanced technologies to improve the specificity and sensitivity of this review system, such as identifying and integrating more patient characteristics or parameters, adjusting clinical rules with time, and applying natural language processing. Overall, our results show that the prescription review system can be used as an important supplement to the services of a manual reviewer, thereby improving review efficiency and saving human resources. Clinical rules can identify the prescription errors that cannot be detected by manual reviews in real-time based on patient characteristic data, detection values, and flexibly adjustable clinical rules. High acceptance rate and modification of prescriptions may be relevant to highly customised and localised clinical rules. However, some new challenges, such as mechanical reviewing and alert fatigue, may be introduced, and further research is required for optimisation. What this paper adds What is already known on this subject Computerised provider order entry embedded in a clinical decision support system can match patient-specific data with a knowledge base to achieve basic functions, such as limited drug–drug interaction checking, basic dosage guidance, and drug–allergy interaction checking. There is a lack of practical information on real-time clinical rules implemented in a large number of obstetrics, gynaecology, and paediatric outpatient prescriptions. What this study adds We developed a real-time prescription review system to prevent prescription errors that has more than 15 000 medication-related clinical rules and can be used as an important aid for a manual reviewer. Clinical rules can identify prescription errors that manual review cannot detect and ensure real-time review efficiency in high-volume outpatient prescription settings. The high acceptance rate and modification of prescriptions may be relevant to highly customised and localised clinical rules. How this study might affect research, practice and/or policy Our prescription review system could improve review efficiency and save human resources. Computerised provider order entry embedded in a clinical decision support system can match patient-specific data with a knowledge base to achieve basic functions, such as limited drug–drug interaction checking, basic dosage guidance, and drug–allergy interaction checking. There is a lack of practical information on real-time clinical rules implemented in a large number of obstetrics, gynaecology, and paediatric outpatient prescriptions. We developed a real-time prescription review system to prevent prescription errors that has more than 15 000 medication-related clinical rules and can be used as an important aid for a manual reviewer. Clinical rules can identify prescription errors that manual review cannot detect and ensure real-time review efficiency in high-volume outpatient prescription settings. The high acceptance rate and modification of prescriptions may be relevant to highly customised and localised clinical rules. Our prescription review system could improve review efficiency and save human resources. |
The American Society of Hematology Health Equity Compendium: examining health equity across the | 0618b77b-e283-4994-a13f-4d0f47947eb0 | 11401199 | Internal Medicine[mh] | |
Starve a cold or feed a fever? Identifying cellular metabolic changes following infection and exposure to SARS-CoV-2 | f197d97e-5aec-4f16-8a5f-bac378734171 | 11819565 | Biochemistry[mh] | As obligate intracellular parasites, viruses co-opt host cellular materials, machinery, and metabolism to facilitate viral replication . Metabolic changes in response to virus infection result in extensive alterations to cellular physiology and often mirror changes seen in cancer cells . Both RNA and DNA viruses reprogram different aspects of host metabolism including increased glycolysis, elevated pentose phosphate activity, amino acid generation and lipid synthesis . Viral hijacking of host metabolism and subversion of metabolic defenses can lead to increased viral replication and host damage, resulting in long-term health consequences, such as those seen in severe cases of COVID-19, following infection with the novel coronavirus, SARS-CoV-2. This is supported by analysis of SARS-CoV-2 positive patient serum that has shown acute and long-term changes in metabolites and further metabolic disorder . To better understand what metabolic changes occur during SARS-CoV-2 infection and how this may relate to severe disease outcomes, we implemented global metabolomic profiling to analyze thousands of metabolites using LC-MS to detect disease-associated changes to the cellular environment . Metabolites serve as intermediates for cellular physiology and include hormones, oligonucleotides, peptides, and other molecular products of cellular biochemical reactions that represent the current physiological state of a cell . Global metabolomic profiling can therefore provide an unbiased view of metabolic shifts induced during and in response to viral infection . To elucidate changes to cellular metabolism associated with SARS-CoV-2 viral replication and those changes associated with virus exposure we infected and profiled A549 cells, a human lung cell line. A549 cells are frequently used to evaluate viral infection for many respiratory viruses but are not intrinsically susceptible to SARS-CoV-2 infection, as they lack endogenous expression of the viral receptor, ACE2 . However, expression of human ACE2 protein on A549 (ACE2-A549) cells renders them fully susceptible to SARS-CoV-2 . By comparing A549 and ACE2-A549 cells inoculated with SARS-CoV-2 we can identify and separate metabolic shifts induced by active viral replication from those induced by the host cells response to virus exposure. Here, we describe distinct metabolic changes in to both ACE2-A549 and A549 cells triggered by SARS-CoV-2 exposure. Amino acid metabolism, glutathione, and urea cycle metabolic pathways were significantly altered in cells that support productive SARS-CoV-2 infection (ACE2-A549 cells). In contrast A549 cells that are not susceptible to infection but were exposed to a high inoculating dose of SARS-CoV-2 had significant changes in fatty acid anabolic and catabolic pathways as well as leukotriene metabolism. These results mirror the metabolic shifts found in serum from patients suffering from severe COVID-19 . Thus, our findings point to metabolite features associated with both active infection and exposure to virus. Understanding how cellular metabolism is reprogrammed following SARS-CoV-2 infection will allow identification of factors responsible for severe disease and aid in the development of antiviral therapies. Cells and viruses E6-Vero, A549, ACE2-A549 cells. E6 Vero cells were obtained from ATCC (Manassas, VA) and grown in DMEM supplemented with 10% FBS, 1% pen-strep. A549 cells were the obtained from Chang Lab. ACE2-A549 cells were obtained from BEI Resources (NR-53821). A549 cells were propagated in Hams F-12 (Corning) media supplemented with 10% fetal bovine serum (HyClone) and 1X Penicillin/Streptomycin (Fisher Scientific). ACE2-A549 cells were supplemented with 100ug/mL Blasticidin (Gibco). SARS-CoV-2 strain WA01 was obtained from BEI Resources (NR-52281). Viral stocks were propagated and titered on E6 Vero cells in DMEM supplemented with 2% FBS and 1% pen-strep. Viral stocks were made by collecting media from infected cell cultures showing extensive cytopathic effect and centrifuged 1,000 RCF for 5 minutes to remove cellular debris. The clarified viral supernatant was then used for all experimental infection. For determination of viral infectivity by plaque assay, E6 Vero cells were cultured then incubated with viral inoculum at limiting dilutions. Dilutions employed resulted in a minimum threshold of detection at 500 plaque forming units (PFU)/mL. Following inoculation, cells were over-layered with 1% methylcellulose, DMEM supplemented with 2% FBS and 1% pen-strep and incubated for 3-4 days . Cells were then fixed and stained with 0.5% methylene blue/70% ethanol solution. Plaques were counted and the overall titer was calculated. Western blot detection Both A549 or ACE2-A549 cells were cultured and collected with PBS supplemented with 0.5 mM EDTA. Cells were pelleted by centrifugation then resuspended in protein extraction buffer [10 mM Tris–HCl (pH 7.5), 10 mM NaCl, 1.5 mM MgCl2, 1% NP40] supplemented with Mini-complete, EDTA Free Protease Inhibitor Cocktail (Roche Applied Science, Indianapolis, ID). Protein concentrations determined by BCA assay, and then 20 µg were loaded and separated on a 10% polyacrylamide gel. Subsequently, proteins were transferred to PVDF membranes and incubated with a goat anti-human ACE2 mAb (R&D Systems Cat # AF933) and secondary rabbit-anti goat AF488-conjugated mAb (ThermoFisher, Cat # A27012). After detection, the blot was stripped and re-probed with mouse anti-beta Actin and a goat anti-mouse DyLight550 secondary (ThermoFisher, Cat # 84540). Membranes were imaged on a Cytiva Amersham Typhoon scanner (Bucks, UK) using the Cy2 and Cy3 channels. Immunofluorescence detection of infection Prior to infection, cells were seeded at 2 x 10 4 per well of an eight-chamber coverslip (Labtek Cat. No. 155411, Nunc International, Rochester, NY). At indicated times post infection, cells were then fixed with 4% paraformaldehyde in PBS for 30 minutes, washed thoroughly with PBS, and blocked in 2% bovine serum albumin (BSA) prior to antibody incubations. Primary and secondary antibodies were diluted in a PBS supplemented with 0.5% saponin, 0.125% BSA as described , and incubated for one hour at room temperature. Primary mouse anti-nucleocapsid (Thermofisher, Cat # MA1-7403) was diluted 1:500, followed by goat anti-mouse IgG labeled with Dylight 550 ThermoFisher, Cat # 84540) at 1:500. DNA was counterstained with Hoescht 3342 at 1:5000 dilution. Actin filaments stained with phalloidin-488 (Thermofisher, Cat # A12379) at 1:500. Stained cells were imaged on a Nikon Ti-Eclipse inverted epifluorescent microscope (Nikon Instruments, Melville, NY) equipped with an iXon 896 EM-CCD (Andor Technology Ltd., Belfast, Northern Ireland) camera. Fluorescence detection used a SpectraX LED light engine (Lumencor, Beaverton, OR) with paired excitation filters, dichroic mirrors, and emission filters (Prior Scientific, Rockland, MA). Images were acquired with either Plan Fluor 20 phase contrast (Ph) air objective or CFI Plan Apochromat Lambda 60x Oil immersion objective. All imaging experiments were performed a minimum of two times. Metabolite extractions ACE2-A549 and A549 cells were cultured in 6-well plates to approximately 90% confluency. Cells were then inoculated with SARS-CoV-2 for one hour, washed with PBS, then fed with fresh media. Cells were harvested at 0-, 6-, and 16-hours post-inoculum removal with 6 replicate wells harvested separately at each time point. At each collection, cells were washed with PBS, suspended with trypsin-EDTA for 5 minutes, collected and centrifuged for 5 minutes. Trypsin-EDTA was removed, and cell pellets were washed with an equi-volume of PBS before repeated centrifugation. PBS was removed and cells were resuspended in 100% methanol. Samples were vortexed in 10 x 1 sec bursts before being placed in −80 °C freezer. Vortexing and freezing was repeated 3 times to maximize macromolecule precipitation. Subsequently, methanol extracts were subjected to centrifugation at 8,000 rcf for 10 minutes to pellet cell debris and precipitate proteins. The supernatant containing the metabolites was transferred to a separate tube and dried by vacuum concentration to remove solvents. Dried metabolites were resuspended in 100 μL mass spectrometry grade 50:50 (v/v) water: acetonitrile solution immediately prior to high performance liquid chromatography-mass spectrometry (HPLC-MS) analysis. Untargeted metabolomic analysis Extracted metabolites were analyzed using HPLC-MS (Agilient 6538 Q-TOF mass spectrometer) in positive mode (resolution: ~ 20 ppm, accuracy: ~ 5 ppm, possible ionization adducts: H + , Na + ) using a Cogent Diamond Hydride HILIC column (150 x 2.1 mm). LC-MS data, consisting of mass-to-charge (m/z) values and their peak intensities, were processed and exported using MSConvert and XCMS (S1 Table). All data was log transformed and autoscaled prior to analysis using MetaboAnalyst . Statistical analyses performed included hierarchical cluster analysis (HCA), principal component analysis (PCA), partial least-squares discriminant analysis (PLS-DA), variable importance in projection (VIP) scores, volcano plot, fold change, and heatmap analysis. Pathway analysis was performed to map differentially expressed metabolite features to biological pathways using the Functional Analysis function in MetaboAnalyst (pathway library: KEGG, mass tolerance: 5 ppm, positive mode) . Pathway significance was determined using FDR-corrected significance levels of 0.05. For metabolomic data and downstream pathway analyses, there were 35 samples total (6 samples per timepoint for each cell line except for only 5 samples for the t6 timepoint in ACE2 cells). In total, 1,085 metabolite features were co-detected across all samples. To examine differences in regulation patterns across timepoints, ANOVA analysis was performed. The results of this analysis are that 152 and 372 metabolite features were differentially regulated across timepoints in ACE2-A549 cells and A549 cells alone, respectively. From here, we took these differentially-regulated features and performed pathway enrichment analyses. Thus 152 features were used for ACE2-A549 pathways and 372 features for A549 pathways. For features that are differentially regulated across timepoints of ACE2-A549 cells, 13 pathways were identified. Conversely, 8 pathways were differentially regulated across timepoints of A549 cells. E6-Vero, A549, ACE2-A549 cells. E6 Vero cells were obtained from ATCC (Manassas, VA) and grown in DMEM supplemented with 10% FBS, 1% pen-strep. A549 cells were the obtained from Chang Lab. ACE2-A549 cells were obtained from BEI Resources (NR-53821). A549 cells were propagated in Hams F-12 (Corning) media supplemented with 10% fetal bovine serum (HyClone) and 1X Penicillin/Streptomycin (Fisher Scientific). ACE2-A549 cells were supplemented with 100ug/mL Blasticidin (Gibco). SARS-CoV-2 strain WA01 was obtained from BEI Resources (NR-52281). Viral stocks were propagated and titered on E6 Vero cells in DMEM supplemented with 2% FBS and 1% pen-strep. Viral stocks were made by collecting media from infected cell cultures showing extensive cytopathic effect and centrifuged 1,000 RCF for 5 minutes to remove cellular debris. The clarified viral supernatant was then used for all experimental infection. For determination of viral infectivity by plaque assay, E6 Vero cells were cultured then incubated with viral inoculum at limiting dilutions. Dilutions employed resulted in a minimum threshold of detection at 500 plaque forming units (PFU)/mL. Following inoculation, cells were over-layered with 1% methylcellulose, DMEM supplemented with 2% FBS and 1% pen-strep and incubated for 3-4 days . Cells were then fixed and stained with 0.5% methylene blue/70% ethanol solution. Plaques were counted and the overall titer was calculated. Both A549 or ACE2-A549 cells were cultured and collected with PBS supplemented with 0.5 mM EDTA. Cells were pelleted by centrifugation then resuspended in protein extraction buffer [10 mM Tris–HCl (pH 7.5), 10 mM NaCl, 1.5 mM MgCl2, 1% NP40] supplemented with Mini-complete, EDTA Free Protease Inhibitor Cocktail (Roche Applied Science, Indianapolis, ID). Protein concentrations determined by BCA assay, and then 20 µg were loaded and separated on a 10% polyacrylamide gel. Subsequently, proteins were transferred to PVDF membranes and incubated with a goat anti-human ACE2 mAb (R&D Systems Cat # AF933) and secondary rabbit-anti goat AF488-conjugated mAb (ThermoFisher, Cat # A27012). After detection, the blot was stripped and re-probed with mouse anti-beta Actin and a goat anti-mouse DyLight550 secondary (ThermoFisher, Cat # 84540). Membranes were imaged on a Cytiva Amersham Typhoon scanner (Bucks, UK) using the Cy2 and Cy3 channels. Prior to infection, cells were seeded at 2 x 10 4 per well of an eight-chamber coverslip (Labtek Cat. No. 155411, Nunc International, Rochester, NY). At indicated times post infection, cells were then fixed with 4% paraformaldehyde in PBS for 30 minutes, washed thoroughly with PBS, and blocked in 2% bovine serum albumin (BSA) prior to antibody incubations. Primary and secondary antibodies were diluted in a PBS supplemented with 0.5% saponin, 0.125% BSA as described , and incubated for one hour at room temperature. Primary mouse anti-nucleocapsid (Thermofisher, Cat # MA1-7403) was diluted 1:500, followed by goat anti-mouse IgG labeled with Dylight 550 ThermoFisher, Cat # 84540) at 1:500. DNA was counterstained with Hoescht 3342 at 1:5000 dilution. Actin filaments stained with phalloidin-488 (Thermofisher, Cat # A12379) at 1:500. Stained cells were imaged on a Nikon Ti-Eclipse inverted epifluorescent microscope (Nikon Instruments, Melville, NY) equipped with an iXon 896 EM-CCD (Andor Technology Ltd., Belfast, Northern Ireland) camera. Fluorescence detection used a SpectraX LED light engine (Lumencor, Beaverton, OR) with paired excitation filters, dichroic mirrors, and emission filters (Prior Scientific, Rockland, MA). Images were acquired with either Plan Fluor 20 phase contrast (Ph) air objective or CFI Plan Apochromat Lambda 60x Oil immersion objective. All imaging experiments were performed a minimum of two times. ACE2-A549 and A549 cells were cultured in 6-well plates to approximately 90% confluency. Cells were then inoculated with SARS-CoV-2 for one hour, washed with PBS, then fed with fresh media. Cells were harvested at 0-, 6-, and 16-hours post-inoculum removal with 6 replicate wells harvested separately at each time point. At each collection, cells were washed with PBS, suspended with trypsin-EDTA for 5 minutes, collected and centrifuged for 5 minutes. Trypsin-EDTA was removed, and cell pellets were washed with an equi-volume of PBS before repeated centrifugation. PBS was removed and cells were resuspended in 100% methanol. Samples were vortexed in 10 x 1 sec bursts before being placed in −80 °C freezer. Vortexing and freezing was repeated 3 times to maximize macromolecule precipitation. Subsequently, methanol extracts were subjected to centrifugation at 8,000 rcf for 10 minutes to pellet cell debris and precipitate proteins. The supernatant containing the metabolites was transferred to a separate tube and dried by vacuum concentration to remove solvents. Dried metabolites were resuspended in 100 μL mass spectrometry grade 50:50 (v/v) water: acetonitrile solution immediately prior to high performance liquid chromatography-mass spectrometry (HPLC-MS) analysis. Extracted metabolites were analyzed using HPLC-MS (Agilient 6538 Q-TOF mass spectrometer) in positive mode (resolution: ~ 20 ppm, accuracy: ~ 5 ppm, possible ionization adducts: H + , Na + ) using a Cogent Diamond Hydride HILIC column (150 x 2.1 mm). LC-MS data, consisting of mass-to-charge (m/z) values and their peak intensities, were processed and exported using MSConvert and XCMS (S1 Table). All data was log transformed and autoscaled prior to analysis using MetaboAnalyst . Statistical analyses performed included hierarchical cluster analysis (HCA), principal component analysis (PCA), partial least-squares discriminant analysis (PLS-DA), variable importance in projection (VIP) scores, volcano plot, fold change, and heatmap analysis. Pathway analysis was performed to map differentially expressed metabolite features to biological pathways using the Functional Analysis function in MetaboAnalyst (pathway library: KEGG, mass tolerance: 5 ppm, positive mode) . Pathway significance was determined using FDR-corrected significance levels of 0.05. For metabolomic data and downstream pathway analyses, there were 35 samples total (6 samples per timepoint for each cell line except for only 5 samples for the t6 timepoint in ACE2 cells). In total, 1,085 metabolite features were co-detected across all samples. To examine differences in regulation patterns across timepoints, ANOVA analysis was performed. The results of this analysis are that 152 and 372 metabolite features were differentially regulated across timepoints in ACE2-A549 cells and A549 cells alone, respectively. From here, we took these differentially-regulated features and performed pathway enrichment analyses. Thus 152 features were used for ACE2-A549 pathways and 372 features for A549 pathways. For features that are differentially regulated across timepoints of ACE2-A549 cells, 13 pathways were identified. Conversely, 8 pathways were differentially regulated across timepoints of A549 cells. Differing susceptibility and productivity of A549 cells lines for SARS-CoV-2 To study metabolic shifts during SARS-CoV-2 infection, we employed A549 cells, a human lung carcinoma cell line. A549 cells are not susceptible to SARS-CoV2 infection and must be modified to express human ACE2 to allow for entry and replication . To confirm human ACE2 expression, cell extracts from A549 cells engineered to express ACE2 (ACE2-A549) and the progenitor A549 cells were subjected to western blot detection . To evaluate the capacity to support SARS-CoV-2 infection, ACE2-A549 and A549 cells were infected and analyzed to explore differences in susceptibility and relative timing of viral replication. Our analysis focused on 6 hpi, a timepoint suggested to be prior to the onset of progeny virus production, and 16 hpi, when productive viral replication should be close to its peak . Infections were performed at an MOI of 10 to ensure homogenous infection and exposure of all cells to sufficient infectious inoculum. The extent of viral infection and replication was analyzed using indirect immunofluorescent detection of the SARS-CoV-2 nucleocapsid (N) protein . ACE2-A549 cells displayed extensive N protein expression at both 6 and 16 hpi. In contrast, unmodified A549 cells displayed no N protein staining, indicating a complete lack of infection and replication following SARS-CoV-2 inoculation. An important element to analyzing metabolic profiles is the relative “health” of the cell, especially at later time points in the viral lifecycle. To evaluate whether 16 hpi exhibits extensive cell deterioration, we analyzed the distribution of actin filaments in infected cells. Cells were counterstained with both SARS-CoV-2 nucleocapsid (N) protein and phalloidin to image for the presence of actin filament assemblies during infection . We observed a distribution of cellular morphologies, but many cells retain actin filament assemblies and adhesions to the cell surface similar to the A549 cells that do not support productive replication (S1 Fig). The distribution of cellular morphology and actin staining suggests that the cells are not undergoing extensive cytopathic effect by 16 hpi. To understand how the selected timepoints correlate with viral replication, we analyzed viral titers from supernatant with (total) and without (extracellular) cellular fractions, . We compared the detection of plaque forming units (PFU) from ACE2-A549 and A549 cells at all three timepoints. In the ACE2-A549 cells, viral titer does not increase until 16 hours post inoculation in both the total and extracellular samples. A decrease at 6 hpi in the extracellular samples reflects the uptake of viral inoculum. The subsequent increase in titer at 16 hpi correlates with the release of virus. For A549 cells there was no increase in viral titer in either the total or extracellular samples beyond what is detected after inoculation of cells. These observations are consistent with the reported lack of susceptibility and permissiveness of A549 cells to SARS-CoV-2 . The differences in infection, replication, and release of infectious virus supports our selection in timepoints to explore metabolic changes in cells that can support robust productive virus infection and in cells that do not allow viral entry. Experimental design to assess metabolic differences following SARS-CoV-2 infection Global metabolomic profiling was performed in cells infected with or exposed to SARS-CoV-2. Our analysis focuses on three critical timepoints: 0 hours post infection (hpi), or immediately after inoculum removal, 6 hpi and 16 hpi . In our experimental approach, ACE2-A549 and A549 cells were inoculated at an MOI of 10 with SARS-CoV-2 to ensure homogeneity across the population of cells critical for the subsequent extraction and detection of metabolites. The cells were collected and processed at 0, 6, and 16 hpi, to analyze temporal changes in the metabolic landscape over the course of the viral lifecycle . Metabolites were extracted and processed for LC-MS metabolite detection . Samples were analyzed via LC-MS to identify molecules smaller than ~ 1000 Da, which can include hormones, oligonucleotides, peptides, and other molecular products of cellular biochemical reactions . A total of 1085 metabolites were detected across all samples and were included in all analyses. Data analyses were performed using MetaboAnalyst allowing for quantification of untargeted metabolites and identification of changes in the metabolomic phenotypes at each of the different time points . Metabolic profiling of ACE2-A549 cells during SARS-CoV-2 infection We began by analyzing metabolic changes in ACE2-A549 cells that support productive SARS-CoV-2 infection. Changes in the global metabolomic profiles of infected ACE2-A549 cells were determined using unsupervised PCA and supervised PLS-DA ( and ). From these analyses, the variance between each time point was greater than the variance between replicates within each group. We observed that the first 2 PLS-DA components represented 42% of the overall variance, further demonstrating that the three time points are distinct from each other. This is much greater than the expected 0.03% variance that would be expected from a uniformly random distribution of metabolites. Taken together, our data suggests that greater metabolomic changes occur over time although some overlap between samples is observed. Both analyses confirm that clear, non-random, differences exist between productively infected ACE2-A549 cells harvested at different each time points. To further examine metabolomic patterns that significantly change during SARS-CoV-2 infection in ACE2-A549 cells, we performed ANOVA to assess changes in metabolomic between cells harvested at 0, 6, and 16 hpi. From this analysis, 152 metabolite features with an FDR-corrected p-value < 0.05 were differentially regulated between timepoint groups. Heatmap analysis of these ANOVA metabolite features revealed temporal changes in metabolite phenotypes from 0 to 16 hpi . The variance between samples can also be observed when each metabolite feature is plotted for the individual samples (S2 Fig). Clustering analysis of similarly altered metabolite feature produces 4 main classes: reduced at 16 hpi (class 1), increased at 16 hpi (class 2), reduced at 6 hpi (class 3), and increased at 6 hpi (class 4). The majority of altered metabolite features, belonging to class 1, had the highest abundance at 0 hpi and progressively decreased from 6 to 16 hpi, suggesting a trajectory of depletion during the course of SARS-CoV-2 replication. In contrast, class 2 metabolite features increased in abundance from 0 to 16 hpi. For classes 3 and 4, metabolite features with detected changes at 6 hpi often returned to baseline abundance by 16 hpi. To derive additional biological relevance, the 152 metabolite features distinguished by ANOVA were then manually searched by m/z value in METLIN to make putative metabolite identifications . Identified putative metabolites were found in the four main classes . Select metabolites are presented to highlight some of the metabolic changes detected during infection . Each boxplot depicts the normalized fold-change of the putative metabolite description for each class with average values represented as yellow diamonds and individual replicates within each time point represented as black spots. These metabolite features may be associated with some flux or alteration in utilization of intermediates in metabolic pathways. Distinct metabolic phenotypes of A549 cells during SARS-CoV-2 infection We next sought to separate metabolic changes identified during productive infection from changes that may result from responses due to virus exposure. To accomplish this, we analyzed A549 cells, which do not express ACE2 and are refractory to infection at 0, 6 and 16 hpi after exposure to SARS-CoV-2 inoculum . As before, we evaluated metabolomic phenotypes using PCA and PLS-DA and observed larger ellipses that somewhat overlap. PLS-DA components 1 and 2 represented 43.2% of the overall variance in the dataset. Similar to the ACE2-A549 cells, these findings suggest that metabolomic differences between A549 cells harvested at different timepoints exist, despite no detectable infection or viral replication. ( and ). From the inoculated A549 cells, we identified 377 metabolite features that had an ANOVA FDR-corrected p-value < 0.05. To further visualize metabolic dysregulation across 0, 6, and 16 hpi, heatmap analysis was performed . The distribution and consistency of detected metabolite features across the replicate samples can be seen in the expanded heatmap (S3 Fig). Unlike with ACE2-A549 cells, the majority of metabolic changes are increasing quantities of metabolite features that peak at 6 hpi and remain elevated through 16 hpi . The second prominent class of metabolite features exhibits a transient increase in detection at 6 hpi, followed by reduced detection at or near levels seen at 0 hpi. The remaining significantly changed metabolite features vary with peak detection seen either at 0 or 16 hpi. Overall, the majority of changes in these cells likely represent changes in response to the inoculum that decrease by the later time points after inoculation. Comparison of metabolites between ACE2 and A549 cells We sought to further understand the differing metabolic responses between A549 cells exposed to SARS-CoV-2 and ACE2-A549 cells infected with SARS-CoV-2. Statistically significant features distinguished by ANOVA analyses for both comparisons were investigated to identify metabolite features that were either shared or unique between the two cell types. Of the 152 significant metabolite features from ACE2-A549 cells and 377 significant metabolite features from A549 cells, only 47 were significantly changed following SARS-CoV-2 exposure in both cell types . In addition, pathway analysis of metabolite features that are dysregulated during SARS-CoV-2 replication was performed . In total, 13 metabolic pathways were altered in ACE2-A549 cells with productive SARS-CoV-2 replication. A majority, 10 pathways, were involved in amino acid metabolism including alanine, aspartate, cysteine, glutamate, glycine, histidine, lysine, methionine, and threonine . Additional metabolic pathways detected included glycerophospholipid metabolism, C5-branched dibasic acid metabolism, and ascorbate metabolism. Identifying if these altered pathways are caused by productive viral replication cannot be assessed from this data alone. We compared pathway changes in cells that cannot be productively infected by SARS-CoV-2. Performing pathway analysis on the 377 metabolite features distinguished by ANOVA from the A549 cells inoculated with SARS-CoV-2 identified 8 distinct pathways. Curiously, none of these pathways overlap with those identified in from ACE2-A549 analysis . The 8 pathways unique to A549 cells mapped exclusively to lipid metabolism and included: fatty acid oxidation, activation, and metabolism; di-unsaturated fatty acid beta-oxidation, de novo fatty acid biosynthesis, omega-3 fatty acid metabolism, and carnitine shuttle . Interestingly, metabolite features associated with immunomodulatory leukotrienes were also detected. Overall, these extensive changes to lipids, specifically fatty acids, suggest a change not only towards an inflammatory state, but also a shift in energy source by the cells following exposure to a non-productive infection. Taken together, changes in metabolism between two cell lines over a 16-hour period of exposure to SARS-CoV-2 demonstrate metabolomic differences in a range of individual metabolites and pathways. Specifically, amino acid related-pathways were dysregulated in ACE2-A540 while lipid-related pathways were dysregulated in A549 cells exposed to SARS-COV2 . To study metabolic shifts during SARS-CoV-2 infection, we employed A549 cells, a human lung carcinoma cell line. A549 cells are not susceptible to SARS-CoV2 infection and must be modified to express human ACE2 to allow for entry and replication . To confirm human ACE2 expression, cell extracts from A549 cells engineered to express ACE2 (ACE2-A549) and the progenitor A549 cells were subjected to western blot detection . To evaluate the capacity to support SARS-CoV-2 infection, ACE2-A549 and A549 cells were infected and analyzed to explore differences in susceptibility and relative timing of viral replication. Our analysis focused on 6 hpi, a timepoint suggested to be prior to the onset of progeny virus production, and 16 hpi, when productive viral replication should be close to its peak . Infections were performed at an MOI of 10 to ensure homogenous infection and exposure of all cells to sufficient infectious inoculum. The extent of viral infection and replication was analyzed using indirect immunofluorescent detection of the SARS-CoV-2 nucleocapsid (N) protein . ACE2-A549 cells displayed extensive N protein expression at both 6 and 16 hpi. In contrast, unmodified A549 cells displayed no N protein staining, indicating a complete lack of infection and replication following SARS-CoV-2 inoculation. An important element to analyzing metabolic profiles is the relative “health” of the cell, especially at later time points in the viral lifecycle. To evaluate whether 16 hpi exhibits extensive cell deterioration, we analyzed the distribution of actin filaments in infected cells. Cells were counterstained with both SARS-CoV-2 nucleocapsid (N) protein and phalloidin to image for the presence of actin filament assemblies during infection . We observed a distribution of cellular morphologies, but many cells retain actin filament assemblies and adhesions to the cell surface similar to the A549 cells that do not support productive replication (S1 Fig). The distribution of cellular morphology and actin staining suggests that the cells are not undergoing extensive cytopathic effect by 16 hpi. To understand how the selected timepoints correlate with viral replication, we analyzed viral titers from supernatant with (total) and without (extracellular) cellular fractions, . We compared the detection of plaque forming units (PFU) from ACE2-A549 and A549 cells at all three timepoints. In the ACE2-A549 cells, viral titer does not increase until 16 hours post inoculation in both the total and extracellular samples. A decrease at 6 hpi in the extracellular samples reflects the uptake of viral inoculum. The subsequent increase in titer at 16 hpi correlates with the release of virus. For A549 cells there was no increase in viral titer in either the total or extracellular samples beyond what is detected after inoculation of cells. These observations are consistent with the reported lack of susceptibility and permissiveness of A549 cells to SARS-CoV-2 . The differences in infection, replication, and release of infectious virus supports our selection in timepoints to explore metabolic changes in cells that can support robust productive virus infection and in cells that do not allow viral entry. Global metabolomic profiling was performed in cells infected with or exposed to SARS-CoV-2. Our analysis focuses on three critical timepoints: 0 hours post infection (hpi), or immediately after inoculum removal, 6 hpi and 16 hpi . In our experimental approach, ACE2-A549 and A549 cells were inoculated at an MOI of 10 with SARS-CoV-2 to ensure homogeneity across the population of cells critical for the subsequent extraction and detection of metabolites. The cells were collected and processed at 0, 6, and 16 hpi, to analyze temporal changes in the metabolic landscape over the course of the viral lifecycle . Metabolites were extracted and processed for LC-MS metabolite detection . Samples were analyzed via LC-MS to identify molecules smaller than ~ 1000 Da, which can include hormones, oligonucleotides, peptides, and other molecular products of cellular biochemical reactions . A total of 1085 metabolites were detected across all samples and were included in all analyses. Data analyses were performed using MetaboAnalyst allowing for quantification of untargeted metabolites and identification of changes in the metabolomic phenotypes at each of the different time points . We began by analyzing metabolic changes in ACE2-A549 cells that support productive SARS-CoV-2 infection. Changes in the global metabolomic profiles of infected ACE2-A549 cells were determined using unsupervised PCA and supervised PLS-DA ( and ). From these analyses, the variance between each time point was greater than the variance between replicates within each group. We observed that the first 2 PLS-DA components represented 42% of the overall variance, further demonstrating that the three time points are distinct from each other. This is much greater than the expected 0.03% variance that would be expected from a uniformly random distribution of metabolites. Taken together, our data suggests that greater metabolomic changes occur over time although some overlap between samples is observed. Both analyses confirm that clear, non-random, differences exist between productively infected ACE2-A549 cells harvested at different each time points. To further examine metabolomic patterns that significantly change during SARS-CoV-2 infection in ACE2-A549 cells, we performed ANOVA to assess changes in metabolomic between cells harvested at 0, 6, and 16 hpi. From this analysis, 152 metabolite features with an FDR-corrected p-value < 0.05 were differentially regulated between timepoint groups. Heatmap analysis of these ANOVA metabolite features revealed temporal changes in metabolite phenotypes from 0 to 16 hpi . The variance between samples can also be observed when each metabolite feature is plotted for the individual samples (S2 Fig). Clustering analysis of similarly altered metabolite feature produces 4 main classes: reduced at 16 hpi (class 1), increased at 16 hpi (class 2), reduced at 6 hpi (class 3), and increased at 6 hpi (class 4). The majority of altered metabolite features, belonging to class 1, had the highest abundance at 0 hpi and progressively decreased from 6 to 16 hpi, suggesting a trajectory of depletion during the course of SARS-CoV-2 replication. In contrast, class 2 metabolite features increased in abundance from 0 to 16 hpi. For classes 3 and 4, metabolite features with detected changes at 6 hpi often returned to baseline abundance by 16 hpi. To derive additional biological relevance, the 152 metabolite features distinguished by ANOVA were then manually searched by m/z value in METLIN to make putative metabolite identifications . Identified putative metabolites were found in the four main classes . Select metabolites are presented to highlight some of the metabolic changes detected during infection . Each boxplot depicts the normalized fold-change of the putative metabolite description for each class with average values represented as yellow diamonds and individual replicates within each time point represented as black spots. These metabolite features may be associated with some flux or alteration in utilization of intermediates in metabolic pathways. We next sought to separate metabolic changes identified during productive infection from changes that may result from responses due to virus exposure. To accomplish this, we analyzed A549 cells, which do not express ACE2 and are refractory to infection at 0, 6 and 16 hpi after exposure to SARS-CoV-2 inoculum . As before, we evaluated metabolomic phenotypes using PCA and PLS-DA and observed larger ellipses that somewhat overlap. PLS-DA components 1 and 2 represented 43.2% of the overall variance in the dataset. Similar to the ACE2-A549 cells, these findings suggest that metabolomic differences between A549 cells harvested at different timepoints exist, despite no detectable infection or viral replication. ( and ). From the inoculated A549 cells, we identified 377 metabolite features that had an ANOVA FDR-corrected p-value < 0.05. To further visualize metabolic dysregulation across 0, 6, and 16 hpi, heatmap analysis was performed . The distribution and consistency of detected metabolite features across the replicate samples can be seen in the expanded heatmap (S3 Fig). Unlike with ACE2-A549 cells, the majority of metabolic changes are increasing quantities of metabolite features that peak at 6 hpi and remain elevated through 16 hpi . The second prominent class of metabolite features exhibits a transient increase in detection at 6 hpi, followed by reduced detection at or near levels seen at 0 hpi. The remaining significantly changed metabolite features vary with peak detection seen either at 0 or 16 hpi. Overall, the majority of changes in these cells likely represent changes in response to the inoculum that decrease by the later time points after inoculation. We sought to further understand the differing metabolic responses between A549 cells exposed to SARS-CoV-2 and ACE2-A549 cells infected with SARS-CoV-2. Statistically significant features distinguished by ANOVA analyses for both comparisons were investigated to identify metabolite features that were either shared or unique between the two cell types. Of the 152 significant metabolite features from ACE2-A549 cells and 377 significant metabolite features from A549 cells, only 47 were significantly changed following SARS-CoV-2 exposure in both cell types . In addition, pathway analysis of metabolite features that are dysregulated during SARS-CoV-2 replication was performed . In total, 13 metabolic pathways were altered in ACE2-A549 cells with productive SARS-CoV-2 replication. A majority, 10 pathways, were involved in amino acid metabolism including alanine, aspartate, cysteine, glutamate, glycine, histidine, lysine, methionine, and threonine . Additional metabolic pathways detected included glycerophospholipid metabolism, C5-branched dibasic acid metabolism, and ascorbate metabolism. Identifying if these altered pathways are caused by productive viral replication cannot be assessed from this data alone. We compared pathway changes in cells that cannot be productively infected by SARS-CoV-2. Performing pathway analysis on the 377 metabolite features distinguished by ANOVA from the A549 cells inoculated with SARS-CoV-2 identified 8 distinct pathways. Curiously, none of these pathways overlap with those identified in from ACE2-A549 analysis . The 8 pathways unique to A549 cells mapped exclusively to lipid metabolism and included: fatty acid oxidation, activation, and metabolism; di-unsaturated fatty acid beta-oxidation, de novo fatty acid biosynthesis, omega-3 fatty acid metabolism, and carnitine shuttle . Interestingly, metabolite features associated with immunomodulatory leukotrienes were also detected. Overall, these extensive changes to lipids, specifically fatty acids, suggest a change not only towards an inflammatory state, but also a shift in energy source by the cells following exposure to a non-productive infection. Taken together, changes in metabolism between two cell lines over a 16-hour period of exposure to SARS-CoV-2 demonstrate metabolomic differences in a range of individual metabolites and pathways. Specifically, amino acid related-pathways were dysregulated in ACE2-A540 while lipid-related pathways were dysregulated in A549 cells exposed to SARS-COV2 . In this study, we sought to understand the nature of cellular metabolic shifts in response to SARS-CoV-2 infection. To distinguish changes associated with viral replication from exposure to infectious virus, we compared ACE2-expressing A549 cells that are susceptible and support productive viral replication with A549 cells that are not susceptible to infection. We chose time points that represent early, intermediate, and late stages of viral replication to evaluate the temporal changes in metabolites following inoculation. Our metabolic pathway analysis found 152 and 377 significantly changed metabolites in ACE2-A549 and A549 cells, respectively. Surprisingly, there was limited overlap in altered metabolites or altered metabolic pathways between cells undergoing productive infection and those exposed to infectious virus. Critically, we identified alterations in pathways that are potentially involved in either productive viral infection or in cellular anti-viral responses to SARS-CoV-2. Consequences of productive viral infection The initial focus of our analysis was the changes to cellular metabolism induced by active viral replication. The ACE2-A549 cells are a widely used model that we observed to be both susceptible to SARS-CoV-2 infection and permissive for productive viral replication . We analyzed metabolic shifts immediately after virion entry (0 hpi), a mid-point of viral replication prior to virion production (6 hpi), and a late timepoint when new virions are being released from infected cells (16 hpi) . Overall, we identified four different classes of metabolites based on the relative increase or decrease in detection between each time point. We used these changes to identify pathways that were altered by active SARS-CoV-2 replication. Most notably, we observed that the majority of altered metabolic pathways were associated with amino acid metabolism. Within the identified pathways, we identified L-Glutamic Acid as a major putative metabolite that was significantly reduced from 0 to 16 hpi. A previous study demonstrated that SARS-CoV-2 infection rewires carbon entry into the TCA cycle . Mullen et al. showed that oxidative metabolism of glutamine through the TCA cycle was reduced during SARS-CoV-2 infection in favor of pyruvate utilization via pyruvate carboxylase . This shift increased levels of oxaloacetate and also maintain synthesis of aspartate, which is used to synthesize pyrimidine nucleotides . Interestingly, both glutamate and aspartate metabolic pathways were considered significant in our analysis of SARS-CoV-2 infected ACE2-A549 cells. In addition to changes to L-glutamic acid and aspartate, the metabolic shifts observed in SARS-CoV-2 infected ACE2-A549 cells reflect similar changes observed in COVID-19 positive patient serum samples . It is notable that our cell culture model identified similar metabolic pathways being disrupted during SARS-CoV-2 infection, even at 6 hpi. This indicates that metabolic screening of different laboratory-based model systems may be able to accurately generate data on potential biomarkers for SARS-CoV-2 and potentially other infectious diseases. Metabolites responding to viral inoculation While metabolic changes due to active viral replication are important, not all cells within a tissue or organ system are equally susceptible to virus infection. Thus, we hypothesized that uninfected cells that do not support SARS-CoV-2 infection can respond to virus exposure with altered metabolism leading to further metabolic dysfunction. To test this hypothesis, we analyzed the metabolic profile of A549 cells that do not support viral entry following exposure to the same infectious virus inoculum as the ACE2-A549 cells. Our data confirmed that lack of ACE2 expression in A549 cells resulted in a complete lack of infection and replication following SARS-CoV-2 inoculation. While we have no evidence of viral replication in A549 cells, the resulting changes in the metabolic profile of these cells indicates that they are responding to the virus inoculum. Specifically, fatty acid catabolic (β-oxidation) and anabolic (de novo fatty acid synthesis) pathways were significantly altered following A549 cell exposure to infectious virus. Lipid dysregulation has been a hallmark of COVID-19 pathology in patients and a hallmark of disease severity and progression . Both pathways converge on acetyl-CoA, a critical molecule in the breakdown of fatty acids and the synthesis of other lipid types, such as cholesterol, which can be transformed into other steroids with pro and anti-inflammatory mechanisms . This is further supported by the significant number of identified metabolites associated with leukotriene and omega-3 fatty acid metabolism within the A549 cells. Increased leukotriene production is connected to COVID-19 through transcriptional and metabolic studies from patient serum and infected monocytes . Another significant metabolite in our profile of the A549 cells was palmitoyl-CoA, a major component in the synthesis of ceramide and sphingolipids . Previous studies of COVID-19 patient serum samples found distinct increases in sphingosine and ceramides . These increases were distinct between patients with mild disease and those in intensive care . The data suggests that uninfected cells respond to SARS-CoV-2, altering metabolic profiles and possibly increasing the production of pro- or anti-inflammatory biomolecules and enzyme cofactors . Thus, even uninfected cells may be contributing to the overall pathology observed in COVID-19 patients. The initial focus of our analysis was the changes to cellular metabolism induced by active viral replication. The ACE2-A549 cells are a widely used model that we observed to be both susceptible to SARS-CoV-2 infection and permissive for productive viral replication . We analyzed metabolic shifts immediately after virion entry (0 hpi), a mid-point of viral replication prior to virion production (6 hpi), and a late timepoint when new virions are being released from infected cells (16 hpi) . Overall, we identified four different classes of metabolites based on the relative increase or decrease in detection between each time point. We used these changes to identify pathways that were altered by active SARS-CoV-2 replication. Most notably, we observed that the majority of altered metabolic pathways were associated with amino acid metabolism. Within the identified pathways, we identified L-Glutamic Acid as a major putative metabolite that was significantly reduced from 0 to 16 hpi. A previous study demonstrated that SARS-CoV-2 infection rewires carbon entry into the TCA cycle . Mullen et al. showed that oxidative metabolism of glutamine through the TCA cycle was reduced during SARS-CoV-2 infection in favor of pyruvate utilization via pyruvate carboxylase . This shift increased levels of oxaloacetate and also maintain synthesis of aspartate, which is used to synthesize pyrimidine nucleotides . Interestingly, both glutamate and aspartate metabolic pathways were considered significant in our analysis of SARS-CoV-2 infected ACE2-A549 cells. In addition to changes to L-glutamic acid and aspartate, the metabolic shifts observed in SARS-CoV-2 infected ACE2-A549 cells reflect similar changes observed in COVID-19 positive patient serum samples . It is notable that our cell culture model identified similar metabolic pathways being disrupted during SARS-CoV-2 infection, even at 6 hpi. This indicates that metabolic screening of different laboratory-based model systems may be able to accurately generate data on potential biomarkers for SARS-CoV-2 and potentially other infectious diseases. While metabolic changes due to active viral replication are important, not all cells within a tissue or organ system are equally susceptible to virus infection. Thus, we hypothesized that uninfected cells that do not support SARS-CoV-2 infection can respond to virus exposure with altered metabolism leading to further metabolic dysfunction. To test this hypothesis, we analyzed the metabolic profile of A549 cells that do not support viral entry following exposure to the same infectious virus inoculum as the ACE2-A549 cells. Our data confirmed that lack of ACE2 expression in A549 cells resulted in a complete lack of infection and replication following SARS-CoV-2 inoculation. While we have no evidence of viral replication in A549 cells, the resulting changes in the metabolic profile of these cells indicates that they are responding to the virus inoculum. Specifically, fatty acid catabolic (β-oxidation) and anabolic (de novo fatty acid synthesis) pathways were significantly altered following A549 cell exposure to infectious virus. Lipid dysregulation has been a hallmark of COVID-19 pathology in patients and a hallmark of disease severity and progression . Both pathways converge on acetyl-CoA, a critical molecule in the breakdown of fatty acids and the synthesis of other lipid types, such as cholesterol, which can be transformed into other steroids with pro and anti-inflammatory mechanisms . This is further supported by the significant number of identified metabolites associated with leukotriene and omega-3 fatty acid metabolism within the A549 cells. Increased leukotriene production is connected to COVID-19 through transcriptional and metabolic studies from patient serum and infected monocytes . Another significant metabolite in our profile of the A549 cells was palmitoyl-CoA, a major component in the synthesis of ceramide and sphingolipids . Previous studies of COVID-19 patient serum samples found distinct increases in sphingosine and ceramides . These increases were distinct between patients with mild disease and those in intensive care . The data suggests that uninfected cells respond to SARS-CoV-2, altering metabolic profiles and possibly increasing the production of pro- or anti-inflammatory biomolecules and enzyme cofactors . Thus, even uninfected cells may be contributing to the overall pathology observed in COVID-19 patients. Cellular models for SARS-CoV-2 infection are incredibly important for the initial testing of interventions that directly target viral replication. Through our metabolomic profiling, we identified metabolites and metabolic profiles that are associated with both active viral infection and exposure to infectious virus. Our analysis identified a range of metabolites and metabolic pathways altered by productive viral replication. Further work will be required to understand if these metabolites promoting viral replication are specific for SARS-CoV-2 infection and the relationship to changes in primary cells and organs. While cellular metabolism is often thought to be manipulated by the virus for its own ends, it is also connected to antiviral responses . Cells can produce antiviral metabolites or inhibit metabolic pathways to hinder viral replication . As with productive replication, further experiments that either promote shifts in leukotrienes or other inflammatory molecules will need to be performed to characterize their effects on SARS-CoV-2 infection. It is also possible that the identified metabolite profiles can be developed as biomarkers of infection that could be used for surveillance testing or as a predictive tool for risk evaluation of severe disease. The similarities of our results to metabolic shifts observed in patients suggest a potential platform for methodological development. Discriminatory metabolites defining infection can be correlated to patient metabolic profiles to facilitate our understanding of SARS-CoV-2 induced pathologies. Through both the immediate findings and the development of more complex models, we hope to increase our understanding of how SARS-CoV-2 replication and spread correlates to disease. Through that understanding, we can then find better therapeutics to limit morbidity and mortality from COVID-19. S1 Fig Cellular morphology during productive and non-productive SARS-CoV-2 infection. A) Further images of ACE2-A549 and A549 cells from Fig 1B. Presented at three-channel merged images of the transmitted light (greyscale), SARS-CoV-2 anti-nucleocapsid antibody (red), and Dapi (blue, nuclei). Scale bar 100 µm. B) Further images of infected ACE2-A549 and A549 cells at 16hpi as in . The three channel merged images are of SARS-CoV-2 anti-nucleocapsid antibody (red), actin filaments stained with phalloidin (green) and Dapi (blue, nuclei). Scale bars are 10 µm. (EPS) S2 Fig Heatmap analysis of significant metabolites (n = 152) for all ACE2-A549 samples at each time point. (EPS) S3 Fig Heatmap analysis of significant metabolites (n = 377) for all A549 samples at each time point. (EPS) S1 Table Raw mass spec values for all samples and timepoints. (XLSX) |
Artificial intelligence in ophthalmology: The path to the real-world clinic | 690c1586-9ffb-4f40-b4ba-748e48911f3f | 10394169 | Ophthalmology[mh] | In recent years, artificial intelligence (AI), including machine learning and deep learning , has made a great impact on society worldwide. This is stimulated by the advent of powerful algorithms, exponential data growth, and computing hardware advances. , In the medical and healthcare fields, numerous studies have validated that AI exhibited robust performance in disease diagnoses and treatment response prediction. , , , , , , For example, a deep-learning system developed on computed tomography (CT) images, can distinguish patients with COVID-19 pneumonia from patients with other common types of pneumonia and normal controls with an area under the curve (AUC) of 0.971. A machine-learning model trained using dual-energy CT radiomics provides a significant additive value for response prediction (AUC = 0.75) in metastatic melanoma patients prior to immunotherapy. In ophthalmology, the application of AI is very promising, given that the diagnoses and therapeutic monitoring of ocular diseases often rely heavily on image recognition . Based on this technique, diabetic retinopathy (DR), glaucoma, and age-related macular degeneration (AMD) can be accurately detected from fundus images, , , and keratitis, pterygium, and cataract can be precisely identified from slit-lamp images. , , Detailed information that describes different imaging types for different purposes in ophthalmology and corresponding AI applications is summarized in . In addition, AI may support eye doctors in generating individualized views of patients along their care pathways and guide clinical decisions. For instance, the visual prognosis after 12 months in neovascular AMD patients receiving ranibizumab can be predicted by an AI-based model that is developed using their clinical data (e.g., ocular coherence tomography [OCT] and best-corrected visual acuity [BCVA]) collected at baseline and in the first 3 months. This method may assist eye doctors in better managing the expectations of patients appropriately during their treatment process. Although there are many reasons to be hopeful for this transformation brought on by AI, hurdles remain to the successful deployment of AI in real-world clinical settings. In this review, we first retrace the current main AI applications in ophthalmology. Second, we describe the major challenges of AI clinical translation. Third, we discuss avenues that could facilitate the real implementation of AI into clinical practice. By stressing issues in the context of present AI applications for clinical ophthalmology, we wish to provide concepts to help promote significative investigations that will finally translate to real-world clinical use. DR The prevalence of diabetes has tripled in the past two decades worldwide. It can cause microvascular damage and retinal dysfunction as a result of chronic exposure to hyperglycemia, and 34.6% of people with diabetes develop DR, which is a leading cause of vision loss in working-age adults (20–74 years). , , In 2019, approximately 160 million of the population suffered from some form of DR, of whom 47 million suffered from sight-threatening DR. By 2045, this number is projected to increase to 242 million for DR and 71 million for sight-threatening DR. Early identification and timely treatment of sight-threatening DR can reduce 95% of blindness from this cause. Therefore, DR screening programs for patients with diabetes are suggested by the World Health Organization. However, conducting these screening programs on a large scale often requires a great deal of manpower and material and financial resources, which is difficult to realize in many low-income and middle-income countries. For this reason, exploring an approach that can reduce costs and increase the efficiency of DR screening programs should be a high priority. The emergence of AI provides potential new solutions, such as applying AI to retinal imaging for automated DR screening and referral . Using deep learning, numerous studies have developed intelligent systems that can accurately detect DR from fundus images. Gulshan et al. developed a deep-learning system using 128,175 fundus images (69,573 subjects) and evaluated the system in two external datasets with 11,711 fundus images (5,871 subjects). Their system achieved an AUC over 0.99 in DR screening. Ting et al. reported a deep-learning system with an AUC of 0.936, a sensitivity of 90.5%, and a specificity of 91.6% in identifying referable DR, and an AUC of 0.958, a sensitivity of 100%, and a specificity of 91.1% in discerning sight-threatening DR. Tang et al. established a deep-learning system for detecting referable DR and sight-threatening DR from ultra-widefield fundus (UWF) images, which had a larger retinal field of view and contained more information about lesions (especially peripheral lesions) compared with traditional fundus images. The AUCs of this system for identifying both referable DR and sight-threatening DR were over 0.9 in the external validation datasets. Justin et al. trained a UWF-image-based deep-learning model using ResNet34, and this model had an AUC of 0.915 for DR detection in the test set. Their results indicated that the model developed based on UWF images may be more accurate than that based on traditional fundus images because only the UWF-image-based model could detect peripheral DR lesions without pharmacologic pupil dilation. In addition to screening for patients with referable DR (i.e., moderate and worse DR) and sight-threatening DR, detecting early-stage DR is also crucial. Evidence indicates that proper intervention to keep glucose, lipid profiles, and blood pressure under control at an early stage can significantly delay the DR progression and even reverse mild non-proliferative DR (NPDR) to the DR-free stage. Dai et al. reported a deep-learning system named DeepDR with robust performance in detecting early to late stages of DR. Their system was developed based on 466,247 fundus images (121,342 diabetic patients) that were graded for mild NPDR, moderate NPDR, severe NPDR, proliferative DR (PDR), and non-DR by a centered reading group including 133 certified ophthalmologists. The evaluation was performed on 209,322 images collected from three external datasets, China National Diabetic Complications Study (CNDCS), Nicheng Diabetes Screening Project (NDSP), and Eye Picture Archive Communication System (EyePACS), and the average AUCs of the system in these datasets were 0.940, 0.944, and 0.943, respectively. Besides, predicting the onset and progression of DR is essential for the mitigation of the rising threat of DR. Bora et al. created a deep-learning system using image data obtained from 369,074 patients to predict the risk of patients with diabetes developing DR within 2 years, and the system achieved an AUC of 0.70 in the external validation set. This automated risk-stratification tool may help optimize DR screening intervals and decrease costs while improving vision-related outcomes. Arcadu et al. developed a predictive DR progression algorithm (AUC = 0.79) based on 14,070 stereoscopic seven-field fundus images to provide a timely referral for a fast DR-progressing patient, enabling initiation of treatment prior to irreversible vision loss occurring. Glaucoma Glaucoma, characterized by cupping of the optic disc and visual field impairment, is the most frequent cause of irreversible blindness, affecting more than 70 million people worldwide. , , Due to population growth and aging globally, the number of patients with glaucoma will increase to 112 million by 2040. Most vision loss caused by glaucoma can be prevented via early diagnosis and timely treatment. However, identifying glaucoma at an early stage, particularly for primary open-angle glaucoma (POAG), normal tension glaucoma (NTG), and chronic primary angle-closure glaucoma (CPACG), is challenging for the following two reasons. First, POAG, NTG, and CPACG are often painless, and visual field defects are inconspicuous at an early stage. Therefore, self-detection of these types of glaucoma by affected people usually occurs at a relatively late stage when central visual acuity is reduced. , Second, the primary approach to detect glaucoma is the examination of the optic disc and retinal nerve fiber layer by a glaucoma specialist through ophthalmoscopy or fundus images. , , Such manual optic disc assessment is time consuming and labor intensive, which is infeasible to implement in large populations. Accordingly, improvements in screening methods for glaucoma are necessary. AI may pave the road for cost-effective glaucoma screening programs, such as detecting glaucoma from fundus images or OCT images in an automated fashion . Li et al. reported a deep-learning system with excellent performance in detecting referable glaucomatous optic neuropathy (GON) from fundus images. Specifically, they adopted the Inception-v3 algorithm to train the system and evaluated it in 8,000 images. Their system achieved an AUC of 0.986 with a sensitivity of 95.6% and a specificity of 92.0% for discerning referable GON. As fundus imaging is intrinsically a two-dimensional (2D) imaging modality observing the surface of the optic nerve head but glaucoma is a three-dimensional (3D) disease with depth-resolved structural changes, fundus imaging may not able to reach a level of accuracy that could be acquired through OCT, a 3D imaging modality. , Ran et al. trained and tested a 3D deep-learning system using 6,921 spectral-domain OCT volumes of optic disc cubes from 1,384,200 2D cross-sectional scans. Their 3D system reached an AUC of 0.969 in detecting GON, significantly outperforming a 2D deep-learning system trained with fundus images (AUC, 0.921). This 3D system also had performance comparable with two glaucoma specialists with over 10 years of experience. The heatmaps indicated that the features leveraged by the 3D system for GON detection were similar to those leveraged by glaucoma specialists. Primary angle-closure glaucoma is avoidable if the progress of angle closure can be stopped at the early stages. Fu et al. developed a deep-learning system using 4,135 anterior-segment OCT images from 2,113 individuals for the automated angle-closure detection. The system achieved an AUC of 0.96 with a sensitivity of 0.90 and a specificity of 0.92, which were better than those of the qualitative feature-based system (AUC, 0.90; sensitivity, 0.79; specificity, 0.87). These results indicate that deep learning may mine a broader range of details of anterior-segment OCT images than the qualitative features (e.g., angle opening distance, angle recess area, and iris area) determined by clinicians. AI can also be used to predict glaucoma progression. Yousefi et al. reported an unsupervised machine-learning method to identify longitudinal glaucoma progression based on visual fields from 2,085 eyes of 1,214 subjects. They found that this machine-learning analysis detected progressing eyes earlier (3.5 years) than other methods such as global mean deviation (5.2 years), region-wise (4.5 years), and point-wise (3.9 years). Wang et al. proposed an AI approach, the archetype method, to detect visual field progression in glaucoma with an accuracy of 0.77. Moreover, this AI approach had a significantly higher agreement (kappa, 0.48) with the clinician assessment than other existing methods (e.g., the permutation of point-wise linear regression). AMD AMD, a disease that affects the macular area of the retina, often causes progressive loss of central vision. Age is the strongest risk factor for AMD and almost all late AMD cases happen in people at ages over 60 years. With the aging population, AMD will continue to be a major cause of vision impairment worldwide. The number of AMD patients will reach 288 million in 2040, denoting the substantial global burden of AMD. Consequently, screening for patients with AMD (especially neovascular AMD) and providing suitable medical interventions in a timely manner can reduce vision loss and improve patient visual outcomes. AI has the potential to facilitate the automated detection of AMD and prediction of AMD progression . Peng et al. constructed and tested a deep-learning system (DeepSeeNet) using 59,302 fundus images from the longitudinal follow-up of 4,549 subjects from the Age-Related Eye Disease Study (AREDS). DeepSeeNet performed well on patient-based multi-class classification with AUCs of 0.94, 0.93, and 0.97 in detecting large drusen, pigmentary abnormalities, and late AMD, respectively. Burlina et al. reported a deep-learning system established by AlexNet based on over 130,000 fundus images from 4,613 patients to screen for referable AMD, and their system achieved an average AUC of 0.95. The referable AMD in their study refers to eyes with one of the following conditions: (1) large drusen (size larger than 125 μm); (2) multiple medium-sized drusen and pigmentation abnormalities; (3) choroidal neovascularization (CNV); (4) geographic atrophy. AI also has the potential to predict the possibility of progression to late AMD, guiding high-risk patients to start preventive care early (e.g., eating healthy food, abandoning smoking, and taking supplements) and assisting clinicians to decide the interval of the patient’s follow-up examination. In patients diagnosed with wet AMD in one eye, Yim et al. introduced an AI system to predict conversion to wet AMD in the second eye. Their system was constructed by a segmentation network, diagnosis network, and prediction network based on 130,327 3D OCT images and corresponding automatic tissue maps for predicting progression to wet AMD within a clinically actionable time window (6 months). The system achieved 80% sensitivity at 55% specificity and 34% sensitivity at 90% specificity. As both genetic and environmental factors can affect the etiology of AMD, Yan et al. developed an AI approach with a modified deep convolutional neural network (CNN) using 52 AMD-associated genetic variants and 31,262 fundus images from 1,351 individuals from the AREDS to predict whether an eye would progress to late AMD. Their results showed that the approach based on both fundus images and genotypes could predict late AMD progression with an AUC of 0.85, whereas the approach based on fundus images alone achieved an AUC of 0.81. Other retinal diseases Numerous studies also have found that AI could be applied to promote the automated detection of other retinal diseases from clinical images to provide timely referrals for positive cases, solving the issues caused by the unbalanced distribution of ophthalmic medical resources. Milea et al. developed a deep-learning system using 14,341 fundus images to detect papilledema. This system achieved an AUC of 0.96 in the external test dataset consisting of 1,505 images. Brown et al. established a deep-learning system based on 5,511 retinal images captured by RetCam to diagnose plus disease in retinopathy of prematurity (ROP), a leading cause of blindness in childhood. The AUC of their system was 0.98 with a sensitivity of 93% and a specificity of 93%. In terms of detecting peripheral retinal diseases, such as lattice degeneration and retinal breaks, Li et al. trained models with four different deep-learning algorithms (InceptionResNetV2, ResNet50, InceptionV3, and VGG16) using 5,606 UWF images. They found that InceptionResNetV2 had the best performance, which achieved an AUC of 0.996 with 98.7% sensitivity and 99.2% specificity. In addition, AI has also been employed in the automated identification of retinal detachment, pathologic myopia, polypoidal choroidal vasculopathy, etc. The prevalence of diabetes has tripled in the past two decades worldwide. It can cause microvascular damage and retinal dysfunction as a result of chronic exposure to hyperglycemia, and 34.6% of people with diabetes develop DR, which is a leading cause of vision loss in working-age adults (20–74 years). , , In 2019, approximately 160 million of the population suffered from some form of DR, of whom 47 million suffered from sight-threatening DR. By 2045, this number is projected to increase to 242 million for DR and 71 million for sight-threatening DR. Early identification and timely treatment of sight-threatening DR can reduce 95% of blindness from this cause. Therefore, DR screening programs for patients with diabetes are suggested by the World Health Organization. However, conducting these screening programs on a large scale often requires a great deal of manpower and material and financial resources, which is difficult to realize in many low-income and middle-income countries. For this reason, exploring an approach that can reduce costs and increase the efficiency of DR screening programs should be a high priority. The emergence of AI provides potential new solutions, such as applying AI to retinal imaging for automated DR screening and referral . Using deep learning, numerous studies have developed intelligent systems that can accurately detect DR from fundus images. Gulshan et al. developed a deep-learning system using 128,175 fundus images (69,573 subjects) and evaluated the system in two external datasets with 11,711 fundus images (5,871 subjects). Their system achieved an AUC over 0.99 in DR screening. Ting et al. reported a deep-learning system with an AUC of 0.936, a sensitivity of 90.5%, and a specificity of 91.6% in identifying referable DR, and an AUC of 0.958, a sensitivity of 100%, and a specificity of 91.1% in discerning sight-threatening DR. Tang et al. established a deep-learning system for detecting referable DR and sight-threatening DR from ultra-widefield fundus (UWF) images, which had a larger retinal field of view and contained more information about lesions (especially peripheral lesions) compared with traditional fundus images. The AUCs of this system for identifying both referable DR and sight-threatening DR were over 0.9 in the external validation datasets. Justin et al. trained a UWF-image-based deep-learning model using ResNet34, and this model had an AUC of 0.915 for DR detection in the test set. Their results indicated that the model developed based on UWF images may be more accurate than that based on traditional fundus images because only the UWF-image-based model could detect peripheral DR lesions without pharmacologic pupil dilation. In addition to screening for patients with referable DR (i.e., moderate and worse DR) and sight-threatening DR, detecting early-stage DR is also crucial. Evidence indicates that proper intervention to keep glucose, lipid profiles, and blood pressure under control at an early stage can significantly delay the DR progression and even reverse mild non-proliferative DR (NPDR) to the DR-free stage. Dai et al. reported a deep-learning system named DeepDR with robust performance in detecting early to late stages of DR. Their system was developed based on 466,247 fundus images (121,342 diabetic patients) that were graded for mild NPDR, moderate NPDR, severe NPDR, proliferative DR (PDR), and non-DR by a centered reading group including 133 certified ophthalmologists. The evaluation was performed on 209,322 images collected from three external datasets, China National Diabetic Complications Study (CNDCS), Nicheng Diabetes Screening Project (NDSP), and Eye Picture Archive Communication System (EyePACS), and the average AUCs of the system in these datasets were 0.940, 0.944, and 0.943, respectively. Besides, predicting the onset and progression of DR is essential for the mitigation of the rising threat of DR. Bora et al. created a deep-learning system using image data obtained from 369,074 patients to predict the risk of patients with diabetes developing DR within 2 years, and the system achieved an AUC of 0.70 in the external validation set. This automated risk-stratification tool may help optimize DR screening intervals and decrease costs while improving vision-related outcomes. Arcadu et al. developed a predictive DR progression algorithm (AUC = 0.79) based on 14,070 stereoscopic seven-field fundus images to provide a timely referral for a fast DR-progressing patient, enabling initiation of treatment prior to irreversible vision loss occurring. Glaucoma, characterized by cupping of the optic disc and visual field impairment, is the most frequent cause of irreversible blindness, affecting more than 70 million people worldwide. , , Due to population growth and aging globally, the number of patients with glaucoma will increase to 112 million by 2040. Most vision loss caused by glaucoma can be prevented via early diagnosis and timely treatment. However, identifying glaucoma at an early stage, particularly for primary open-angle glaucoma (POAG), normal tension glaucoma (NTG), and chronic primary angle-closure glaucoma (CPACG), is challenging for the following two reasons. First, POAG, NTG, and CPACG are often painless, and visual field defects are inconspicuous at an early stage. Therefore, self-detection of these types of glaucoma by affected people usually occurs at a relatively late stage when central visual acuity is reduced. , Second, the primary approach to detect glaucoma is the examination of the optic disc and retinal nerve fiber layer by a glaucoma specialist through ophthalmoscopy or fundus images. , , Such manual optic disc assessment is time consuming and labor intensive, which is infeasible to implement in large populations. Accordingly, improvements in screening methods for glaucoma are necessary. AI may pave the road for cost-effective glaucoma screening programs, such as detecting glaucoma from fundus images or OCT images in an automated fashion . Li et al. reported a deep-learning system with excellent performance in detecting referable glaucomatous optic neuropathy (GON) from fundus images. Specifically, they adopted the Inception-v3 algorithm to train the system and evaluated it in 8,000 images. Their system achieved an AUC of 0.986 with a sensitivity of 95.6% and a specificity of 92.0% for discerning referable GON. As fundus imaging is intrinsically a two-dimensional (2D) imaging modality observing the surface of the optic nerve head but glaucoma is a three-dimensional (3D) disease with depth-resolved structural changes, fundus imaging may not able to reach a level of accuracy that could be acquired through OCT, a 3D imaging modality. , Ran et al. trained and tested a 3D deep-learning system using 6,921 spectral-domain OCT volumes of optic disc cubes from 1,384,200 2D cross-sectional scans. Their 3D system reached an AUC of 0.969 in detecting GON, significantly outperforming a 2D deep-learning system trained with fundus images (AUC, 0.921). This 3D system also had performance comparable with two glaucoma specialists with over 10 years of experience. The heatmaps indicated that the features leveraged by the 3D system for GON detection were similar to those leveraged by glaucoma specialists. Primary angle-closure glaucoma is avoidable if the progress of angle closure can be stopped at the early stages. Fu et al. developed a deep-learning system using 4,135 anterior-segment OCT images from 2,113 individuals for the automated angle-closure detection. The system achieved an AUC of 0.96 with a sensitivity of 0.90 and a specificity of 0.92, which were better than those of the qualitative feature-based system (AUC, 0.90; sensitivity, 0.79; specificity, 0.87). These results indicate that deep learning may mine a broader range of details of anterior-segment OCT images than the qualitative features (e.g., angle opening distance, angle recess area, and iris area) determined by clinicians. AI can also be used to predict glaucoma progression. Yousefi et al. reported an unsupervised machine-learning method to identify longitudinal glaucoma progression based on visual fields from 2,085 eyes of 1,214 subjects. They found that this machine-learning analysis detected progressing eyes earlier (3.5 years) than other methods such as global mean deviation (5.2 years), region-wise (4.5 years), and point-wise (3.9 years). Wang et al. proposed an AI approach, the archetype method, to detect visual field progression in glaucoma with an accuracy of 0.77. Moreover, this AI approach had a significantly higher agreement (kappa, 0.48) with the clinician assessment than other existing methods (e.g., the permutation of point-wise linear regression). AMD, a disease that affects the macular area of the retina, often causes progressive loss of central vision. Age is the strongest risk factor for AMD and almost all late AMD cases happen in people at ages over 60 years. With the aging population, AMD will continue to be a major cause of vision impairment worldwide. The number of AMD patients will reach 288 million in 2040, denoting the substantial global burden of AMD. Consequently, screening for patients with AMD (especially neovascular AMD) and providing suitable medical interventions in a timely manner can reduce vision loss and improve patient visual outcomes. AI has the potential to facilitate the automated detection of AMD and prediction of AMD progression . Peng et al. constructed and tested a deep-learning system (DeepSeeNet) using 59,302 fundus images from the longitudinal follow-up of 4,549 subjects from the Age-Related Eye Disease Study (AREDS). DeepSeeNet performed well on patient-based multi-class classification with AUCs of 0.94, 0.93, and 0.97 in detecting large drusen, pigmentary abnormalities, and late AMD, respectively. Burlina et al. reported a deep-learning system established by AlexNet based on over 130,000 fundus images from 4,613 patients to screen for referable AMD, and their system achieved an average AUC of 0.95. The referable AMD in their study refers to eyes with one of the following conditions: (1) large drusen (size larger than 125 μm); (2) multiple medium-sized drusen and pigmentation abnormalities; (3) choroidal neovascularization (CNV); (4) geographic atrophy. AI also has the potential to predict the possibility of progression to late AMD, guiding high-risk patients to start preventive care early (e.g., eating healthy food, abandoning smoking, and taking supplements) and assisting clinicians to decide the interval of the patient’s follow-up examination. In patients diagnosed with wet AMD in one eye, Yim et al. introduced an AI system to predict conversion to wet AMD in the second eye. Their system was constructed by a segmentation network, diagnosis network, and prediction network based on 130,327 3D OCT images and corresponding automatic tissue maps for predicting progression to wet AMD within a clinically actionable time window (6 months). The system achieved 80% sensitivity at 55% specificity and 34% sensitivity at 90% specificity. As both genetic and environmental factors can affect the etiology of AMD, Yan et al. developed an AI approach with a modified deep convolutional neural network (CNN) using 52 AMD-associated genetic variants and 31,262 fundus images from 1,351 individuals from the AREDS to predict whether an eye would progress to late AMD. Their results showed that the approach based on both fundus images and genotypes could predict late AMD progression with an AUC of 0.85, whereas the approach based on fundus images alone achieved an AUC of 0.81. Numerous studies also have found that AI could be applied to promote the automated detection of other retinal diseases from clinical images to provide timely referrals for positive cases, solving the issues caused by the unbalanced distribution of ophthalmic medical resources. Milea et al. developed a deep-learning system using 14,341 fundus images to detect papilledema. This system achieved an AUC of 0.96 in the external test dataset consisting of 1,505 images. Brown et al. established a deep-learning system based on 5,511 retinal images captured by RetCam to diagnose plus disease in retinopathy of prematurity (ROP), a leading cause of blindness in childhood. The AUC of their system was 0.98 with a sensitivity of 93% and a specificity of 93%. In terms of detecting peripheral retinal diseases, such as lattice degeneration and retinal breaks, Li et al. trained models with four different deep-learning algorithms (InceptionResNetV2, ResNet50, InceptionV3, and VGG16) using 5,606 UWF images. They found that InceptionResNetV2 had the best performance, which achieved an AUC of 0.996 with 98.7% sensitivity and 99.2% specificity. In addition, AI has also been employed in the automated identification of retinal detachment, pathologic myopia, polypoidal choroidal vasculopathy, etc. Cataract In the past 20 years, although the prevalence of cataracts has been decreasing due to the increasing rates of cataract surgery because of improved techniques and active surgical initiatives, it still affects 95 million people worldwide. Cataract remains the leading cause of blindness (accounting for 50% of blindness), especially in low-income and middle-income countries. Therefore, exploring a set of strategies to promote cataract screening and related ophthalmic services is imperative. Recent advancements in AI may help achieve this goal, such as diagnosis and quantitative classification of age-related cataract from slit-lamp images . Keenan et al. trained deep-learning models, named DeepLensNet, to detect and quantify nuclear sclerosis (NS) from 45° slit-lamp images and cortical lens opacity (CLO) and posterior subcapsular cataract (PSC) from retroillumination images. NS grading was considered on 0.9–7.1 scale. CLO and PSC grading were both considered as percentages. In the full test set, mean squared error values for DeepLensNet were 0.23 for NS, 13.1 for CLO, and 16.6 for PSC. The results indicate that this framework can perform automated and quantitative classification of cataract severity with high accuracy, which has the potential to increase the accessibility of cataract evaluation globally. Except for slit-lamp images, Tham et al. found that fundus images could also be used to develop an AI system for cataract screening. Based on 25,742 fundus images, they constructed a framework with ResNet50 and XGBoost classifier for the automated detection of visually significant cataracts (BCVA < 20/60), achieving AUCs of 0.916–0.965 in three external test sets. One merit of this system is that it can screen for cataracts with a single imaging modality, which is different from the traditional method that requires slit-lamp and retroillumination images alongside BCVA measurement. The other merit is that this system can be readily integrated into existing fundus-image-based AI systems, allowing simultaneous screening for other posterior-segment diseases. Other than cataract screening, AI can also offer real-time guidance for phacoemulsification cataract surgery (PCS). Nespolo et al. invented a computer vision-based platform using a region-based CNN (Faster R-CNN) built on ResNet50, a k-means clustering technique, and an optical-flow-tracking technology to enhance the surgeon experience during the PCS. Specifically, this platform can be used to receive frames from the video source, locate the pupil, discern the surgical phase being performed, and provide visual feedback to the surgeon in real time. The results showed that the platform achieved AUCs of 0.996, 0.972, 0.997, and 0.880 for capsulorhexis, phacoemulsification, cortex removal, and idle-phase recognition, respectively, with a dice score of 90.23% for pupil segmentation and a mean processing speed of 97 frames per second. A usability survey suggested that most surgeons would be willing to perform PCS for complex cataracts with this platform and thought it was accurate and helpful. Keratitis Keratitis is a major global cause of corneal blindness, often affecting marginalized populations. The burden of corneal blindness on patients and the wider community can be huge, particularly as it tends to occur in people at a younger age than other blinding eye diseases such as AMD and cataracts. Keratitis can get worse quickly with time, which may lead to permanent visual impairment and even corneal perforation. Early detection and timely management of keratitis can halt the disease progression, resulting in a favorable prognosis. Li et al. found that AI had high accuracy in screening for keratitis and other corneal abnormalities from slit-lamp images. In terms of the deep-learning algorithms, they used Inception-v3, DenseNet121, and ResNet50, with DensNet121 performing best. To be specific, the optimal algorithm DenseNet121 reached AUCs of 0.988–0.997, 0.982–0.990, and 0.988–0.998 for the classification of keratitis, other corneal abnormalities (e.g., corneal dystrophies, corneal degeneration, corneal tumors), and normal cornea, respectively, in three external test datasets. Interestingly, their system also performed well on cornea images captured by smartphone under the super-macro mode, with an AUC of 0.967, a sensitivity of 91.9%, and a specificity of 96.9% in keratitis detection. This smartphone-based approach will be extremely cost-effective and convenient for proactive keratitis screening by high-risk people (e.g., farmers and contact lens wearers) if it can be applied to clinical practice. To give prompt and precise treatment to patients with infectious keratitis, Xu et al. proposed a sequential-level deep-learning system that could effectively discriminate among bacterial keratitis, fungal keratitis, herpes simplex virus stromal keratitis, and other corneal diseases (e.g., phlyctenular keratoconjunctivitis, acanthamoeba keratitis, corneal papilloma), with an overall diagnostic accuracy of 80%, outperforming the mean diagnostic accuracy (49.27%) achieved by 421 ophthalmologists. The strength of this system was that it could extract the detailed patterns of the cornea region and assign local features to an ordered set to conform to the spatial structure and thereby learn the global features of the corneal image to perform diagnosis, which achieved better performance than conventional CNNs. Major AI applications in keratitis diagnosis are described in . Keratoconus Keratoconus is a progressive corneal ectasia with central or paracentral stroma thinning and corneal protrusion, resulting in irreversible visual impairment due to irregular corneal astigmatism or the loss of corneal transparency. Early identification of keratoconus, especially subclinical keratoconus, and subsequent treatment (e.g., corneal crosslinking and intrastromal corneal ring segments) are crucial to stabilize the disease and improve the visual prognosis. Advanced keratoconus can be detected by classic clinical signs (e.g., Vogt’s striae, Munson’s sign, Fleischer ring) through slit-lamp examination or by corneal topographical characteristics such as increased corneal refractive power, steeper radial axis tilt, and inferior-superior (I-S) corneal refractive asymmetry from corneal topographical maps. However, the detection of subclinical keratoconus remains challenging. AI may accurately diagnose subclinical keratoconus and keratoconus and predict their progress trends . Luna et al. reported machine-learning techniques, decision tree, and random forest for the diagnosis of subclinical keratoconus based on Pentacam topographic and Corvis biomechanical metrics, such as the flattest keratometry curvature, steepest keratometry curvature, stiffness parameter at the first flattening, and corneal biomechanical index. The optimal model achieved an accuracy of 89% with a sensitivity of 93% and a specificity of 86%. Meanwhile, they found that the stiffness parameter at the first flattening was the most important determinant in identifying subclinical keratoconus. Timemy et al. introduced a hybrid deep-learning construct for the detection of keratoconus. This model was developed using corneal topographic maps from 204 normal eyes, 215 keratoconus eyes, and 123 subclinical keratoconus eyes and was tested in an independent dataset including 50 normal eyes, 50 keratoconus eyes, and 50 subclinical keratoconus eyes. The proposed model reached an accuracy of 98.8% with an AUC of 0.99 and F1 score of 0.99 for the two-class task (normal vs. keratoconus) and an accuracy of 81.5% with an AUC of 0.93 and F1 score of 0.81 for the three-class task (normal vs. keratoconus vs. subclinical keratoconus). Early and accurate prediction of progress trends in keratoconus is critical for the prudent and cost-effective use of corneal crosslinking and the determination of timing of follow-up visits. García et al. reported a time-delay neural network to predict keratoconus progression using two prior tomography measurements from Pentacam. This network received six characteristics as input (e.g., average keratometry, the steepest radius of the front surface, and the average radius of the back surface), evaluated in two consecutive examinations, forecasted the future values, and obtained the result (stable or suspect progressive) leveraging the significance of the variation from the baseline. The average positive and negative predictive values of the network were 71.4% and 80.2%, indicating it had the potential to assist clinicians to make a personalized management plan for patients with keratoconus. Other anterior-segment diseases A large number of studies have also proved the possibility of using AI to detect other anterior-segment diseases. For example, Chase et al. demonstrated that the deep-learning system, developed by a VGG19 network based on 27,180 anterior-segment OCT images, was able to identify dry eye diseases, with 84.62% accuracy, 86.36% sensitivity, and 82.35% specificity. The performance of this system was significantly better than some clinical dry eye tests, such as Schirmer’s test and corneal staining, and was comparable with that of tear break-up time and Ocular Surface Disease Index. In addition, Zhang et al. developed a deep-learning system for detecting obstructive meibomian gland dysfunction (MGD) and atrophic MGD using 4,985 in vivo laser confocal microscope images and validated the system on 1,663 images. The accuracy, sensitivity, and specificity of the system for obstructive MGD were 97.3%, 88.8%, and 95.4%, respectively; for atrophic MGD, 98.6%, 89.4%, and 98.4%, respectively; and for healthy controls, 98.0%, 94.5%, and 92.6%, respectively. Moreover, Li et al. introduced an AI system based on Faster R-CNN and DenseNet121 to detect malignant eyelid tumors from photographic images captured by ordinary digital cameras. In an external test set, the average precision score of the system was 0.762 for locating eyelid tumors and the AUC was 0.899 for discerning malignant eyelid tumors. In the past 20 years, although the prevalence of cataracts has been decreasing due to the increasing rates of cataract surgery because of improved techniques and active surgical initiatives, it still affects 95 million people worldwide. Cataract remains the leading cause of blindness (accounting for 50% of blindness), especially in low-income and middle-income countries. Therefore, exploring a set of strategies to promote cataract screening and related ophthalmic services is imperative. Recent advancements in AI may help achieve this goal, such as diagnosis and quantitative classification of age-related cataract from slit-lamp images . Keenan et al. trained deep-learning models, named DeepLensNet, to detect and quantify nuclear sclerosis (NS) from 45° slit-lamp images and cortical lens opacity (CLO) and posterior subcapsular cataract (PSC) from retroillumination images. NS grading was considered on 0.9–7.1 scale. CLO and PSC grading were both considered as percentages. In the full test set, mean squared error values for DeepLensNet were 0.23 for NS, 13.1 for CLO, and 16.6 for PSC. The results indicate that this framework can perform automated and quantitative classification of cataract severity with high accuracy, which has the potential to increase the accessibility of cataract evaluation globally. Except for slit-lamp images, Tham et al. found that fundus images could also be used to develop an AI system for cataract screening. Based on 25,742 fundus images, they constructed a framework with ResNet50 and XGBoost classifier for the automated detection of visually significant cataracts (BCVA < 20/60), achieving AUCs of 0.916–0.965 in three external test sets. One merit of this system is that it can screen for cataracts with a single imaging modality, which is different from the traditional method that requires slit-lamp and retroillumination images alongside BCVA measurement. The other merit is that this system can be readily integrated into existing fundus-image-based AI systems, allowing simultaneous screening for other posterior-segment diseases. Other than cataract screening, AI can also offer real-time guidance for phacoemulsification cataract surgery (PCS). Nespolo et al. invented a computer vision-based platform using a region-based CNN (Faster R-CNN) built on ResNet50, a k-means clustering technique, and an optical-flow-tracking technology to enhance the surgeon experience during the PCS. Specifically, this platform can be used to receive frames from the video source, locate the pupil, discern the surgical phase being performed, and provide visual feedback to the surgeon in real time. The results showed that the platform achieved AUCs of 0.996, 0.972, 0.997, and 0.880 for capsulorhexis, phacoemulsification, cortex removal, and idle-phase recognition, respectively, with a dice score of 90.23% for pupil segmentation and a mean processing speed of 97 frames per second. A usability survey suggested that most surgeons would be willing to perform PCS for complex cataracts with this platform and thought it was accurate and helpful. Keratitis is a major global cause of corneal blindness, often affecting marginalized populations. The burden of corneal blindness on patients and the wider community can be huge, particularly as it tends to occur in people at a younger age than other blinding eye diseases such as AMD and cataracts. Keratitis can get worse quickly with time, which may lead to permanent visual impairment and even corneal perforation. Early detection and timely management of keratitis can halt the disease progression, resulting in a favorable prognosis. Li et al. found that AI had high accuracy in screening for keratitis and other corneal abnormalities from slit-lamp images. In terms of the deep-learning algorithms, they used Inception-v3, DenseNet121, and ResNet50, with DensNet121 performing best. To be specific, the optimal algorithm DenseNet121 reached AUCs of 0.988–0.997, 0.982–0.990, and 0.988–0.998 for the classification of keratitis, other corneal abnormalities (e.g., corneal dystrophies, corneal degeneration, corneal tumors), and normal cornea, respectively, in three external test datasets. Interestingly, their system also performed well on cornea images captured by smartphone under the super-macro mode, with an AUC of 0.967, a sensitivity of 91.9%, and a specificity of 96.9% in keratitis detection. This smartphone-based approach will be extremely cost-effective and convenient for proactive keratitis screening by high-risk people (e.g., farmers and contact lens wearers) if it can be applied to clinical practice. To give prompt and precise treatment to patients with infectious keratitis, Xu et al. proposed a sequential-level deep-learning system that could effectively discriminate among bacterial keratitis, fungal keratitis, herpes simplex virus stromal keratitis, and other corneal diseases (e.g., phlyctenular keratoconjunctivitis, acanthamoeba keratitis, corneal papilloma), with an overall diagnostic accuracy of 80%, outperforming the mean diagnostic accuracy (49.27%) achieved by 421 ophthalmologists. The strength of this system was that it could extract the detailed patterns of the cornea region and assign local features to an ordered set to conform to the spatial structure and thereby learn the global features of the corneal image to perform diagnosis, which achieved better performance than conventional CNNs. Major AI applications in keratitis diagnosis are described in . Keratoconus is a progressive corneal ectasia with central or paracentral stroma thinning and corneal protrusion, resulting in irreversible visual impairment due to irregular corneal astigmatism or the loss of corneal transparency. Early identification of keratoconus, especially subclinical keratoconus, and subsequent treatment (e.g., corneal crosslinking and intrastromal corneal ring segments) are crucial to stabilize the disease and improve the visual prognosis. Advanced keratoconus can be detected by classic clinical signs (e.g., Vogt’s striae, Munson’s sign, Fleischer ring) through slit-lamp examination or by corneal topographical characteristics such as increased corneal refractive power, steeper radial axis tilt, and inferior-superior (I-S) corneal refractive asymmetry from corneal topographical maps. However, the detection of subclinical keratoconus remains challenging. AI may accurately diagnose subclinical keratoconus and keratoconus and predict their progress trends . Luna et al. reported machine-learning techniques, decision tree, and random forest for the diagnosis of subclinical keratoconus based on Pentacam topographic and Corvis biomechanical metrics, such as the flattest keratometry curvature, steepest keratometry curvature, stiffness parameter at the first flattening, and corneal biomechanical index. The optimal model achieved an accuracy of 89% with a sensitivity of 93% and a specificity of 86%. Meanwhile, they found that the stiffness parameter at the first flattening was the most important determinant in identifying subclinical keratoconus. Timemy et al. introduced a hybrid deep-learning construct for the detection of keratoconus. This model was developed using corneal topographic maps from 204 normal eyes, 215 keratoconus eyes, and 123 subclinical keratoconus eyes and was tested in an independent dataset including 50 normal eyes, 50 keratoconus eyes, and 50 subclinical keratoconus eyes. The proposed model reached an accuracy of 98.8% with an AUC of 0.99 and F1 score of 0.99 for the two-class task (normal vs. keratoconus) and an accuracy of 81.5% with an AUC of 0.93 and F1 score of 0.81 for the three-class task (normal vs. keratoconus vs. subclinical keratoconus). Early and accurate prediction of progress trends in keratoconus is critical for the prudent and cost-effective use of corneal crosslinking and the determination of timing of follow-up visits. García et al. reported a time-delay neural network to predict keratoconus progression using two prior tomography measurements from Pentacam. This network received six characteristics as input (e.g., average keratometry, the steepest radius of the front surface, and the average radius of the back surface), evaluated in two consecutive examinations, forecasted the future values, and obtained the result (stable or suspect progressive) leveraging the significance of the variation from the baseline. The average positive and negative predictive values of the network were 71.4% and 80.2%, indicating it had the potential to assist clinicians to make a personalized management plan for patients with keratoconus. A large number of studies have also proved the possibility of using AI to detect other anterior-segment diseases. For example, Chase et al. demonstrated that the deep-learning system, developed by a VGG19 network based on 27,180 anterior-segment OCT images, was able to identify dry eye diseases, with 84.62% accuracy, 86.36% sensitivity, and 82.35% specificity. The performance of this system was significantly better than some clinical dry eye tests, such as Schirmer’s test and corneal staining, and was comparable with that of tear break-up time and Ocular Surface Disease Index. In addition, Zhang et al. developed a deep-learning system for detecting obstructive meibomian gland dysfunction (MGD) and atrophic MGD using 4,985 in vivo laser confocal microscope images and validated the system on 1,663 images. The accuracy, sensitivity, and specificity of the system for obstructive MGD were 97.3%, 88.8%, and 95.4%, respectively; for atrophic MGD, 98.6%, 89.4%, and 98.4%, respectively; and for healthy controls, 98.0%, 94.5%, and 92.6%, respectively. Moreover, Li et al. introduced an AI system based on Faster R-CNN and DenseNet121 to detect malignant eyelid tumors from photographic images captured by ordinary digital cameras. In an external test set, the average precision score of the system was 0.762 for locating eyelid tumors and the AUC was 0.899 for discerning malignant eyelid tumors. AI has the potential to detect hidden information that clinicians are normally unable to perceive from digital health data. In ophthalmology, with the continuous advancement of AI technologies, the application of AI based on retinal images has extended from the detection of multiple fundus diseases to the screening for systemic diseases. These breakthroughs can be attributed to the following three reasons: (1) the unique anatomy of the eye offers an accessible “window” for the in vivo visualization of microvasculature and cerebral neurons; (2) the retina manifestations can be signs of many systemic diseases, such as diabetes and heart disease; (3) retinal changes can be recorded through non-invasive digital fundus imaging, which is low cost and widely available in different levels of medical institutions. Cardiovascular disease Cardiovascular disease (CVD) is a leading cause of death globally, taking an estimated 17.9 million lives annually. Overt retinal vascular damage (such as retinal hemorrhages) and subtle changes (such as retinal arteriolar narrowing) are markers of CVD. To improve present risk-stratification approaches for CVD events, Rim et al. developed and validated a deep-learning-based cardiovascular risk-stratification system using 216,152 retinal images from five datasets from Singapore, South Korea, and the United Kingdom. This system achieved an AUC of 0.742 in predicting the presence of coronary artery calcium (a preclinical marker of atherosclerosis and strongly associated with the risk of CVD). Poplin et al. reported that deep-learning models trained on data from 284,355 patients could extract new information from retinal images to predict cardiovascular risk factors, such as age (mean absolute error [MAE] within 3.26 years), gender (AUC = 0.97), systolic blood pressure (MAE within 11.23 mm Hg), smoking status (AUC = 0.71), and major adverse cardiac events (AUC = 0.70). Meanwhile, they demonstrated that the deep-learning models generated each prediction using anatomical features, such as the retinal vessels or the optic disc. Chronic kidney disease and type 2 diabetes Chronic kidney disease (CKD) is a progressive disease with high morbidity and mortality that occurs in the general adult population, particularly in people with diabetes and hypertension. Type 2 diabetes is another common chronic disease that accounts for nearly 90% of the 537 million cases of diabetes worldwide. Early diagnosis and proactive management of CKD and diabetes are critical in reducing microvascular and macrovascular complications and mortality burden. As CKD and diabetes have manifestations in the retina, retinal images can be used to detect and monitor these diseases. Zhang et al. reported that deep-learning models developed based on 115,344 retinal images from 56,672 patients were able to detect CKD and type 2 diabetes solely from retinal images or in combination with clinical metadata (e.g., age, sex, body mass index, and blood pressure) with AUCs of 0.85–0.93. The models can also be utilized to predict estimated glomerular filtration rates and blood-glucose levels, with MAEs of 11.1–13.4 mL min −1 per 1.73 m 2 and 0.65–1.1 mmol L −1 , respectively. Sabanayagam et al. established a deep-learning algorithm using 12,790 retinal images to screen for CKD. In this study, the model trained solely by retinal images achieved AUCs of 0.733–0.911 in validation and testing datasets, indicating the feasibility of employing retinal photography as an adjunctive screening tool for CKD in community and primary care settings. Alzheimer’s disease Alzheimer’s disease (AD), a progressive neurodegenerative disease, is the most common type of dementia in the elderly worldwide and is becoming one of the most lethal, expensive, and burdening diseases of this century. Diagnosis of AD is complex and normally involves expensive and sometimes invasive tests (such as amyloid positron emission tomography [PET] imaging and cerebrospinal fluid assays), which are not usually available outside of highly specialized clinical institutions. The retina is an extension of the central nervous system and offers a distinctively accessible insight into brain pathology. Research has found potentially measurable structural, vascular, and metabolic changes in the retina at the early stages of AD. Therefore, using noninvasive and low-cost retinal photography to detect AD is feasible. Cheung et al. demonstrated that a deep-learning model had the capability to identify AD from retinal images alone. They trained, validated, and tested the model using 12,949 retinal images from 648 AD patients and 3,240 individuals without the disease. The model had accuracies ranging from 79.6% to 92.1% and AUCs ranging from 0.73 to 0.91 for detecting AD in testing datasets. In the datasets with PET information, the model can also distinguish between participants who were β-amyloid positive and those who were β-amyloid negative, with accuracies ranging from 80.6% to 89.3% and AUCs ranging from 0.68 to 0.86. This study showed that a retinal-image-based deep-learning algorithm had high accuracy in detecting AD and this approach could be used to screen for AD in a community setting. Challenges in the AI clinical translation Although AI systems have shown great performance in a wide variety of retrospective studies, relatively few of them have been translated into clinical practice. Many challenges, such as the generalizability of AI systems, still exist and stand in the path of true clinical adoption of AI tools . In this section, we highlight some critical challenges and the research that has already been conducted to tackle these issues. Cardiovascular disease (CVD) is a leading cause of death globally, taking an estimated 17.9 million lives annually. Overt retinal vascular damage (such as retinal hemorrhages) and subtle changes (such as retinal arteriolar narrowing) are markers of CVD. To improve present risk-stratification approaches for CVD events, Rim et al. developed and validated a deep-learning-based cardiovascular risk-stratification system using 216,152 retinal images from five datasets from Singapore, South Korea, and the United Kingdom. This system achieved an AUC of 0.742 in predicting the presence of coronary artery calcium (a preclinical marker of atherosclerosis and strongly associated with the risk of CVD). Poplin et al. reported that deep-learning models trained on data from 284,355 patients could extract new information from retinal images to predict cardiovascular risk factors, such as age (mean absolute error [MAE] within 3.26 years), gender (AUC = 0.97), systolic blood pressure (MAE within 11.23 mm Hg), smoking status (AUC = 0.71), and major adverse cardiac events (AUC = 0.70). Meanwhile, they demonstrated that the deep-learning models generated each prediction using anatomical features, such as the retinal vessels or the optic disc. Chronic kidney disease (CKD) is a progressive disease with high morbidity and mortality that occurs in the general adult population, particularly in people with diabetes and hypertension. Type 2 diabetes is another common chronic disease that accounts for nearly 90% of the 537 million cases of diabetes worldwide. Early diagnosis and proactive management of CKD and diabetes are critical in reducing microvascular and macrovascular complications and mortality burden. As CKD and diabetes have manifestations in the retina, retinal images can be used to detect and monitor these diseases. Zhang et al. reported that deep-learning models developed based on 115,344 retinal images from 56,672 patients were able to detect CKD and type 2 diabetes solely from retinal images or in combination with clinical metadata (e.g., age, sex, body mass index, and blood pressure) with AUCs of 0.85–0.93. The models can also be utilized to predict estimated glomerular filtration rates and blood-glucose levels, with MAEs of 11.1–13.4 mL min −1 per 1.73 m 2 and 0.65–1.1 mmol L −1 , respectively. Sabanayagam et al. established a deep-learning algorithm using 12,790 retinal images to screen for CKD. In this study, the model trained solely by retinal images achieved AUCs of 0.733–0.911 in validation and testing datasets, indicating the feasibility of employing retinal photography as an adjunctive screening tool for CKD in community and primary care settings. Alzheimer’s disease (AD), a progressive neurodegenerative disease, is the most common type of dementia in the elderly worldwide and is becoming one of the most lethal, expensive, and burdening diseases of this century. Diagnosis of AD is complex and normally involves expensive and sometimes invasive tests (such as amyloid positron emission tomography [PET] imaging and cerebrospinal fluid assays), which are not usually available outside of highly specialized clinical institutions. The retina is an extension of the central nervous system and offers a distinctively accessible insight into brain pathology. Research has found potentially measurable structural, vascular, and metabolic changes in the retina at the early stages of AD. Therefore, using noninvasive and low-cost retinal photography to detect AD is feasible. Cheung et al. demonstrated that a deep-learning model had the capability to identify AD from retinal images alone. They trained, validated, and tested the model using 12,949 retinal images from 648 AD patients and 3,240 individuals without the disease. The model had accuracies ranging from 79.6% to 92.1% and AUCs ranging from 0.73 to 0.91 for detecting AD in testing datasets. In the datasets with PET information, the model can also distinguish between participants who were β-amyloid positive and those who were β-amyloid negative, with accuracies ranging from 80.6% to 89.3% and AUCs ranging from 0.68 to 0.86. This study showed that a retinal-image-based deep-learning algorithm had high accuracy in detecting AD and this approach could be used to screen for AD in a community setting. Although AI systems have shown great performance in a wide variety of retrospective studies, relatively few of them have been translated into clinical practice. Many challenges, such as the generalizability of AI systems, still exist and stand in the path of true clinical adoption of AI tools . In this section, we highlight some critical challenges and the research that has already been conducted to tackle these issues. Data issues in developing robust AI systems Data sharing Large datasets are required to facilitate the development of a robust AI system. The lack of high-quality public datasets that are truly representative of real-world clinical practice stands in the path of clinical translation of AI systems. Data sharing might be a good solution but it generates ethical and legal challenges in general. Even if data are obtained in an anonymized manner, it can potentially put patient privacy at risk. Protecting patient privacy and acquiring approval for data use are important rules to comply with. Unfortunately, these rules may hinder data sharing among different medical research groups, not to mention making the data publicly available. The adoption of federated learning is a good alternative to training AI models with diverse data from multiple clinical institutions without the centralization of data. This strategy can address the issue that data reside in different institutions, removing barriers to data sharing and circumventing the problem of patient privacy . Meanwhile, federated learning can facilitate rapid data science collaboration and thus improve the robustness of AI systems. In addition, controlled-access data sharing, an approach that requires a request for access to datasets to be approved, is another alternative solution for researchers to acquire data to solve relevant research issues while protecting participant privacy. Data annotation Accurate clinical data annotations are crucial for the development of reliable AI systems, , as annotation inconsistencies may lead to unpredictable clinical consequences, such as erroneous classifications. Several approaches have been leveraged to resolve disagreements between graders and obtain ground truth. One method consists of adopting the majority decision from a panel of three or more professional graders. Another consists of recruiting two or more professional graders to label data independently and then employing another senior grader to arbitrate disagreements, with the senior grader’s decision used as the ground truth. Third, some data annotations can be conducted using the recognized gold standard. For example, Li et al. annotated images of benign and malignant eyelid tumors based on unequivocal histopathological diagnoses. Normally, annotations can be divided into two categories: the annotation of interest regions in images (e.g., retinal hemorrhages, exudates, and drusen) and clinical annotations (e.g., disease classification, treatment response, vision prognosis). Conducting manual annotations for large-scale datasets before the model training is a considerably time-consuming and labor-intensive task that needs a lot of professional graders or ophthalmologists, hindering the construction of robust AI systems. , , , Therefore, exploring techniques to promote the efficient production of annotations is important, although manual annotations are still necessary. Training models using weakly supervised learning may be a good approach to reduce the workload of manual annotations. , For image segmentation, weak supervision requires sparse manual annotations of small interest regions using dots via experts, whereas full supervision needs dense annotations, in which all pixels of images are manually labeled. , Playout et al. have reported that weak supervision in combination with advanced learning in model training can achieve performance comparable with fully supervised models for retinal lesion segmentation in fundus images. Standardization of clinical data collection In the past two decades, health systems have heavily invested in the digitalization of every aspect of their operation. This transformation has resulted in unprecedented growth in the volume of medical electronic data, facilitating the development of AI-based medical devices. Although the size of datasets has increased, data collection is not done in a standardized manner, affecting the ready utilization of these data for AI model training and testing. This issue leads to a growing number of multicentric efforts to deal with the large variability in examination items, the timing of laboratory tests, image quality, etc. To improve the usability of data, standardization of clinical data collection should be implemented to generate high-quality data with complete and consistent information to support the development of robust medical AI products. , For example, medical text data collection should include basic information such as age, gender, and examination date, and health examination records should have all examination items and complete results. Besides, as low-quality image data often result in a loss of diagnostic information and affect AI-based image analyses, image quality assessment is necessary at the stage of data allocation to filter out low-quality images, which could improve the performance of AI-based diagnostic models in real-world settings. To address this issue, Liu et al. developed a deep-learning-based flow-cytometry-like image quality classifier for the automated, high-throughput, and multidimensional classification of fundus image quality, which can detect low-quality images in real time and then guide a photographer to acquire high-quality images immediately. Li et al. reported a deep-learning-based image-quality-control system that could discern low-quality slit-lamp images. This system can be used as a prescreening tool to filter out low-quality images and ensure that only high-quality images will be transferred to the subsequent AI-based diagnostic systems. Shen et al. established a new multi-task domain adaptation framework for the automated fundus image quality assessment. The proposed framework can offer interpretable quality assessment with both quantitative scores and quality visualization, which outperforms different state-of-the-art methods. Dai et al. demonstrated that the AI-based image quality assessment could reduce the proportion of poor-quality images and significantly improve the accuracy of an AI model for DR diagnosis. Real-world performance of AI systems Recently, several reports showed that AI systems in practice were less helpful than retrospective studies described. , For example, Kanagasingam et al. evaluated their deep-learning-based system for DR screening based on retinal images in a real-world primary care clinic. They found that the system had a high false-positive rate (7.9%) and a low positive predictive value (12%). The possible reason for this unsatisfactory performance of the system is that the incidence of DR in the primary care clinic was 1%, whereas their AI system was developed using retrospective data in which the incidence of DR was much higher (33.3%). Long et al. developed an AI platform that had 98.25% accuracy in childhood cataract diagnosis and 92.86% accuracy in treatment suggestions in external datasets retrospectively collected from three hospitals. However, when they applied the platform to unselected real-world datasets prospectively obtained from five hospitals, accuracies decreased to 87.4% in cataract diagnosis and 70.8% in treatment determination. The possible explanation for this phenomenon is that the retrospective datasets often undergo extensive filtering and cleaning, which makes them less representative of real-world clinical practice. Randomized controlled trials (RCTs) and prospective research can bridge such gaps between theory and practice, showing the true performance of AI systems in real healthcare settings and demonstrating how useful the systems are for the clinic. Data sharing Large datasets are required to facilitate the development of a robust AI system. The lack of high-quality public datasets that are truly representative of real-world clinical practice stands in the path of clinical translation of AI systems. Data sharing might be a good solution but it generates ethical and legal challenges in general. Even if data are obtained in an anonymized manner, it can potentially put patient privacy at risk. Protecting patient privacy and acquiring approval for data use are important rules to comply with. Unfortunately, these rules may hinder data sharing among different medical research groups, not to mention making the data publicly available. The adoption of federated learning is a good alternative to training AI models with diverse data from multiple clinical institutions without the centralization of data. This strategy can address the issue that data reside in different institutions, removing barriers to data sharing and circumventing the problem of patient privacy . Meanwhile, federated learning can facilitate rapid data science collaboration and thus improve the robustness of AI systems. In addition, controlled-access data sharing, an approach that requires a request for access to datasets to be approved, is another alternative solution for researchers to acquire data to solve relevant research issues while protecting participant privacy. Data annotation Accurate clinical data annotations are crucial for the development of reliable AI systems, , as annotation inconsistencies may lead to unpredictable clinical consequences, such as erroneous classifications. Several approaches have been leveraged to resolve disagreements between graders and obtain ground truth. One method consists of adopting the majority decision from a panel of three or more professional graders. Another consists of recruiting two or more professional graders to label data independently and then employing another senior grader to arbitrate disagreements, with the senior grader’s decision used as the ground truth. Third, some data annotations can be conducted using the recognized gold standard. For example, Li et al. annotated images of benign and malignant eyelid tumors based on unequivocal histopathological diagnoses. Normally, annotations can be divided into two categories: the annotation of interest regions in images (e.g., retinal hemorrhages, exudates, and drusen) and clinical annotations (e.g., disease classification, treatment response, vision prognosis). Conducting manual annotations for large-scale datasets before the model training is a considerably time-consuming and labor-intensive task that needs a lot of professional graders or ophthalmologists, hindering the construction of robust AI systems. , , , Therefore, exploring techniques to promote the efficient production of annotations is important, although manual annotations are still necessary. Training models using weakly supervised learning may be a good approach to reduce the workload of manual annotations. , For image segmentation, weak supervision requires sparse manual annotations of small interest regions using dots via experts, whereas full supervision needs dense annotations, in which all pixels of images are manually labeled. , Playout et al. have reported that weak supervision in combination with advanced learning in model training can achieve performance comparable with fully supervised models for retinal lesion segmentation in fundus images. Standardization of clinical data collection In the past two decades, health systems have heavily invested in the digitalization of every aspect of their operation. This transformation has resulted in unprecedented growth in the volume of medical electronic data, facilitating the development of AI-based medical devices. Although the size of datasets has increased, data collection is not done in a standardized manner, affecting the ready utilization of these data for AI model training and testing. This issue leads to a growing number of multicentric efforts to deal with the large variability in examination items, the timing of laboratory tests, image quality, etc. To improve the usability of data, standardization of clinical data collection should be implemented to generate high-quality data with complete and consistent information to support the development of robust medical AI products. , For example, medical text data collection should include basic information such as age, gender, and examination date, and health examination records should have all examination items and complete results. Besides, as low-quality image data often result in a loss of diagnostic information and affect AI-based image analyses, image quality assessment is necessary at the stage of data allocation to filter out low-quality images, which could improve the performance of AI-based diagnostic models in real-world settings. To address this issue, Liu et al. developed a deep-learning-based flow-cytometry-like image quality classifier for the automated, high-throughput, and multidimensional classification of fundus image quality, which can detect low-quality images in real time and then guide a photographer to acquire high-quality images immediately. Li et al. reported a deep-learning-based image-quality-control system that could discern low-quality slit-lamp images. This system can be used as a prescreening tool to filter out low-quality images and ensure that only high-quality images will be transferred to the subsequent AI-based diagnostic systems. Shen et al. established a new multi-task domain adaptation framework for the automated fundus image quality assessment. The proposed framework can offer interpretable quality assessment with both quantitative scores and quality visualization, which outperforms different state-of-the-art methods. Dai et al. demonstrated that the AI-based image quality assessment could reduce the proportion of poor-quality images and significantly improve the accuracy of an AI model for DR diagnosis. Large datasets are required to facilitate the development of a robust AI system. The lack of high-quality public datasets that are truly representative of real-world clinical practice stands in the path of clinical translation of AI systems. Data sharing might be a good solution but it generates ethical and legal challenges in general. Even if data are obtained in an anonymized manner, it can potentially put patient privacy at risk. Protecting patient privacy and acquiring approval for data use are important rules to comply with. Unfortunately, these rules may hinder data sharing among different medical research groups, not to mention making the data publicly available. The adoption of federated learning is a good alternative to training AI models with diverse data from multiple clinical institutions without the centralization of data. This strategy can address the issue that data reside in different institutions, removing barriers to data sharing and circumventing the problem of patient privacy . Meanwhile, federated learning can facilitate rapid data science collaboration and thus improve the robustness of AI systems. In addition, controlled-access data sharing, an approach that requires a request for access to datasets to be approved, is another alternative solution for researchers to acquire data to solve relevant research issues while protecting participant privacy. Accurate clinical data annotations are crucial for the development of reliable AI systems, , as annotation inconsistencies may lead to unpredictable clinical consequences, such as erroneous classifications. Several approaches have been leveraged to resolve disagreements between graders and obtain ground truth. One method consists of adopting the majority decision from a panel of three or more professional graders. Another consists of recruiting two or more professional graders to label data independently and then employing another senior grader to arbitrate disagreements, with the senior grader’s decision used as the ground truth. Third, some data annotations can be conducted using the recognized gold standard. For example, Li et al. annotated images of benign and malignant eyelid tumors based on unequivocal histopathological diagnoses. Normally, annotations can be divided into two categories: the annotation of interest regions in images (e.g., retinal hemorrhages, exudates, and drusen) and clinical annotations (e.g., disease classification, treatment response, vision prognosis). Conducting manual annotations for large-scale datasets before the model training is a considerably time-consuming and labor-intensive task that needs a lot of professional graders or ophthalmologists, hindering the construction of robust AI systems. , , , Therefore, exploring techniques to promote the efficient production of annotations is important, although manual annotations are still necessary. Training models using weakly supervised learning may be a good approach to reduce the workload of manual annotations. , For image segmentation, weak supervision requires sparse manual annotations of small interest regions using dots via experts, whereas full supervision needs dense annotations, in which all pixels of images are manually labeled. , Playout et al. have reported that weak supervision in combination with advanced learning in model training can achieve performance comparable with fully supervised models for retinal lesion segmentation in fundus images. In the past two decades, health systems have heavily invested in the digitalization of every aspect of their operation. This transformation has resulted in unprecedented growth in the volume of medical electronic data, facilitating the development of AI-based medical devices. Although the size of datasets has increased, data collection is not done in a standardized manner, affecting the ready utilization of these data for AI model training and testing. This issue leads to a growing number of multicentric efforts to deal with the large variability in examination items, the timing of laboratory tests, image quality, etc. To improve the usability of data, standardization of clinical data collection should be implemented to generate high-quality data with complete and consistent information to support the development of robust medical AI products. , For example, medical text data collection should include basic information such as age, gender, and examination date, and health examination records should have all examination items and complete results. Besides, as low-quality image data often result in a loss of diagnostic information and affect AI-based image analyses, image quality assessment is necessary at the stage of data allocation to filter out low-quality images, which could improve the performance of AI-based diagnostic models in real-world settings. To address this issue, Liu et al. developed a deep-learning-based flow-cytometry-like image quality classifier for the automated, high-throughput, and multidimensional classification of fundus image quality, which can detect low-quality images in real time and then guide a photographer to acquire high-quality images immediately. Li et al. reported a deep-learning-based image-quality-control system that could discern low-quality slit-lamp images. This system can be used as a prescreening tool to filter out low-quality images and ensure that only high-quality images will be transferred to the subsequent AI-based diagnostic systems. Shen et al. established a new multi-task domain adaptation framework for the automated fundus image quality assessment. The proposed framework can offer interpretable quality assessment with both quantitative scores and quality visualization, which outperforms different state-of-the-art methods. Dai et al. demonstrated that the AI-based image quality assessment could reduce the proportion of poor-quality images and significantly improve the accuracy of an AI model for DR diagnosis. Recently, several reports showed that AI systems in practice were less helpful than retrospective studies described. , For example, Kanagasingam et al. evaluated their deep-learning-based system for DR screening based on retinal images in a real-world primary care clinic. They found that the system had a high false-positive rate (7.9%) and a low positive predictive value (12%). The possible reason for this unsatisfactory performance of the system is that the incidence of DR in the primary care clinic was 1%, whereas their AI system was developed using retrospective data in which the incidence of DR was much higher (33.3%). Long et al. developed an AI platform that had 98.25% accuracy in childhood cataract diagnosis and 92.86% accuracy in treatment suggestions in external datasets retrospectively collected from three hospitals. However, when they applied the platform to unselected real-world datasets prospectively obtained from five hospitals, accuracies decreased to 87.4% in cataract diagnosis and 70.8% in treatment determination. The possible explanation for this phenomenon is that the retrospective datasets often undergo extensive filtering and cleaning, which makes them less representative of real-world clinical practice. Randomized controlled trials (RCTs) and prospective research can bridge such gaps between theory and practice, showing the true performance of AI systems in real healthcare settings and demonstrating how useful the systems are for the clinic. Although numerous studies reported that their AI systems showed robust performance in detecting eye diseases and had the potential to be applied in clinics, most AI-based medical devices had not yet been authorized for market distribution for clinical management of diseases such as AMD, glaucoma, and cataracts. One of the most important reasons for this is that the generalizability of AI systems to populations of different ethnicities and different countries, different clinical application scenarios, and images captured using different types of cameras remains uncertain. Lots of AI studies only evaluated their systems in data from a single source, hence the systems often performed poorly in real-world datasets that had more sources of variation than the datasets utilized in research papers. To improve the generalizability of AI systems, first, we need to build large, multicenter, and multiethnic datasets for system development and evaluation. Milea et al. developed a deep-learning system for papilledema detection using data collected from 19 sites in 11 countries and evaluated the system in data obtained from five other sites in five countries. The AUCs of their system in internal and external test datasets were 0.99 and 0.96, verifying that the system had broad generalizability. In addition, transfer learning, a technique that aims to transfer knowledge from one task to a different but related task, can help decrease generalization errors of AI systems via reusing the weights of a pretrained model, particularly when faced with tasks with limited data. Kermany et al. demonstrated that AI systems trained with a transfer-learning algorithm had good performance and generalizability in the diagnosis of common diseases from different types of images, such as detecting diabetic macular edema from OCT images (accuracy = 98.2%) and pediatric pneumonia from chest X-ray images (accuracy = 92.8%). Third, the generalizability of AI networks can be improved by utilizing a data-augmentation (DA) strategy that creates more training samples for increasing the diversity of the training data. Zhou et al. proposed an approach named DA-based feature alignment that could consistently and significantly improve the out-of-distribution generalizability (up to +16.3% mean of clean AUC) of AI algorithms in glaucoma detection from fundus images. Fourth, an AI algorithm trained based on lesion labels can broaden its generalizability in disease detection. Li et al. reported that the algorithm trained with the image-level classification labels and the anatomical and pathological labels displayed better performance and generalizability than that trained with only the image-level classification labels in diagnosing ophthalmic disorders from slit-lamp images (accuracies, 99.22%–79.47% versus 90.14%–47.19%). AI systems are often described as black boxes due to the nature of these systems (being trained instead of being explicitly programmed). It is difficult for clinicians to understand the precise underlying functioning of the systems. As a result, correcting some erroneous behaviors might be difficult, and acceptance by clinicians as well as regulatory approval might be hampered. Decoding AI for clinicians can mitigate such uncertainty. This challenge has provided a stimulus for research groups and industries to focus on explainable AI. Techniques that enable a good understanding of the working principle of AI systems are developed. For instance, Niu et al. reported a method that could enhance the interpretability of an AI system in detecting DR. To be specific, they first define novel pathological descriptors leveraging activated neurons of the DR detector to encode both the appearance and spatial information of lesions. Then, they proposed a novel generative adversarial network (GAN), Patho-GAN, to visualize the signs that the DR detector identified as evidence to make a prediction. Xu et al. developed an explainable AI system for diagnosing fungal keratitis from in vivo confocal microscopy images based on gradient-weighted class activation mapping (Grad-CAM) and guided Grad-CAM techniques. They found that the assistance from the explainable AI system could boost ophthalmologists’ performance beyond what was achievable by the ophthalmologist alone or with the black-box AI assistance. Overall, these interpretation frameworks may facilitate AI acceptance for clinical usage. The performance of AI systems has the potential to degrade over time as the characteristics of the world, such as disease distribution, population characteristics, health infrastructure, and cyber technologies, are changing all the time. This requires that AI systems should have the ability of lifelong continuous learning to keep and even improve their performance over time. The continuous learning technique, meta-learning, which aims to improve the AI algorithm itself, is a potential approach to address this issue. Although medical AI systems can help physicians in clinics, such as disease diagnosis, recommendations for treatment, and prognosis prediction, it is still unclear whether healthcare providers, developers, sellers, or regulators should be held accountable if an AI system makes mistakes in real-world clinical practice even after being thoroughly clinically validated. For example, AI systems may miss a retinal disease in a fundus image or recommend an incorrect treatment strategy. As a result, patients may be injured. In this case, we have to determine who is responsible for this incident. The allocation of liability makes clear not only whether and from whom patients acquire redress but also whether, potentially, AI systems will make their way into clinical practice. At present, the suggested solution is to treat medical AI systems as a confirmatory tool rather than as a source of ways to improve care. In other words, a physician should check every output from the medical AI systems to ensure the results that meet and follow the standard of care. Therefore, the physician would be held liable if malpractice occurs due to using these systems. This strategy may minimize the potential value of medical AI systems as some systems may perform better than even the best physicians but the physicians would choose to ignore the AI recommendation when it conflicts with standard practice. Consequently, the approach that can balance the safety and innovation of medical AI needs to be further explored. Generally, AI models were directly trained using existing open-sourced machine-learning packages frequently utilized by others to address the issue of interest without additional customization or refinement. This approach may limit the optimal performance of AI applications as no generalized solution exists in most cases. To improve the performance of AI tools, in-depth knowledge of clinical problems as well as the features of AI algorithms is indispensable. Therefore, applicable customization of the algorithms should be conducted according to the specific challenges of each problem, which usually needs interdisciplinary collaboration among ophthalmologists, computer scientists (e.g., AI experts), policymakers, and others. Although AI studies have seen enormous progress in the past decade, they are predominantly based on fixed datasets and stationary environments. The performance of AI systems is often fixed by the time they are developed. However, the world is not stationary, which requires that AI systems should have the ability as clinicians to improve themselves constantly and evolve to thrive in dynamic learning settings. Continual learning techniques, such as gradient-based learning, modular neural network, and meta-learning, may enable AI models to obtain specialized solutions without forgetting previous ones, namely learning over a lifetime, as a clinician does. These techniques may take AI to a higher level by improving learning efficiency and enabling knowledge transfer between related tasks. In addition to current diagnostic and predictive tasks, AI methods can also be employed to support ophthalmologists with additional information impossible to obtain by sole visual inspection. For instance, the objective quantification of the area of corneal ulcer via a combination of segmentation and detection techniques can assist ophthalmologists in precisely evaluating whether the treatment is effective on patients in follow-up visits. If the area becomes smaller, it indicates that the condition has improved. Otherwise, it denotes that the condition has worsened and treatment strategies may need to change. To date, AI is not immune to the garbage-in, garbage-out weakness, even with big data. Appropriate data preprocessing to acquire high-quality training sets is critical to the success of AI systems. , While AI systems have good performance (e.g., detecting corneal diseases) in high-quality images, they often perform poorly in low-quality images. , Nevertheless, the performance of human doctors in low-quality images is better than that of an AI system, exposing a vulnerability of the AI system. As low-quality images are inevitable in real-world settings, , exploring approaches that can improve the performance of AI systems in low-quality images is needed to enhance the robustness of AI-based products in clinical practice. A lot of studies have drawn overly optimistic conclusions based on AI systems’ good performance on external validation datasets. However, such results are not evidence of the clinical usefulness of AI systems. Well-conducted and well-reported prospective studies are essential to provide proof to truly demonstrate the added value of AI systems in ophthalmology and pave the way to clinical implementation. Recent guidelines, such as the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT)-AI extension, Consolidated Standards of Reporting Trials (CONSORT)-AI extension, and Standards for Reporting of Diagnostic Accuracy Study (STARD)-AI, may improve the design, transparency, reporting, and nuanced conclusions of AI studies, rigorously validating the usefulness of medical AI, and ultimately improving the quality of patient care. , , In addition, an international team established an evaluation framework termed Translational Evaluation of Healthcare AI (TEHAI) focusing on the assessment of translational aspects of AI systems in medicine. The evaluation components (e.g., capability, utility, and adoption) of TEHAI can be used at any stage of the development and deployment of medical AI systems. Patient privacy and data security are major concerns in medical AI development and application. Several approaches may help address these issues. First, sensitive data should be obtained and used in research with patient consent, and anonymization and aggregation strategies should be adopted to obscure personal details. Any clinical institution should handle patient data responsibly, for example, by utilizing appropriate security protocols. Second, differential privacy (DP), a data-perturbation-based privacy approach that is able to retain the global information of a dataset while reducing information about a single individual, can be employed to reduce privacy risks and protect data security. Based on this approach, an outside observer cannot infer whether a specific individual was utilized for acquiring a result from the dataset. Third, homomorphic encryption, an encryption scheme that allows computation on encrypted data, is widely treated as a gold standard for data security. This approach has successfully been applied to AI algorithms and to the data that allow secure and joint computation. Recently, regulatory agencies, such as the Food and Drug Administration (FDA), have proposed a regulatory framework to evaluate the safety and effectiveness of AI-based medical devices during the initial premarket review. Specifically, manufacturers have to illustrate what aspects they intend to achieve through AI devices, and how the devices will learn and change while remaining effective and safe, as well as strategies to reduce performance loss. This regulatory framework is good guidance for research groups to better develop and report their AI-based medical products. Conclusions AI in ophthalmology has made huge strides over the past decade. Plenty of studies have shown that the performance of AI is equal to and even superior to that of ophthalmologists in many diagnostic and predictive tasks. However, much work remains to be done before deploying AI products from bench to bedside. Issues such as real-world performance, generalizability, and interpretability of AI systems are still insufficiently investigated and will require more attention in future studies. The solution of data sharing, data annotation, and other related problems will facilitate the development of more robust AI products. Strategies such as customization of AI algorithms for a specific clinical task and utilization of continual learning techniques may further improve AI’s ability to serve patients. RCTs and prospective studies following special guidelines (e.g., SPIRIT-AI extension, STARD-AI, and FDA’s guidance) can rigorously demonstrate whether AI devices would bring a positive impact to real healthcare settings, contributing to the clinical translation of these devices. Although this field is not completely mature yet, we hope AI will play an important role in the future of ophthalmology, making healthcare more efficient, accurate, and accessible, especially in regions lacking ophthalmologists. AI in ophthalmology has made huge strides over the past decade. Plenty of studies have shown that the performance of AI is equal to and even superior to that of ophthalmologists in many diagnostic and predictive tasks. However, much work remains to be done before deploying AI products from bench to bedside. Issues such as real-world performance, generalizability, and interpretability of AI systems are still insufficiently investigated and will require more attention in future studies. The solution of data sharing, data annotation, and other related problems will facilitate the development of more robust AI products. Strategies such as customization of AI algorithms for a specific clinical task and utilization of continual learning techniques may further improve AI’s ability to serve patients. RCTs and prospective studies following special guidelines (e.g., SPIRIT-AI extension, STARD-AI, and FDA’s guidance) can rigorously demonstrate whether AI devices would bring a positive impact to real healthcare settings, contributing to the clinical translation of these devices. Although this field is not completely mature yet, we hope AI will play an important role in the future of ophthalmology, making healthcare more efficient, accurate, and accessible, especially in regions lacking ophthalmologists. |
Cytology of a seminoma in a koi ( | e7806baa-6eea-46c5-b609-54ea2da511ea | 11315794 | Pathology[mh] | Koi ( Cyprinus carpio ) is an ornamental variety of common carp frequently kept as pets in domestic ponds worldwide. Due to their tame behavior and long lifespan, they have high economic and emotional value as pets (Sirri et al. ). To date, only a few studies have reported the prevalence of tumors in koi (Sirri et al. ). Some tumors in fish have been attributed to genetic factors (Meierjohann and Schartl ; Nairn et al. ); others were associated with viral infection (Hanson et al. ; Coffee et al. ) or with environmental contamination (Fabacher and Baumann ; Baumann et al. ; Harshbarger and Clark ). However, since the presence of tumors in koi populations includes just sporadic case reports of tumors worldwide (Knüsel et al. ; Sirri et al. , ; Stegeman et al. ), data about the prevalence or significance of neoplastic lesions in koi are still missing (Ferraro et al. ). Among the tumors, neoplastic lesions of internal organs are particularly represented, with case numbers increasing over the last years (Ott Knüsel et al. ). In cyprinids, a high prevalence of spontaneous gonadal neoplasms has been reported in hybrids of goldfish Carassius auratus L. × common carp Cyprinus carpio L. (Sonstegard ; Leatherland and Sonstegard ; Dickman and Steele ; Granado-Lorencio et al. ; Down and Leatherland ; Sirri et al. ). According to data collected by breeders and examination of various previous documents, ovarian neoplasms in ornamental koi Cyprinus carpio L. are similar to those described in wild goldfish × carp hybrids: they are common in sexually mature females and originate from the ovary, although the cellular origin is often difficult to determine (Groff ). Of the gonadal tumors described in the literature in fish, ovarian tumors are the most reported, while testicular tumors are more rarely described (H. Schmidt-Posthaus & R. Knüsel unpubl. Data; Sirri et al. ). In particular, in 2010 a case of spontaneous testicular tumor was described by Sirri et al. and was classified by histological and immunohistochemical investigation according to the WHO International Histological Classification of the Tumors of the Genital System in use for mammals as diffused and classical seminoma (Kennedy et al. ; Sirri et al. ). Despite what is already reported in the literature, there are no cases in which the application of cytology has been used as a diagnostic tool to obtain an initial diagnosis at the surgical site to be confirmed later by diagnostic methods such as histology or immunohistochemistry. Therefore, the present study is the first case in which cytology is used for this purpose. In the present case, a koi was presented to the referring veterinarian due to coelomic swelling. The carp underwent surgery, which revealed an enlargement of removed testes. Testes measured 19 x 10.5 x 9 cm and 20.5 x 6 x 3.5 cm were cocoonlike and yellow whitish. Some cytological samples were performed. Cytological samples consisted of imprints obtained by placing the mass on the slide and stained with Diff Quick stain. Then, testicular samples were collected, fixed in 10% neutral buffered formalin, and serial sections were obtained and stained with Hematoxylin-Eosin (H&E) for histological examination as previously described (Armando et al. ). The cytological samples were highly cellular, poorly hemodiluted, and composed of a mixed cellular population mainly consisting of atypical cells admixed with occasional lymphocytes and embedded in a moderate amount of bluish fluid (Fig. A). Most cells consisted of round to oval cells with distinct margins, intermediate to high nucleus-cytoplasmic ratio, and moderate to scant homogeneous bluish cytoplasm. Nuclei were round and eccentric, with coarsely dispersed chromatin and occasionally a single prominent nucleolus. Anisocytosis and anisokariosis were moderate, and mitoses were rare. Numerous bi- and multinucleated atypical cells were also observed (Fig. B). Histologically, the parenchyma of both testicles was diffusely effaced and replaced by a densely cellular, multilobular, poorly demarcated, unencapsulated, infiltrative neoplasm. The neoplasm was composed of round cells arranged in sheets and small clusters, which were variably supported by a thin fibrovascular stroma. Atypical cells were round from 25 to 30 µm in diameter with abundant eosinophilic to amphophilic homogeneous cytoplasm and moderate to high nuclear-cytoplasmic ratio. Nuclei were round to oval, ranged from 15 to 25 µm in diameter, central to paracentral with finely stippled and often marginated chromatin and one occasionally visible eosinophilic nucleolus. Anisocytosis and anisokariosis were moderate to high, and there were 9 mitoses in 2.37 mm 2 ; numerous multinucleated neoplastic cells were also present. Intratumoral necrotic areas were multifocally observed (Fig. ). Considering the spread of koi, and their value as pet animals, an improvement in the veterinary diagnostic algorithmis needed. Neoplastic diseases are described in these animals, and gonadal tumors should be considered in cases of coelomic swelling in koi. In the literature, gonadal tumors in koi are described but as exposed in a recent study by Ott Knüsel et al. conducted in 2016, these tumors are mostly represented by ovarian predominantly sex cord stroma tumors, whereas tumors originating from germ cells account for only 2.5 % of coelomic neoplasms being relatively rare although reported in the literature (Ott Knüsel et al. ). The causes of the onset of these tumors are not yet known. However, the request for particular color varieties has increased the selection and inbreeding of the species; thus, a genetic predisposition has been suggested (Ott Knüsel et al. ). Moreover, since few studies exist on these tumors, environmental factors such as toxic compounds, or viral causes cannot be excluded (Sirri et al. ). In fish, organic pollutants are often absorbed through the gills and skin, and accumulate in lipid-rich tissues, such as liver, brain, gonads, and hypodermal lipid storages. (Baines et al. ) In particular, exposure to substances such as: PAH (7,12-Diniethylbenz[a]anthracene), Ethlynitrosourea, N-methyl-N’-nitro-N-nitrosoguanidine (MNNG), PCB’s, pesticides (ß-endosulfan and α-endosulfan), hydrocarbons (oil), heavy metals, are known to be related to the occurrence of gonadal tumors in fish, particularly seminomas and dysgerminomas (Baines et al. ; Bunton and Wolfe ; Spitsbergen et al. , ). In the present case, the neoplasia described was composed of cells that resemble normal germinal epithelium and have oval nuclei, straight cell borders and distinct Golgi complex,. These aspects, together with the presence of intercellular bridges, as seen in normal germinal cells, are present in seminomas (Maxie ). These histological and cytological features allow the clear distinction of these tumors from the differential diagnoses of other testicular tumors such as interstitial or Sertoli cell tumors. Given the histological and cytological findings observed in this case, the present neoplasia was diagnosed as a spontaneous seminoma. Seminomas in fish are reported in literature and are described as tumors composed of typical germ cells similar to those from humans and equivalent mammalian tumors. This enables the comparative oncologist to classify fish tumors on the same bases as mammal tumors (Masahito et al. ). According to the WHO International Histological Classification of the Tumors of the Genital System of Domestic Animals (Kennedy et al. ), the present seminoma could be classified as diffuse, given the lobular arrangement of neoplastic cells divided by a stromal component infiltrated by lymphocytes suggested a similarity with the diffuse form. However, the high malignancy of our seminoma and the probable origin of the neoplastic cells from undifferentiated seminal cells suggest that the present seminoma is ascribable to the classical type, according to the WHO classification of testicular tumors in humans (Mostofi and Sesterhenn ). As occurs for mammals, gonadal tumors are diagnosed histologically supported by cytological examination. However, in order to confirm the cytological and histological diagnosis and, above all, to classify seminomas according to classifications in human and veterinary medicine, an immunohistochemical panel, tested in a previous case of seminoma in a koi described by Sirri et al. in 2010 together with the PAS staining, is available (Sirri et al. ). This panel included several markers including in particular cytokeratin, vimentin, c-KIT, placental alkaline phosphatase (PLAP), and neuronspecific enolase (NSE), revealing an immunoreactivity of seminomatous germ cells to vimentin, PLAP, and c-KIT, but not to NSE and cytokeratin (Feitz et al. ; Foster and Ladds ; Grieco et al. ; Sirri et al. ). PLAP, which is produced ectopically by a variety of malignant tumors including human seminoma, was found to be a specific antibody for neoplastic cells of a classical histotype (Lange et al. ; Grieco et al. ). c-KIT, which is normally expressed by germ cells, has been validated as a marker to distinguish seminoma from Sertoli cell tumors, as it is also expressed by undifferentiated neoplastic seminal cells (Grieco et al. ; Yu et al. ; Sirri et al. ). However, the use of mammalian antibodies in fish tissues has certain limitations related to their specificity. In addition the immunohistochemical panel is useful for classifying the neoplasm whereas, the cytologic and histopathologic diagnosis is itself quite accurate given the particularity of the neoplasm and its very different appearance from the main differential diagnoses. Therefore, the cytological examination, which is quick, inexpensive and can be performed at the surgical site, is an excellent first-stage diagnostic tool. Little is known about the prognosis of these neoplasms as there is only one case report in the literature of a black sea bass in which surgery was performed to remove a seminoma diagnosed by histological examination. In that case, surgery was successful, as an improvement in the patient’s vital parameters and the absence of a recurrence of the neoplasm during follow-up diagnostic investigations eight weeks after surgery have been described (Weisse et al. ). The cases described in the literature concerning surgical procedures for the removal of seminomas in koi and their post-operative prognosis are rare. The present report does not provide any further information in this respect as the koi died during the surgical procedure. There are currently studies in the literature in which new anaesthetic protocols are being tested with the aim of reducing the already high anaesthesiological risk in fish. This risk depends on several factors such as the sensitivity of these species to anaesthetics, drug dosage, anaesthesia monitoring and post-operative hospitalisation (Gladden et al. ). Seminomas in koi carp are diagnosed histologically and classified immunohistochemically, but cytology, a rapid and cheap exam executable in all veterinary clinical facilities, could be a relevant preliminary diagnostic tool that may influence the entire diagnostic process. |
Accuracy of parotid gland FNA cytology and reliability of the Milan System for Reporting Salivary Gland Cytopathology in clinical practice | d3d63b81-f0f0-4427-88e1-f1f37d59c3b4 | 8453933 | Pathology[mh] | A mass in the parotid gland is potentially neoplastic. The diagnosis of lesions in the parotid can be challenging because there are more than 40 different benign and malignant salivary gland tumors. In addition, various types of metastatic tumors can be found in the intraparotid lymph nodes. Treatment of both benign and malignant tumors mostly relies on surgical resection. Furthermore, a parotid gland mass can be of nonneoplastic origin. Sialadenosis, sialolithiasis, sialadenitis, oncocytosis, sialometaplasia, cysts, and reactive enlarged lymph nodes can all occur in the parotid gland and cause a mass. The primary diagnostic technique used for the preoperative assessment of a parotid gland mass is fine‐needle aspiration cytology (FNAC). The assessment of a mass using FNAC can help the clinician to decide whether surgical intervention is necessary (is it of neoplastic or nonneoplastic origin?). Furthermore, it can play an essential role in determining the timing (is it a benign or malignant neoplasm?) and extent of surgical treatment (is the tumor subtype of a high histopathological grade or low grade?). Unfortunately, FNAC is known to have its limitations. Although some specific tumor types are mostly easily diagnosed, the differentiation between certain types of salivary gland tumors or the distinction between a benign tumor and a malignant tumor can prove challenging. This is, among other things, caused by the overlap in cell types between different salivary gland tumor types and the fact that additional staining is often of little help. On account of the technique, it is also impossible to determine signs of malignancy such as invasive growth and perineural or vaso‐invasive growth. Moreover, there is the possibility of a sampling error. A previous systematic review performed by Liu et al showed that parotid gland FNAC had an overall sensitivity of 78% and an overall specificity of 97.8% for correctly identifying the tumor's dignity (eg, benign, malignant or non‐neoplastic). In the same review, a subgroup analysis was performed for FNAC under ultrasound guidance. This showed a sensitivity of 84.8% and a specificity of 98%. Moreover, 13.3% of all FNACs performed showed indeterminate or nondiagnostic results. As a result of this sensitivity, there is a significant risk for false‐negative results (in other words, malignant tumors falsely diagnosed as benign tumors or nonneoplastic lesions). Furthermore, indeterminate and nondiagnostic FNAC diagnoses lack clarity for the patient and the treating clinician. To provide a more objective and reproducible measure for clinicians, an international group of experts supported by the American Society of Cytopathology and the International Academy of Cytology developed a categorical system in 2018 called the Milan System for Reporting Salivary Gland Cytopathology (MSRSGC). , The MSRSGC contains different diagnostic categories in which FNAC results are subdivided. The American Society of Cytopathology also estimated the risk of malignancy (ROM) within each group and subsequently provided an advised management strategy (Table ). Numerous studies have assessed the accuracy of cytopathological evaluation using the MSRSGC tool; these have concluded that this tool can make a valuable contribution to the management of a parotid gland mass. , , , , , , , , , Unfortunately, most of these studies have lacked large groups and have been mostly single‐institution or bi‐institutional studies. As a result, the claim has been made that large‐scale, multicenter studies are imperative for testing the reliability and validity of the MSRSGC classification. Head and neck (H&N) oncology care in the Netherlands is centralized in 14 hospitals: 8 head and neck oncology centers (HNOCs) and 6 head and neck oncology affiliated centers (HNOACs). HNOCs are tertiary, mostly academic referral centers, whereas HNOACs are general hospitals (GHs) that are closely affiliated with them. The HNOACs use the same treatment protocols as the related HNOCs. No previous studies have compared the diagnostic accuracy of FNAC between hospitals (ie, HNOCs, HNOACs, and GHs). Our primary objective was to study the diagnostic accuracy of FNAC and to test the validity of using the MSRSGC classification in a large nationwide cohort. The secondary objective was to examine the differences in diagnostic accuracy for parotid gland FNAC among HNOCs (n = 8), HNOACs (n = 6), and GHs (n = 36). Accordingly, we contemplate management strategies for GHs based on the possible differences between dedicated H&N centers and GHs. Patients The PALGA database, “the nationwide network and registry of histo‐ and cytopathology in the Netherlands,” was consulted to search for patients who had undergone a salivary gland resection between January 1, 2006, and January 1, 2017. Information on age, the date of examination, the side of the lesion, and the type of hospital where the diagnosis was rendered (HNOC, HNOAC, or GH) was included in the database. Study Approval This study was approved by the scientific and privacy committee of PALGA. Because of the anonymous patient data collection and its retrospective nature, the study did not fall within the remit of the Medical Research Involving Human Subjects Act. Inclusion and Exclusion Histopathological diagnoses were classified according to the 2005 World Health Organization classification for salivary gland tumors, which was most appropriate to the search period. Tumors classified as malignant epithelial tumors, borderline tumors, benign epithelial tumors, other epithelial lesions, or soft tissue lesions according to the World Health Organization classification for salivary gland tumors were included in the study. Metastatic tumors to the parotid were included because the MSRSGC classification was designed to also differentiate between these. Lymphomas were excluded because surgery is not the treatment of choice for lymphoma. Therefore, this group was underrepresented in the histopathological resections, and as a result, these FNACs lacked a definitive diagnosis for comparison. Lesions outside the parotid gland (eg, the submandibular or the sublingual gland) were excluded. Patients were excluded if the FNAC was performed more than 1 year before the resection and if there was a histopathological result of the lesion before the first FNAC because of the bias that this may have presented to the pathologist. Categorization and Analysis The MSRSGC guideline was used to categorize the cytopathological reports retrospectively by one of the authors (S.T.H.R.). In case of uncertainty, a second reviewer (A.C.H.V.E.V.G.) was consulted. Each FNAC result was compared with the definitive histopathological diagnosis as the golden standard; it was stated if the result was discordant or concordant with the FNAC result. In case of revision of the histopathological diagnosis, the revised diagnosis was considered the golden standard. The ROM was determined for each MSRSGC category. Consequently, the sensitivity and specificity for diagnosing malignancy were calculated as measures of diagnostic accuracy. For this analysis, the suspected malignant (MSRSGC V) and malignant (MSRSGC VI) groups were classified as positive cytopathological tests, and the nonneoplastic (MSRSGC II) and benign salivary gland neoplasm (MSRSGC IVa) groups were categorized as negative test results. Indeterminate (MSRSGC III or IVb) and nondiagnostic (MSRSGC I) results were excluded from the analysis of sensitivity and specificity; these results were separately reported. Cytopathological revisions were excluded from the diagnostic accuracy analysis because of the possible bias that these may present. Statistical Analysis Sensitivity and specificity were calculated along with their respective 95% confidence intervals. χ 2 tests were used to test for differences between sensitivity and specificity between the different types of hospitals. Two‐sided P values < .05 were considered statistically significant. Statistical analysis was performed with SPSS version 26.0 (IBM Corp, Armonk, New York). The PALGA database, “the nationwide network and registry of histo‐ and cytopathology in the Netherlands,” was consulted to search for patients who had undergone a salivary gland resection between January 1, 2006, and January 1, 2017. Information on age, the date of examination, the side of the lesion, and the type of hospital where the diagnosis was rendered (HNOC, HNOAC, or GH) was included in the database. This study was approved by the scientific and privacy committee of PALGA. Because of the anonymous patient data collection and its retrospective nature, the study did not fall within the remit of the Medical Research Involving Human Subjects Act. Histopathological diagnoses were classified according to the 2005 World Health Organization classification for salivary gland tumors, which was most appropriate to the search period. Tumors classified as malignant epithelial tumors, borderline tumors, benign epithelial tumors, other epithelial lesions, or soft tissue lesions according to the World Health Organization classification for salivary gland tumors were included in the study. Metastatic tumors to the parotid were included because the MSRSGC classification was designed to also differentiate between these. Lymphomas were excluded because surgery is not the treatment of choice for lymphoma. Therefore, this group was underrepresented in the histopathological resections, and as a result, these FNACs lacked a definitive diagnosis for comparison. Lesions outside the parotid gland (eg, the submandibular or the sublingual gland) were excluded. Patients were excluded if the FNAC was performed more than 1 year before the resection and if there was a histopathological result of the lesion before the first FNAC because of the bias that this may have presented to the pathologist. The MSRSGC guideline was used to categorize the cytopathological reports retrospectively by one of the authors (S.T.H.R.). In case of uncertainty, a second reviewer (A.C.H.V.E.V.G.) was consulted. Each FNAC result was compared with the definitive histopathological diagnosis as the golden standard; it was stated if the result was discordant or concordant with the FNAC result. In case of revision of the histopathological diagnosis, the revised diagnosis was considered the golden standard. The ROM was determined for each MSRSGC category. Consequently, the sensitivity and specificity for diagnosing malignancy were calculated as measures of diagnostic accuracy. For this analysis, the suspected malignant (MSRSGC V) and malignant (MSRSGC VI) groups were classified as positive cytopathological tests, and the nonneoplastic (MSRSGC II) and benign salivary gland neoplasm (MSRSGC IVa) groups were categorized as negative test results. Indeterminate (MSRSGC III or IVb) and nondiagnostic (MSRSGC I) results were excluded from the analysis of sensitivity and specificity; these results were separately reported. Cytopathological revisions were excluded from the diagnostic accuracy analysis because of the possible bias that these may present. Sensitivity and specificity were calculated along with their respective 95% confidence intervals. χ 2 tests were used to test for differences between sensitivity and specificity between the different types of hospitals. Two‐sided P values < .05 were considered statistically significant. Statistical analysis was performed with SPSS version 26.0 (IBM Corp, Armonk, New York). In total, 24,164 patients with a salivary gland resection were gathered from the PALGA database. The inclusion and exclusion process is summarized in Figure . During the study period, 12,898 FNAC aspirates from the parotid gland were taken from a total of 9672 patients who underwent subsequent resection of the lesion within 1 year after the FNAC. The study group consisted of 4807 male patients (49.7%) and 4865 female patients (50.3%). The mean age at first FNAC was 54.8 years (range, 0‐98 years; SD, 15.5 years). The left and right parotid glands were equally involved (50% and 49.7%). In 0.3% of cases, the side was unknown. A small portion of patients (0.9%) had a bilateral mass surgically removed. Twenty‐nine percent of all FNAC aspirates were evaluated in an HNOC, 19.3% were evaluated in an HNOAC, and 51.6% were evaluated in a GH. In 81.8% of the patients, only 1 FNAC was performed. A second FNAC was performed in 15%, and a third FNAC was performed in 2.8%. Only 0.4% of patients had 4 or more aspirates taken (up to 7 FNACs). After resection, the final histopathological diagnosis yielded a malignant neoplasm in 12.4%, a benign tumor in 84.2%, and nonneoplastic disease in 3.4%. The overall distribution of FNAC aspirates classified according to the MSRSGC is provided in Table . The most frequent preoperative FNAC result was a benign salivary gland tumor (61.4%), which was followed by a nondiagnostic result (19.0%), a salivary gland neoplasm of unknown malignant potential (SUMP; 6.4%), and a malignant tumor (4.7%). The same table also shows the ROM for each category. These were high in both the malignant (99.3%) and suspected malignant categories (83%). The SUMP and atypia of unknown significance (AUS) categories had ROMs of 28.6% and 29%, respectively. The ROMs for the nondiagnostic category (12.5%), the nonneoplastic category (10.3%), and benign salivary gland tumors (2.3%) were lower. The majority of all cytopathological smears (57%) were taken and evaluated at a GH. The HNOCs and HNOACs evaluated 21.7% and 21.3%, respectively. The sensitivity of FNAC was highest in the dedicated H&N centers: HNOCs had a sensitivity of 88.1%, HNOACs had a lower sensitivity of 79.7%, and GHs had a sensitivity of 75%. The sensitivity and specificity of FNAC differed significantly between HNOCs and HNOACs ( P = .006 and P = .034, respectively) and between HNOCs and GHs ( P < .001 and P = .002, respectively). When we compared HNOACs and GHs, there were no differences in sensitivity or specificity found ( P = .205 and P = .803, respectively). Table summarizes these results. Most benign tumors and nonneoplastic parotid gland lesions were resected at a GH (Table ). Malignant tumors, on the other hand, were mostly resected at an HNOC; 15.4% of malignant tumors (n = 185) were removed at a GH. Among the 185 malignant tumors removed at a GH, 45 tumors (24.3%) had a preoperative diagnosis of a malignant parotid gland tumor or a suspected malignant parotid gland tumor (MSRSGC V or VI), 33 tumors (17.8%) had a preoperative diagnosis of SUMP (MSRSGC IVb), and 9 tumors (4.9%) had a preoperative diagnosis of AUS (MSRSGC III). The other 98 tumors (53.0%) had a nondiagnostic, nonneoplastic, or benign preoperative FNAC result. FNAC provided a false‐negative diagnosis for some of the resected malignant parotid tumors (MSRSGC II/MSRSGC IVa). The histopathological subtypes associated with the highest false‐negative rates on FNAC before surgery are shown in Table . Histopathologically proven myoepithelial carcinomas were shown to have the highest false‐negative rate (57.1%) on preoperative FNAC. The exact cytopathological subtype diagnoses within MSRSGC class IVa that proved malignant on histopathological examination (false‐negative or false‐benign results) were basal cell adenoma (50%, n = 7), myoepithelioma (33.3%, n = 4), and oncocytoma (9.1%, n = 2). More frequently occurring cytopathological diagnoses such as Warthin tumor and pleomorphic adenoma had false‐negative rates of 2.1% and 1.9%, respectively. The false‐negative results in the Warthin tumor group included mucoepidermoid carcinoma (n = 20), acinic cell carcinoma (n = 17), salivary duct carcinoma (n = 4), metastatic squamous cell carcinoma (n = 2), adenocarcinoma not otherwise specified (n = 1), epithelial‐myoepithelial carcinoma (n = 1), and carcinoma ex pleomorphic adenoma (n = 1). The false‐negative results in the pleomorphic adenoma group included carcinoma ex pleomorphic adenoma (n = 17), epithelial‐myoepithelial carcinoma (n = 16), acinic cell carcinoma (n = 11), myoepithelial carcinoma (n = 8), adenoid cystic carcinoma (n = 9), adenocarcinoma not otherwise specified (n = 5), salivary duct carcinoma (n = 3), basal cell adenocarcinoma (n = 3), secretory carcinoma (n = 2), polymorphic adenocarcinoma (n = 1), sarcoma (n = 1), large cell carcinoma (n = 1), primary squamous cell carcinoma (n = 1), metastatic squamous cell carcinoma (n = 1), and melanoma (n = 1). FNAC is the standard diagnostic test in the management of salivary gland lesions, but as illustrated previously, it is known to have its limitations. Fortunately, the presented ROMs validate the use of the MSRSGC. The distribution of the ROMs in our study (Table ) showed minor differences in comparison with the estimates of the MSRSGC (Table ). The ROMs of MSRSGC III (AUS), MSRSGC V (suspected malignant), and MSRSGC VI (malignant) proved slightly higher. However, these results still justify the management proposed in the MSRSGC, namely to either repeat FNAC or perform surgery for MSRSGC III and to perform surgery for MSRSGC V and VI. Previous studies reporting on the ROMs for MSRSGC categories have shown variation in the distribution of ROMs, especially in the MSRSGC I, II, III, and IVb categories. , , , , , , , , , These studies are summarized in Table . Only studies that included at least 200 patients with histopathological confirmation were included in this summary. The vast majority of these studies also included patients who had no histopathological confirmation, which means that they had only clinical or radiological correlation to verify the assigned MSRSGC category. These patients were mostly represented in the nondiagnostic (MSRSGC I) and nonneoplastic (MSRSGC II) categories. The combined ROMs found in these previous studies are in line with the current results in the Netherlands, which are presented in this study. This proves the validity of our data and the robustness of the MSRSGC classification. The observed variance in ROM between the independent studies can probably be explained by the fact that they had different cohorts that may have varied in the distribution of dignity and tumor types. Metastatic cutaneous squamous cell carcinoma (cSCC) to the parotid, for instance, is one of the tumors that has varying prevalence per geographical zone. The majority of the studies included in Table include metastatic cSCC to the parotid. Because of the relatively easy diagnosis of metastatic cSCC on FNAC, this might positively affect the results in countries where cSCC is more often diagnosed. Because we calculated the combined ROMs, we have adjusted for this possible bias. The combined ROMs of the previous studies and the observed ROMs in our study are slightly different from the ROMs presented by Faquin et al. The ROMs of the nondiagnostic category in both this study and the combined previous studies (12.5% and 14.1%) proved to be lower than the rate of 25% proposed by the authors of the MSRSGC. For the AUS category, both our results and the combined results from previous studies (29% and 34.8%) showed higher ROMs than proposed by the original authors (20%). Lastly, the ROMs of the suspected malignant category (current study, 83%; combined studies, 80.0%) and the malignant category (current study, 99.3%; combined studies, 97.5%) were seemingly higher than the values of 60% and 90% proposed in the original classification. Therefore, we propose changing the expected ROMs of the categories as follows: <15% for the nondiagnostic category, ±30% for the AUS category, >80% for the suspected malignant category, and >95% for the malignant category. These changes, in our opinion, do not affect the original management propositions of the MSRSGC classification. The sensitivity of FNAC was lower in GHs and HNOACs than HNOCs, whereas the specificity was nearly the same. This implicates a higher chance of false‐negative results (and thus missing the diagnosis of malignancy) at less specialized hospitals. There are 2 likely causes for these differences. First, it is well established that the procedure's accuracy is lower when the operators who perform the fine‐needle aspiration (H&N surgeons, radiologists, or pathologists) are less experienced. Second, the combination of a relatively rare diagnosis such as salivary gland carcinoma and the previously mentioned pitfalls in cytopathology (different tumor types with overlapping in morphology and cell types with only subtle or no differences between tumors, limited benefit from additional staining, and a lack of clear cytomorphological signs of malignancy in some malignant tumors) make the pathologist's experience in the assessment of salivary gland FNAC material also likely to influence the diagnostic accuracy. However, we have to note that even the most experienced H&N pathologists can find the diagnosis of some salivary gland tumors, on both cytopathology and histopathology, troublesome. The histopathological subtypes that were most often misdiagnosed on FNAC were myoepithelial carcinoma, epithelial‐myoepithelial carcinoma, mucoepidermoid carcinoma, carcinoma ex pleomorphic adenoma, acinic cell carcinoma, and adenoid cystic carcinoma. Myoepithelial carcinoma, adenoid cystic carcinoma, and epithelial‐myoepithelial carcinoma show a resemblance to pleiomorphic adenomas because of shared cell types, and they often lack overt features of malignancy. Acinic cell carcinoma is often mistaken on cytopathology for normal salivary gland tissue because it can show a close resemblance to normal acinic cells. The difficulty in the correct cytopathological diagnosis of (low‐grade) mucoepidermoid carcinoma is mostly due to sampling error because this tumor type consists of both an epithelial component and a cystic component. If only or mainly the cystic component is sampled, the tumor can be mistaken for a mucous cyst or Warthin tumor, whereas the epithelial component contains a variety of histopathological patterns, which may show overlapping with pleomorphic adenoma. , Carcinoma ex pleomorphic adenoma is also prone to false‐negative results, partly because of an overlap in morphological components with other tumor types and partly because of a considerable risk of sampling error when only the benign segment of the tumor is sampled. , Also, there were several cytopathological diagnoses carrying a relatively high risk for a false‐negative result. Although the total number of diagnoses was small, 50% of the preoperatively assumed basal cell adenomas and 33.3% of the assumed myoepitheliomas turned out to be malignant. This can be explained by the fact that invasive growth, which is the main discriminator between these two and their malignant counterparts (basal cell carcinoma and myoepithelial carcinoma), is never clear on cytopathology. Furthermore, oncocytoma (9.1%) showed relatively high false‐negative rates. Clinicians should, therefore, be warned in case of these 3 cytopathological diagnoses. Cytopathologically diagnosed Warthin tumors and pleomorphic adenomas both carry a low false‐negative rate of ±2%. The majority of false‐negatively diagnosed Warthin tumors were primary salivary gland malignancies such as mucoepidermoid carcinoma and acinic cell carcinoma. The first was probably caused by the cystic components of Warthin tumors, which are also found in low‐grade mucoepidermoid carcinoma, along with the general resemblance of their cell types. The reason that acinic cell carcinoma is often mistaken for a Warthin tumor lies mainly in the prominent (cystic) lymphoid infiltrate seen in both. High‐volume surgery can have a favorable effect on overall survival in the treatment of H&N cancers. , , However, high volumes are not easily achieved in the treatment of rare cancers such as salivary gland carcinoma, let alone for all its distinct subtypes. Centralization of care, therefore, is of the utmost importance for achieving high volumes. Previous studies have shown that the centralization of care for rare diseases can be beneficial to overall survival. Likewise, a recent study showed that major salivary gland carcinoma could also benefit from centralization of care because high‐volume hospitals were shown to have lower rates of positive surgical margins. Therefore, it is our strong belief that malignant parotid tumors should be surgically removed at dedicated high‐volume centers. Unfortunately, our results showed that many resections of malignant parotid gland tumors were performed in GHs (15.4%). In most of the malignant parotid tumors resected at a GH, the patient was not referred to a dedicated H&N center because of a false‐negative preoperative FNAC (MSRSGC I, II, or IVa was found in 53%). However, 17.8% had a preoperative diagnosis of SUMP (MSRSGC IVb), and 4.9% had a preoperative diagnosis of AUS (MSRSGC III). In 24.3%, the preoperative diagnosis even was malignant or suspicious for malignancy (MSRSGC V or VI). Because of the high numbers of surgically removed malignant tumors at GHs, the previously presented ROMs, and the differences in the false‐negative rates and sensitivity among the various treatment facilities, a referral scheme for clinicians at GHs is proposed. The goal is to minimize the chance of false‐negative preoperative cytopathological diagnoses and to ensure that more patients with malignant parotid tumors are treated at a dedicated H&N center. In summary, a nondiagnostic result (MSRSGC I) should warrant repeating FNAC. In cases with the same (nondiagnostic) result, clinical and radiological correlation is necessary. When there is no or low suspicion of a neoplasm, follow‐up can be appropriate. If there is a high suspicion of neoplastic disease, surgery (when a benign tumor is suspected) or referral to a dedicated H&N center (in case of clinical or radiological doubt about the dignity) should be considered. Nonneoplastic results (MSRSGC II) should be carefully evaluated. The clinician must decide after clinical and radiological correlation if follow‐up or repeat FNAC is indicated. If this repeated FNAC yields similar results and the FNAC result does not correspond to the clinical and/or radiological correlation, surgery (when a benign tumor is suspected) or referral to a dedicated H&N center is indicated. In patients with a cytopathological diagnosis of AUS (MSRSGC III), referral to a dedicated H&N hospital should be considered on the basis of the ROM of 30%. Clinicians at dedicated centers are encouraged to repeat the FNAC because of the higher diagnostic accuracy of FNAC at these centers. Benign salivary gland neoplasms (MSRSGC IVa) can be removed safely by an experienced surgeon at a GH. Because of the relatively high ROM (35%) of SUMP (MSRSGC IVb), treating these neoplasms at a dedicated H&N center should be considered. Clinicians at dedicated centers are encouraged to repeat the FNAC because of the higher diagnostic accuracy of FNAC at these centers. We strongly advise that suspected malignant and malignant tumors (MSRSGC V and VI) be treated at a dedicated H&N hospital. This referral scheme is summarized in the flowchart in Figure . On account of the sole inclusion of patients who had a histopathological correlation after their parotid gland FNAC, this study may be limited by the underrepresentation of FNACs in MSRSGC II (nonneoplastic) because these are not always surgically resected. Furthermore, earlier studies have observed that lymphomas predominantly contribute to a higher ROM in the nonneoplastic category. These were also excluded from our analysis because they were underrepresented in this surgically treated cohort. Another limiting factor is that no histopathological reassessment of the resections was performed. A previous study has shown that the reassessment of histopathology in malignant salivary gland tumors is associated with changes in the histopathological subtype, the origin of the tumor, or even dignity in a number of patients. Because of our large sample size, histopathological reassessment was practically impossible to perform. Our study, which to our knowledge is the most extensive retrospective, nationwide study evaluating parotid gland FNACs to date, once more proves the effectiveness and validity of the MSRSGC. Therefore, we strongly recommend the use of this diagnostic tool for reporting on salivary gland cytology. In addition, we present referral guidelines for clinicians at GHs based on our results. These help in minimizing the chance of false‐negative preoperative FNAC results and can further optimize care for patients with parotid gland tumors. No specific funding was disclosed. The authors made no disclosures. Sam T. H. Reerds: Study concept, methodology design, data collection, statistical analysis, supervision of the statistical analysis, writing–original draft, drafting of the management and referral flowchart for general hospitals, and revision and finalization of the manuscript. Adriana C. H. Van Engen–Van Grunsven: Study concept, methodology design, supervision of the statistical analysis, drafting of the management and referral flowchart for general hospitals, and revision and finalization of the manuscript. Frank J. A. van den Hoogen: Drafting of the management and referral flowchart for general hospitals and revision and finalization of the manuscript. Robert P. Takes: Drafting of the management and referral flowchart for general hospitals and revision and finalization of the manuscript. Henri A. M . Marres: Study concept, methodology design, drafting of the management and referral flowchart for general hospitals, and revision and finalization of the manuscript. Jimmie Honings: Study concept, methodology design, supervision of the statistical analysis, writing–original draft, drafting of the management and referral flowchart for general hospitals, and revision and finalization of the manuscript. |
The Effectiveness of a Primary Care Diabetes Education and Self‐Management Program in Ireland: A 6‐Month Follow‐Up Study | a67da6b5-ecbd-4ddd-93f5-30d1e7f5d6ed | 11839741 | Patient Education as Topic[mh] | Introduction Type 2 diabetes mellitus imposes a substantial burden on individuals, healthcare professionals (HCPs), and the broader health system . People living with type 2 diabetes mellitus face daily decisions affecting their health, including dietary choices, medication compliance, and physical activity. These decisions are often made with limited guidance from health care professionals . Structured diabetes self‐management education (DSME) is now integral to optimal diabetes care, alongside glucose‐lowering medications and psychological support . These programmes aim to improve participants knowledge of how to incorporate health‐promoting behaviours into their daily lives. This includes how to improve metabolic control, decisions around medication, and improve quality of life . In addition, group education is cost‐effective and provides opportunities for shared experiences . The Institute of Public Health in Ireland have reported that the increasing prevalence of type 2 diabetes mellitus is driven primarily by the ageing population . The reported prevalence of diagnosed and undiagnosed type 2 diabetes mellitus is 8.4% in those over 50 years, with a higher prevalence in men (10.3%) than women (6.6%) . Prevalence also increases with advancing age . Among individuals in this age group, 26% reported microvascular complications, and 15% reported macrovascular complications . The American Diabetes Association “Standards of Care in Diabetes” advises on therapeutic goals of clinical parameters for glycaemic and weight management control in order to reduce or delay the micro and macrovascular complications of diabetes . Diabetes also exerts a significant financial strain on the Irish health service, costing approximately 10% (€1.7 billion) of the healthcare budget . This thereby supports the need for early initiation of DSME. Given the average Irish life expectancy of 82 years, there is a critical need to centre the ‘end‐to‐end’ treatment of diabetes within the primary care setting . There is an onus on HCPs to help people to understand that DSME should be included as part of their treatment plan in line with the new emerging models of integrated care . The current service model of integrated care for type 2 diabetes mellitus in Ireland advocates for structured patient education to empower individuals, improve quality of life, and promote active self‐management Referrals to DSME is ideally within 3 months of diagnosis . Quality standards for the provision of DSME have been established . Currently, there are three structured education programmes for the management of type 2 diabetes mellitus . These include DISCOVER Diabetes ‐ Type 2 (Diabetes Insights and Self Care Options via Education and Reflection), CODE (Community Orientated Diabetes Education) and DESMOND (Diabetes Education and Self‐Management for Ongoing and Newly Diagnosed) , with DESMOND currently being the only programme to be validated using a randomised controlled trial . However, the impact of DSME on clinical markers within Irish services is inadequately published to date, creating a knowledge gap . Documenting participant outcomes is important to establish the efficacy of DSME interventions and service modelling. This study aims to evaluate the effectiveness of the DESMOND structured education programme in people with newly diagnosed and existing diabetes who attended a locally developed follow‐up session 6 months post‐DESMOND in a primary care centre in the Midwest of Ireland. Methods 2.1 Study Design A retrospective chart review was used to extract patient outcome data from the dietetic records of participants who completed DSME during 2018 in the Health Service Executive Midwest Community Healthcare area. Ethical approval was granted by the HSE University Hospital Limerick Research Ethics Committee. 2.2 Population and DSME Intervention Diabetes Education and Self‐Management for Ongoing and Newly Diagnosed (DESMOND) is an evidence‐based group education program for people with type 2 diabetes mellitus, focusing on long‐term self‐management through behaviour change . Programme completion requires attendance at 6‐h of education either on one whole day or two half days in groups of up to 12 people with type 2 diabetes mellitus who can bring an accompanying person . In the HSE Midwest service, local policy is to offer a follow‐up session after 6‐months. The combination of DESMOND group and the 6‐month locally developed group will hereafter be called ‘The Midwest DSME’. Data was collected from one primary care centre in the Midwest of Ireland. Baseline data were collected from participants prior to attending the first session and again at the 6‐month follow‐up. Participants were eligible for inclusion in this study if they attended the Midwest DSME during 2018. The exclusion criteria were those who did not attend any or only one session. Attendance at both sessions was deemed necessary in order to obtain the relevant clinical measurements (weight and BMI) and evaluate change 6 months post DESMOND. 2.3 Clinical Outcome Measures The primary outcome was glycated haemoglobin (HbA1c) levels. HbA1c reflects average plasma glucose over the previous 8–12 weeks and is the preferred test for assessing glycaemic control. HbA1c values are provided per routine collection of the clinical parameters among participants attending DESMOND, and no additional blood samples were required. Clinical goals used for HbA1c were < 53 mmol/mol (< 7%) for adults and older adults who are otherwise healthy, with few coexisting illnesses and intact cognitive function and functional status . Weight (kg) and BMI (kg/m 2 ) data were measured by a registered dietitian and collected onsite. The goal for weight loss was set at 3%–7% of baseline weight, which is shown to improve glycaemia and other cardiovascular risk factors . 2.4 Statistical Analysis A descriptive data analysis was conducted using SPSS version 28 . Continuous variables were summarised using means, standard deviations, and ranges. Categorical variables were summarised using counts and percentages. Shapiro–Wilk tests were used to test the normality of the data. BMI was categorised according to the WHO classification (underweight, normal weight, overweight and obese) . All missing data was assumed to be randomly missing and were not replaced. Paired sample t ‐tests were used to compare the means of the baseline and post‐intervention results. McNemar chi‐square tests were used to determine the statistical significance of HbA1c values in relation to the 53 mmol/mol glycaemic goal for health before and after the intervention . Statistical significance was assumed when p < 0.05. Study Design A retrospective chart review was used to extract patient outcome data from the dietetic records of participants who completed DSME during 2018 in the Health Service Executive Midwest Community Healthcare area. Ethical approval was granted by the HSE University Hospital Limerick Research Ethics Committee. Population and DSME Intervention Diabetes Education and Self‐Management for Ongoing and Newly Diagnosed (DESMOND) is an evidence‐based group education program for people with type 2 diabetes mellitus, focusing on long‐term self‐management through behaviour change . Programme completion requires attendance at 6‐h of education either on one whole day or two half days in groups of up to 12 people with type 2 diabetes mellitus who can bring an accompanying person . In the HSE Midwest service, local policy is to offer a follow‐up session after 6‐months. The combination of DESMOND group and the 6‐month locally developed group will hereafter be called ‘The Midwest DSME’. Data was collected from one primary care centre in the Midwest of Ireland. Baseline data were collected from participants prior to attending the first session and again at the 6‐month follow‐up. Participants were eligible for inclusion in this study if they attended the Midwest DSME during 2018. The exclusion criteria were those who did not attend any or only one session. Attendance at both sessions was deemed necessary in order to obtain the relevant clinical measurements (weight and BMI) and evaluate change 6 months post DESMOND. Clinical Outcome Measures The primary outcome was glycated haemoglobin (HbA1c) levels. HbA1c reflects average plasma glucose over the previous 8–12 weeks and is the preferred test for assessing glycaemic control. HbA1c values are provided per routine collection of the clinical parameters among participants attending DESMOND, and no additional blood samples were required. Clinical goals used for HbA1c were < 53 mmol/mol (< 7%) for adults and older adults who are otherwise healthy, with few coexisting illnesses and intact cognitive function and functional status . Weight (kg) and BMI (kg/m 2 ) data were measured by a registered dietitian and collected onsite. The goal for weight loss was set at 3%–7% of baseline weight, which is shown to improve glycaemia and other cardiovascular risk factors . Statistical Analysis A descriptive data analysis was conducted using SPSS version 28 . Continuous variables were summarised using means, standard deviations, and ranges. Categorical variables were summarised using counts and percentages. Shapiro–Wilk tests were used to test the normality of the data. BMI was categorised according to the WHO classification (underweight, normal weight, overweight and obese) . All missing data was assumed to be randomly missing and were not replaced. Paired sample t ‐tests were used to compare the means of the baseline and post‐intervention results. McNemar chi‐square tests were used to determine the statistical significance of HbA1c values in relation to the 53 mmol/mol glycaemic goal for health before and after the intervention . Statistical significance was assumed when p < 0.05. Results Two hundred and ten participants attended DESMOND in the primary care centre during 2018. Of these, 66 participants returned for the locally developed 6‐month follow‐up session. There were no statistically significant differences in characteristics (age, weight, BMI) between those who attended just the first session ( n = 144) and those who attended both sessions ( n = 66). Those who returned for the 6‐month follow‐up session had a higher mean HbA1c at baseline compared to those who attended DSME only (59 mmol/mol (7.5%) vs. 53.78 mmol/mol (7.1%), p = 0.046). Of those who attended the Midwest DSME ( n = 66), the mean age was 63 years (SD:11 years). The length of time since diagnosis ranged from 3 months to 11 years. Clinical characteristics are shown in Table . Mean baseline HbA1c was 58.4 mmol/mol (7.5%) (SD: 16 mmol/mol). Mean HbA1c at follow‐up was 52 mmol/mol (6.9%) (SD: 10.43 mmol/mol) ( p = 0.006 ). At follow‐up, 60% of participants had a reduction in HbA1c, with values ranging from 1 to 47 mmol/mol. At baseline, 52% ( n = 27) were below the 53 mmol/mol (7%) glycemic goal recommended for most individuals with type 2 diabetes mellitus. This increased to 71% ( n = 37) at follow‐up ( p < 0.001 ) (Table ). Mean bodyweight at baseline was 93.4 kg (SD: 20 kg). At follow‐up, mean body weight was 92 kg (SD: 19.98 kg). This mean decrease of 1.4 kg represented a 1.5% reduction in mean bodyweight ( p = 0.21 ). A subset analysis of BMI found that the percentage of individuals in the normal weight category (BMI: 18.5–25 kg/m 2 ) increased from 9.4% to 13.2% at follow‐up. This coincides with a corresponding decrease in the overweight BMI category (BMI: 25–29.9 kg/m 2 ) from 30.2% to 26.4%. The percentage of those in the obese category (BMI > 30 kg/m 2 ) remained unchanged at 60.4%. Although the number of participants in the obese category remained unchanged ( n = 32), it was observed that 18 participants (56%) recorded a mean weight loss of 4.4 kg (range 0.8–11.4 kg). Of these 18 individuals, 39% ( n = 7) lost > 5% body weight, a clinically significant marker . Discussion The aim of this study was to evaluate the efficacy of the DESMOND programme in participants who returned for a locally developed 6‐month follow‐up session delivered by registered dietitians and a diabetes nurse specialist on key clinical outcomes. Our results found that HbA1c was reduced by 6.54 mmol/mol (2.7%) at follow‐up. This is comparable to previous work conducted in the UK by Chatterjee et al. who reported a 10 mmol/mol (3.1%) reduction in HbA1c at 6 months post‐intervention. Similar findings have been reported across diverse populations attending other similar DSME programmes including China and Latino adults . As the cohort's mean age was 63 years, 53 mmol/mol (7%) was selected as the glycaemic goal for this study. The ADA recommends an HbA1c target of 53–58 mmol/mol (7%–7.5%) in older adults (> 65 years) who are otherwise healthy with few existing comorbidities . In practice, HbA1c goals are person‐centred and set in collaboration between the individual and healthcare professional . Therefore, these results may potentially underestimate the benefits of DSME, as they do not account for individuals with more lenient glycaemic goals, such as older individuals, those with comorbidities, longer time since diagnosis, or those who cannot independently manage their condition. The UK Prospective Diabetes Study (UKPDS) was one of the largest and longest studies ever undertaken in diabetes with a median follow‐up of 10 years, which found that there was no threshold in terms of clinical benefits of improving HbA1c values . Hence, any improvement in glycaemic control was seen to be beneficial for health . An epidemiological extrapolation of this study showed that each 1% reduction in mean HbA1c was associated with significant decreases in risk for any diabetes end‐point (diabetes‐related mortality, myocardial infarction and microvascular complications) with no threshold of risk being observed for any end point . More recent research using data from the UK Biobank of 471,399 (56% women) individuals without cardiovascular disease established an 18% greater risk of a myocardial infarction in both sexes with each 1% higher HbA1c . In this present study, 60% of participants saw a reduction in HbA1c and, therefore, the potential for future health benefits. We found greater than 90% of participants were overweight/obese at baseline. This may be an indication of the close correlation between increased adiposity and type 2 diabetes mellitus risk . According to the Model of Integrated Care published by the Irish Health Service Executive, weight management is one of many interventons that may hep to optimise glucose control in those with type 2 diabetes mellitus who are overweight/obese . In this study, the mean decrease of 1.4 kg (1.5%) found at follow‐up, albeit not statistically significant, is comparable to previously published work . A modest weight loss improves glycaemic control and reduces the need for glucose‐lowering medication . A meta‐analysis conducted by Franz et al. concluded that a > 5% weight reduction equated to a 7 mmol/mol HbA1c reduction in those with established type 2 diabetes mellitus and up to a 13 mmol/mol reduction in newly diagnosed individuals. Our findings report that 13% of participants experienced weight loss of more than 5%. The ADA has reported that 3%–7% weight loss improves glycaemia in those with diabetes, while also referencing the importance of small attainable weight loss goals targets . It is important to acknowledge that the mean age in this study was 63 years. Central obesity was the strongest characteristic associated with diabetes in older Irish adults . Weight loss in older individuals with type 2 diabetes mellitus has been contra‐indicated by ESPEN as they acknowledge that following a weight‐reducing diet may result in loss of muscle mass and functional decline . The prevalence of malnutrition in older people with diabetes is as high or even higher than matched individuals without diabetes . If weight reduction is beneficial, it should be approached with great care, preferably in an individual session with a dietitian offering personalised support and not general DSME. It is also important to note that DESMOND is not a weight loss intervention, and participants are encouraged to choose their own risk factor for intervention. This could include physical activity, smoking, managing cardiovascular risk factors, or a psychosocial aspect. Participants, in collaboration with the educator, then set goals around their chosen risk factor, which may not have been HbA1c or weight‐related. We noted a high rate of attrition with just 31% ( n = 66) participants returning for the 6‐month follow‐up session. The motivation to attend the follow‐up session may be related to the observed higher HbA1c at baseline among those who attended . The role of HCPs in facilitating participation in DSME at critical stages of care on an ongoing basis among people with diabetes is highlighted to support ongoing self‐management and decision‐making . Since this study was conducted, the Covid‐19 pandemic has had an impact on healthcare globally and has resulted in huge delays in the delivery of DSME. While services are being restored, HCPs continue to face many challenges, including increased waiting lists, delayed diagnoses, or those who remain fearful of entering a healthcare setting environment . Covid‐19 has also had an impact on health‐related behaviours such as physical activity and maintenance of a healthy weight, which may indirectly affect those with diabetes. The need for referral to DSME may be more important now than prior to the pandemic . 4.1 Strengths and Limitations As this is an observational study, we cannot determine cause and effect. This study focused solely on clinical outcomes within the scope of the available data for reporting proposed DSME core outcome measures . However, scant empirical evidence exists regarding the impact of interventions spearheaded by dietitians in real‐world contexts. Consequently, we deemed it imperative to contribute to this body of knowledge, particularly concerning the pragmatic aspects of implementing DSME within primary care settings. In practical terms, DSME educators depend on both electronic patient lab systems and patient self‐reports of clinical outcomes. However, it is noteworthy that lab results frequently lack recent (within 3 months) biochemical data, particularly when they have not been requested at the point of referral to DSME in primary care. Hence, any changes seen cannot be solely attributed to the DESMOND education. It also needs to be considered that changes may have come about due to the participants being highly motivated individuals and opting to attend DSME. It should also be noted that this study took place prior to the Covid‐19 pandemic and prior to the widespread introduction of telemedicine. During the pandemic, DSME moved to online platforms. DSME provided via telemedicine has previously been shown to improve glycemic control, diabetes knowledge, and self‐care adherence behaviour . Future research should examine the effectiveness of telemedicine with regard to improving attendance at DSME sessions. An option for telemedicine may have improved attendance in this present study. Despite the number of participants attending DESMOND initially, only 66 attended the locally developed follow‐up session and had repeat clinical data to be included in this study. Future research should focus on reasons for non‐attendance to DSME and in particular the HCP role in helping participants to understand the importance of DSME as part of treatment plans in integrated care models . Due to demands on the local dietetic service, this study did not have the staffing resources available to collect additional measurements of body composition (e.g., waist circumference, waist‐to‐hip ratio); future research should also evaluate the changes in these anthropometric measurements alongside potential changes in HbA1c. Notably, the influence of gender on attendance rates and trajectory of type 2 diabetes mellitus cannot be ignored. Previous research conducted in this area has shown that in 365 individuals with newly diagnosed type 2 diabetes mellitus, attendance at structured education was independently associated with female gender (Odds Ratio (OR) 1.28, 95% CI 1.05–1.46), lower HbA1c (OR 0.98 mmol/mol 95% CI 0.97–0.99) and non‐smoker status (OR 1.36, 95% CI 1.07–1.55) . Therefore, future research should examine ways to improve attendance for males. Regardless, this is an evaluation of the effectiveness of the Midwest DSME impact on clinical markers in participants who completed the programme. In terms of strengths, this sample was representative of individuals in Ireland with type 2 diabetes mellitus. The DESMOND programme has been previously validated and ensures generalisability of these findings across DESMOND programmes. Strengths and Limitations As this is an observational study, we cannot determine cause and effect. This study focused solely on clinical outcomes within the scope of the available data for reporting proposed DSME core outcome measures . However, scant empirical evidence exists regarding the impact of interventions spearheaded by dietitians in real‐world contexts. Consequently, we deemed it imperative to contribute to this body of knowledge, particularly concerning the pragmatic aspects of implementing DSME within primary care settings. In practical terms, DSME educators depend on both electronic patient lab systems and patient self‐reports of clinical outcomes. However, it is noteworthy that lab results frequently lack recent (within 3 months) biochemical data, particularly when they have not been requested at the point of referral to DSME in primary care. Hence, any changes seen cannot be solely attributed to the DESMOND education. It also needs to be considered that changes may have come about due to the participants being highly motivated individuals and opting to attend DSME. It should also be noted that this study took place prior to the Covid‐19 pandemic and prior to the widespread introduction of telemedicine. During the pandemic, DSME moved to online platforms. DSME provided via telemedicine has previously been shown to improve glycemic control, diabetes knowledge, and self‐care adherence behaviour . Future research should examine the effectiveness of telemedicine with regard to improving attendance at DSME sessions. An option for telemedicine may have improved attendance in this present study. Despite the number of participants attending DESMOND initially, only 66 attended the locally developed follow‐up session and had repeat clinical data to be included in this study. Future research should focus on reasons for non‐attendance to DSME and in particular the HCP role in helping participants to understand the importance of DSME as part of treatment plans in integrated care models . Due to demands on the local dietetic service, this study did not have the staffing resources available to collect additional measurements of body composition (e.g., waist circumference, waist‐to‐hip ratio); future research should also evaluate the changes in these anthropometric measurements alongside potential changes in HbA1c. Notably, the influence of gender on attendance rates and trajectory of type 2 diabetes mellitus cannot be ignored. Previous research conducted in this area has shown that in 365 individuals with newly diagnosed type 2 diabetes mellitus, attendance at structured education was independently associated with female gender (Odds Ratio (OR) 1.28, 95% CI 1.05–1.46), lower HbA1c (OR 0.98 mmol/mol 95% CI 0.97–0.99) and non‐smoker status (OR 1.36, 95% CI 1.07–1.55) . Therefore, future research should examine ways to improve attendance for males. Regardless, this is an evaluation of the effectiveness of the Midwest DSME impact on clinical markers in participants who completed the programme. In terms of strengths, this sample was representative of individuals in Ireland with type 2 diabetes mellitus. The DESMOND programme has been previously validated and ensures generalisability of these findings across DESMOND programmes. Conclusion In a sample cohort of adults with ongoing and newly diagnosed type 2 diabetes mellitus, completion of the DESMOND programme resulted in sustained benefits in HbA1c and BMI clinical outcomes at 6‐month follow‐up. The importance of DSME for type 2 diabetes mellitus is now recognised, and this is reflected in national and international policies . As the rates of type 2 diabetes mellitus continue to rise, further research is required to determine optimal contact time and frequency of sessions required in order to sustain the observed improvement in clinical outcomes. C.S.: formal analysis, methodology, writing – original draft, writing – review and editing. T.O.: conceptualization, data curation, formal analysis, methodology, writing – review and editing. A.G.: Conceptualization, formal analysis, methodology, writing – review and editing. The authors declare no conflicts of interest. |
Modulation of Voltage-Gating and Hysteresis of Lysenin Channels by Cu | 3c6d75ad-4533-4edb-b583-6b57eb6070df | 10455686 | Physiology[mh] | Lysenin is a protein extracted from the coelomic fluid of the earthworm E. fetida , which inserts large β-barrel pores in artificial and natural lipid membranes . The intricate, multi-step mechanism of pore formation implies binding to sphingomyelin (SM), oligomerization into prepores, and prepore-to-pore conversion . The large diameter of the conducting pathway of the pore (~3 nm, ) leads to uncontrolled leakage of ions and molecules; this strong lytic activity dubbed lysenin as a pore-forming toxin (PFT) , although its physiological role is yet to be deciphered. Reconstituted lysenin channels share features specific to ion channels such as large transport rate, regulation, and selectivity . The intrinsic regulatory mechanisms manifest by adjustments of the channel’s conformation and conductance in response to stimuli of physical and chemical origin . Lysenin channels reconstituted in artificial membrane systems comprising anionic lipids present voltage-induced gating at low positive voltages , but this feature vanishes when the channels are reconstituted in neutral membranes . When exposed to multivalent cations, lysenin channels present ligand-induced gating in both charged and neutral lipid membranes . A salient feature of lysenin channels is the prominent hysteresis in conductance and bistability, indicative of molecular memory . Hysteresis manifests when the response to a particular stimulation is not fixed and depends on the history of the system . All voltage-gated ion channels characterized by two states (i.e., open and closed) may present a dynamic hysteresis in conductance, resulting from the slow equilibration of the channels in response to periodic voltage stimuli . However, other intrinsic mechanisms may come into play and endow ion channels with a more persistent hysteresis . Any hysteretic behavior leads to bistability and acquisition of memory emerging from the history-dependent behavior , which paves the way for gaining novel functionalities originating in the non-Markovian distribution of the states explorable by the system under investigations. In the case of lysenin, the observed hysteresis is not dynamic, and seemingly originates in an invariant reopening pathway of the channels previously closed by applied positive voltages . Notably, the reactivation pathway is also temperature-independent, while the inactivation pathway is strongly modulated by temperature variations . This memristor-like behavior manifests at time scales that greatly exceed the hysteresis observed for ion channels . The hysteresis in conductance of lysenin does not vanish when the period of the oscillatory voltage stimulus is much larger than the relaxation time of the channels , which would be the hallmark of dynamic hysteresis . The hysteretic behavior of lysenin channels is intimately linked to two important regulatory mechanisms, unique among PFTs, namely, voltage-induced gating and ligand-induced gating. To date, no reasonable attempt to advance a realistic model of lysenin regulation has been made, and this is justified by the multiple challenges accompanying such an attempt. A realistic model of gating must account for all regulatory features investigated so far regarding voltage gating, ligand gating, and hysteretic behavior. However, the regulatory response to physical and chemical factors is far from uniform. K + ions and higher pH lead to a right shift in the voltage-induced gating, suggesting modulation of voltage regulation by electrostatic interactions . The same monovalent cations adjust the hysteresis in conductance by shifting the voltage needed to close the channels to higher values, yet the reactivation pathway is rather invariant . Multivalent ions present a more intricate interaction with lysenin channels, dominated by the resulting ligand-induced gating . Many trivalent metal cations (lanthanides, Al 3+ , Fe 3+ ) force the individual channels to adopt a fully closed state, while divalent metal cations (Ca 2+ , Mg 2+ , etc.) induce stable conformational changes characterized as sub-conducting states . In most cases, the inhibitory effects are suppressed upon ligand removal by precipitation or chelation, indicative of reversibility. However, Cr 3+ ions present a distinct inhibition pattern, suggesting cooperativity, and the changes are not reversible . In the same line of intricacy, voluminous organic cations (spermidine, spermine) present an inhibition profile resembling divalent metal cations , suggesting that charge density as opposed to charge alone influences the pathway adopted for conformational changes (i.e., full closing, as opposed to sub-conducting). La 3+ ions, at concentrations sufficiently small to render ligand-induced gating negligible, influence hysteresis similarly to monovalent cations: a right shift of the voltage gating and open probability during ascending voltage ramps, and a rather invariant reopening pathway during descending voltage ramps . The interaction with anionic ATP adds another level of intricacy: the conductance is inhibited in a concentration-dependent manner, yet voltage-gating and open probability show a strong right shift for both ascending and descending voltage ramps . These prior investigations are indicative of major hurdles in elaborating a realistic model of gating, yet they may provide clues required for a better understanding of regulatory mechanisms and hysteretic response. We attempted to explain gating and hysteretic behavior by considering electrostatic screening of a voltage-domain sensor moving into the hydrophobic core of the membrane , but no strong evidence for such a mechanism exists. Therefore, any additional clue on lysenin’s functionality modulation by physical or chemical cues may contribute to the development of a realistic model of regulation. To gain new knowledge on the functionality of lysenin channels, we employed electrophysiology approaches to assess the effect of Cu 2+ ions on the voltage regulation and hysteretic behavior. Unlike other multivalent metals that force the channels to close in a single step or to adopt a stable, sub-conducting state , Cu 2+ ions interact with lysenin and lead to full closing in two steps . To verify if this distinct interaction between lysenin channels and Cu 2+ ions influences the hysteresis in conductance, we employed electrophysiology measurements and concluded that the addition of small amounts of Cu 2+ ions not only significantly enhances the hysteretic behavior but also enables its persistent manifestation at zero bias voltage. This result prompted us to further investigate the effect of Cu 2+ ions on the voltage-gating of lysenin channels reconstituted in neutral membranes, for which such feature is abrogated . Surprisingly, Cu 2+ addition led to a full restoration of the voltage-induced gating of lysenin channels reconstituted in neutral membranes, which also manifested a strong hysteresis in conductance. Our experiments were initiated by reconstituting lysenin channels in a planar Bilayer Lipid Membrane (BLM) composed of Asolectin (Aso), Sphingomyelin (SM), and Cholesterol (Chol), bathed by electrolyte solutions (50 mM KCl, 20 mM Hepes, pH 7.2). We opted for a low electrolyte concentration for several reasons: lysenin channels gate at lower voltages , a low conductivity solution enables measuring ionic currents through very large populations of reconstituted channels , and a lower ionic strength may enhance the electrostatic interactions by reducing screening. Channel insertion was monitored at −60 mV bias potential, and the insertion of individual channels was inferred from the stepwise variation of the ionic currents . Each inserted channel led to a variation of the ionic current by ~20 pA, consistent with previous experiments carried out in similar solution and electrical conditions ; the resulting conductance (~0.33 nS/channel) was used to provide a rough yet realistic estimation of the number of reconstituted channels for each of the experiments from the slope of the linear portion of the I-V plots or the macroscopic currents measured at a voltage at which all the channels are fully conducting. The next investigations focused on recording the response of a population of ~1300 lysenin channels to a slow oscillatory voltage stimulus in the positive voltage range with and without addition of Cu 2+ ions. For this task, we measured the ionic currents in response to linearly variable voltage ramps (ascending and descending) ranging from 0 mV to +60 mV with a period of 20 min at a sampling rate of 1 Hz, a protocol customarily used to assess the hysteretic behavior of lysenin’s conductance . In the absence of Cu 2+ ions (control experiment), the lysenin channels showed the typical hysteretic behavior ( a). During the ascending voltage ramp, the currents increased linearly with voltage until ~20 mV, after which the channels started to close, and the ionic currents became negligible at voltages greater than 45 mV. The closed channels started reopening at lower voltages during the descending voltage ramp, and full reopening occurred at voltages under 10 mV. Given the known influence presented by monovalent and trivalent metal cations on the hysteretic conductance of lysenin channels , we expected Cu 2+ ions to elicit a right shift in the voltage gating during the ascending voltage ramp, and an invariant reopening pathway during the descending voltage ramp. However, the I-V plot recorded in response to otherwise identical voltage ramps after Cu 2+ addition was quite different ( a): the channels started closing at a much lower voltage (<10 mV) during the ascending voltage ramp, and minimal reopening occurred during the descending voltage ramp. The hysteresis in conductance persisted but the channels resisted reopening after Cu 2+ addition. A consistent overlap of the I-V plots was observed for three consecutive runs within the same experiment in the absence of Cu 2+ ions. However, the I-V plots recorded after Cu 2+ addition changed significantly following the first run: the closed channels did not reopen after the first run, and only negligible ionic currents were recorded for both ascending and descending voltage ramps ( b). The attempt to record data from independent experiments proved futile, owing to difficulties to replicate the exact same number of inserted lysenin channels in independent experiments. The insertion process is not easily controlled, and the number of inserted channels may differ substantially between experiments carried out in otherwise identical experimental conditions, which would make a statistical analysis of I-V plots meaningless. One may argue that a normalized value, like the open probability, would be suitable for statistical analyses of data from independent experiments; this is not necessarily true, since it was shown that the open probability may depend on the number of inserted channels . Consequently, our results represent typical, single traces recorded in individual experiments. However, each set of paired data showing the effects of Cu 2+ on a particular feature comprised the same membrane, with measurements taken before and after Cu 2+ addition. This limitation is not uncommon: single traces are reported for similar electrophysiology experiments in which the number of channels cannot be precisely controlled , including plots of normalized quantities. For a better understanding of the effects presented by Cu 2+ ions on gating, we repeated the experiments by extending the voltage range from −80 mV to +80 mV and the ramp period to 30 min for a population of ~950 channels. In the absence of Cu 2+ ions, the I-V plot recorded in response to the ascending voltage ramp showed the typical response of lysenin channels to external voltages . From −80 mV to +20 mV, the ohmic relationship between macroscopic currents and voltages indicated that the channels remained in the open state ( a). The channels began to close at voltages exceeding +20 mV; this behavior continued up to the maximum applied voltage of +80 mV. The steep decrease in the macroscopic currents at voltages over +30 mV indicated voltage gating and sustained channel closure; the effectiveness of voltage gating at positive potentials was inferred from the very small ionic currents recorded at voltages greater than +50 mV. After nearly all the channels were closed by the large depolarizing voltage, the application of a descending voltage (from +80 mV to −80 mV) led to the observation of hysteresis in conductance ( a). The reopening of the lysenin channels followed a different pathway: the channels showed a preference for the closed state for a larger voltage range and fully reopened at lower positive voltages. At descending voltages under ~+8 mV, the linear I-V plot indicated an ohmic behavior identical to the I-V plot recorded for ascending voltages, confirming a complete reopening of the channels. The observations inferred from the I-V curves recorded in the absence of Cu 2+ ions were confirmed by plots of the open probability (P open ) determined in the positive voltage range for ascending and descending ramps ( b). To calculate the P open value, we used the ratio between the measured currents and the theoretical maximal currents estimated for the same population open channels assumed in the open state at each applied voltage . These theoretical currents were approximated from the slopes of the linear portion of the response to ascending voltages . Since a P open value of one was obvious at negative voltages and to avoid the division by zero around the origin, the P open was plotted only for positive voltages larger than 1.5 mV. During the ascending voltage ramp, the P open equaled one at low positive voltages, started decreasing as the voltage increased, and approached zero at transmembrane voltages larger than 50 mV. However, for the descending voltage ramp, the plot was shifted to the left, and a full reopening occurred at voltages of a few mV. The midway voltage of activation V 0.5 (i.e., the voltage at which P open = 0.5) for ascending ramps was estimated at ~32 mV, while the descending ramps indicated a smaller V 0.5 value of ~15 mV, which is consistent with earlier reports on the hysteresis and bistability of lysenin channels subjected to slow oscillatory voltages . Substantial qualitative and quantitative differences were observed when the experiment was repeated in the presence of 4 µM Cu 2+ ions added to both sides of the membrane ( c). As a consequence of the ligand-induced gating presented by the interactions between lysenin channels and Cu 2+ ions , the recorded ionic currents were lower for the entire voltage range. The I-V plot recorded during ascending voltage ramps was linear from −80 mV to ~+8 mV, which is indicative of an ohmic behavior and the absence of voltage-induced gating within this range. We concluded that the Cu 2+ addition adjusted the voltage at which the channels started to close, and gating occurred at much smaller voltages than in the absence of Cu 2+ ions. This is not the typical behavior of lysenin channels exposed to monovalent and multivalent metal cations, for which the addition induces a rightward shift in the voltage gating behavior during ascending voltages . The channels practically closed at +20 mV and remained in this state for applied voltages up to +80 mV. A peculiar behavior regarding the effects of added Cu 2+ ions was observed during the descending voltage ramps. The ionic currents measured from +80 mV to 0 mV were very small, indicating that lysenin channels resisted opening at any positive voltage. This is also contrasted with the typical behavior of lysenin channels exposed to multivalent metal cations, in which case the reactivation pathway, although indicative of hysteresis, is invariant and unaffected by ion additions . The addition of Cu 2+ ions not only elicited an early closing during ascending voltage ramps but also forced the voltage-closed channels to stay in that state for the entire range of positive potentials during descending voltage ramps. In addition, it seems like the channels remained closed even at negative voltages, and sustained reopening began at voltages under −10 mV. The nonlinear shape of the descending I-V plot recorded at negative voltages together with the consistently lower values of the ionic currents suggested that the reopening of the channels continued at negative potentials and that not all the channels reopened within the timeframe of the experiment. Since the channels did not fully reopen during the descending voltage ramp, we used the linear portion of the I-V plot recorded during the ascending voltage ramp as reference to determine P open since all the channels, except the ones closed due to ligand gating, were in the open state at negative voltages. The analysis of the P open at positive voltages ( d) confirmed our observations inferred from the I-V plots. The midway voltage of activation V 0.5 for the ascending voltage ramp was estimated at ~+15 mV; however, the P open measured for the descending voltage ramp was nearly zero for the entire range of positive voltages. The experimental investigations of the effects of Cu 2+ ions presented above showed major changes in the hysteresis and bistability of lysenin channels, indicative of memory capabilities. Nonetheless, an important remaining question was the behavior of lysenin channels in the absence of any applied voltage (i.e., 0 mV). The I-V and P open plots for ascending ramps suggested a full opening when 0 mV was applied to channels that had previously been in an open state. However, for descending voltage ramps, the plots suggested that at 0 mV, the channels would remain in the closed state if their prior state was closed. To test this assumption, we employed the AC/DC setup presented in the and determined the status of the channels regarding various voltages, prior electrical conditions, and conducting states. The resulting recording, taken of a large population of channels without Cu 2+ ions and exposed to simultaneous AC/DC excitation, is shown in for each voltage condition. Before starting the experiment, the channels were biased for a few minutes at −60 mV to ensure that all channels were in the open state. The recording started immediately upon the application of 0 mV DC. In these conditions, the amplitude of the AC current through the ~2400 open channels was ~790 pA . The manual application of a +60 mV step voltage gradually reduced the amplitude of the AC signal down to ~375 pA, indicating a reduction in the macroscopic conductance due to channel closing induced by the applied voltage. After a few minutes, the reapplication of 0 mV led to channel reopening, which was observed as a quick increase in the amplitude of the AC current. In less than one minute, the amplitude of the AC signal recovered the initial value (~790 pA), indicative of a quick reinstatement of the original conductance and full channel reopening. This experiment suggested that, in the absence of Cu 2+ ions, the channels were open at 0 mV, started closing at positive voltages, and rapidly re-instated their initial conductance after removal of the DC voltage stimulus. A few important observations can be made from a further analysis of the channel conductance estimated from the trace shown in . Firstly, the signal to noise ratio recorded at 0 mV before and after channel closure seemed much smaller than the one recorded during the application of the +60 mV step voltage. The larger value of the electrical noise while channels were closing most likely originated in the conformational open-close fluctuations manifested during voltage application. At +60 mV, the amplitude of the AC current decreased exponentially (as anticipated) but owing to the large volume of data (sampling frequency 500 Hz), we opted not to wait until a constant amplitude was achieved (i.e., indicative of steady state). Regardless, one may reasonably estimate that many of the channels were still in the open conformation, which was not observed in the I-V and P open plots shown in and . The explanation for this behavior is related to the high density of the channels in the target membrane . To successfully complete these experiments and obtain a reasonable amplitude of the AC current for a 1 mV stimulation, we needed to reconstitute more channels in the target membrane. However, given the propensity of lysenin to prefer oligomerization into lipid rafts , high local channel densities may be achieved with only a few thousand channels, which may lead to adjustments of voltage-induced gating and the prevention of full closing even at high positive voltages . Our next experiment addressed the influence of Cu 2+ ions on lysenin channel’s opening, reopening, and bistability in response to applied voltages while accounting for the history of the channels. To do this, we utilized the same experimental system and used the AC currents to determine the status of the channels exposed to 4 µM Cu 2+ and additional DC voltages . Before recording, the membrane was biased by −60 mV for several minutes to ensure that all the channels were in the open state. The application of the AC signal at 0 mV DC voltage indicated an ionic current amplitude of ~710 pA. The application of +60 mV DC to the membrane led to a fast decline of the AC current amplitude to ~35 pA. After channel closure, we monitored the amplitude of the AC current at 0 mV bias voltage. As the experiment utilized a fast-sampling rate, we recorded a large amount of data. To mitigate potential problems with the acquisition system and further data analysis, we temporarily paused the recording during the experiment while maintaining all other electrical and solution conditions. As shows, a very small number of the lysenin channels previously closed by the application of the +60 mV potential reopened after removing the voltage stimulus. At 0 mV, the previously closed channels maintained their state for an extended time, as seen by the notably low AC current amplitude, which was monitored for more than 30 min. The small, stable amplitude of the AC current suggested the channels exposed to Cu 2+ ions and previously closed by voltage remained in the closed state for a remarkably long time, if not indefinitely, at a 0 mV bias potential. To verify if the channels were permanently closed (which would be in contradiction with the results shown in the I-V plots), we applied a −60 mV transmembrane voltage for about one minute, which elicited a fast increase in the AC currents, which was indicative of channel reopening. The same maximum amplitude of the AC current was recorded after a consequent application of 0 mV; the recovery of the amplitude of the AC current to the same value we recorded at 0 mV before closing the channels suggested that the channels fully reopened upon the application of the hyperpolarization voltage. The power spectrum for each portion of the trace recorded at a particular voltage and channel conformation indicated the presence of the 10 Hz AC signal . The experiments indicated that Cu 2+ ions elicit not only major changes in hysteresis but also adjustments of the voltage-elicited response: faster closing, slower reopening, and a left-shifted midway voltage of activation. This is opposite to what is known of the influence of monovalent and trivalent metal cations on lysenin channels, which usually manifests by a rightward shift of the midway voltage of activation and slow down the response to positive voltage stimuli in response to ascending voltage ramps . These discrepancies suggest Cu 2+ may present a very different interaction with lysenin channels and act by promoting channel closure even when this feature would be weakened or suppressed by experimental conditions. In this line, lysenin channels are well known for losing their voltage-induced gating upon reconstitution in bilayer lipid membranes composed of electrically neutral lipids . To investigate the influence of Cu 2+ ions on the voltage gating of lysenin channels reconstituted in neutral bilayers, we employed support lipid membranes in which we replaced the anionic Aso with the neutral Diphytanoil-Phosphatidylcholine (DiPhyt-PC) as the major membrane component . In these experimental conditions, the suppression of voltage gating for the ~2000 inserted channels was seen in the quasi-linear I-V plot recorded from −60 mV to +60 mV . The linear I-V plot constructed for lysenin channels reconstituted in neutral membranes and in the absence of Cu 2+ ions indicated that no voltage-induced gating manifested for the entire voltage range employed in this experiment, confirming prior investigations . After the addition of Cu 2+ to both sides (4 μM final concentration in each reservoir), the lysenin channels maintained the ohmic behavior for negative voltages and up to ~+10 mV. Surprisingly, the ionic currents decreased significantly for larger positive voltages (exceeding +10 mV), indicative of channel closing by voltage gating. The small values of the ionic currents measured at applied voltages exceeding 20 mV implied that the channels closed completely, suggesting that the Cu 2+ addition restored the voltage gating properties. In addition, the P open determined for voltage ramps in the positive voltage range ( b) indicated a complete recovery of the hysteretic behavior, including the arresting of the channels in the closed state at positive voltages and the absence of reopening at 0 mV. Our findings raise more questions than provide answers with regard to deciphering the complex gating mechanisms presented by lysenin channels. Voltage gating and its absence in neutral membranes, as well as hysteresis, have been reported for more than two decades . Nonetheless, no real progress was encountered regarding the biophysical mechanisms by which they manifest, which makes a mechanistic interpretation of our data difficult. Although electrostatic interactions seem to play a major role in voltage gating (because it is suppressed in neutral membranes, and modulated by ionic strength), how exactly they lead to gating is not known. Our work did not bring any experimental evidence to indicate that the changes in voltage gating and hysteresis are a consequence of the interaction of Cu 2+ ions with lipids, proteins, or both. Cu 2+ ions lead to a leftward shift of the open probability for both ascending and descending voltage ramps, while other monovalent and multivalent metal cations (i.e., K + and La 3+ ) adjust the hysteresis by a rightward shift of the open probability during ascending voltage ramps and an invariant pathway for descending voltage ramps (channel reopening) . The current structural data do not provide sufficient evidence of an outside domain able to occlude the channel by a ball and chain mechanism, but such occurrence may not be fully excluded. While our findings may help in elucidating the mechanism of lysenin’s gating, substantial investigations are needed to attain this goal. On the other hand, we anticipate our findings to contribute to a better understanding of how such behavior may provide molecular memory capabilities to unicellular or even complex organisms. Lysenin is not an ion channel; as a matter of fact, it is not even a transmembrane protein in its native environment. Lysenin shares salient features of ion channels (i.e., high transport rate, regulation, and selectivity), and it is endowed with strong memristive capabilities. The memory originating in the kinetics of ion channels as well as the implications of memristive properties in learning and the establishment of important neural functions are under intense scrutiny for a better understanding of how molecular memory contributes to fundamental physiological processes . In addition to replicating fundamental features of ion channels, lysenin is much easier to work with, which makes it an excellent experimental model for exploring molecular memory phenomena and their consequences. The lipids used for these experiments were Asolectin (Aso, Sigma-Aldrich, St. Louis, MO, USA), Sphingomyelin (SM, Avanti Polar Lipids, Alabaster, AL, USA), Diphytanoil-Phosphatidylcholine (Diphyt-PC, Avanti Polar Lipids) and cholesterol (Chol, Sigma-Aldrich). The lipids, which were originally in powder form, were solubilized in n-decane (Fisher Scientific, Pittsburgh, PA, USA) to produce lipid mixtures of Aso/SM/Chol at a weight ratio of 10:5:4. Neutral bilayers were prepared by replacing Aso with DiPhyt-PC in the lipid mixture. The planar lipid membrane experimental setup , often used for electrophysiology investigations on lysenin channels , consisted of two insulating polytetrafluoroethylene (PTFE) reservoirs (each of ~1 mL volume) separated by a thin PTFE film (~120 µm thickness) in which a central hole of ~100 µm diameter was created by an electric spark. The reservoirs were filled with electrolyte solutions made of 50 mM KCl (Fisher Scientific) and 20 mM Hepes (pH 7.2, Sigma-Aldrich). For electrical connections, two Ag/AgCl electrodes embedded in salt bridges (2% low melting point agarose-Sigma Aldrich, dissolved in 1 M NaCl—Fisher Scientific) were inserted directly into the electrolyte solutions in the two reservoirs. The Ag/AgCl electrodes were wired to the headstage of the electrophysiology amplifier (Axopatch 200B, Molecular Devices, San Jose, CA, USA), which fed into the DigiData 1440A digitizer (Molecular Devices). The computer-controlled digitizer was used for online visualization, recording, and further analysis with the pClamp 10.7.0.3 software package (Molecular Devices). The recorded data were further analyzed and plotted with the Origin 8.5.1 (OriginLab, Northampton, MA, USA) software package. Membrane formation and stabilization was monitored by estimating the membrane capacitance from the capacitive current measured in response to a triangle-wave signal (provided from a Keithley 3390 function generator), and the seal was verified by applying a DC voltage to the membrane. After a stable membrane was formed (C > 65 pF, R > 100 GΩ), we proceeded with channel insertion. Small amounts of recombinant lysenin were added to the grounded reservoir under continuous stirring (Warner Instruments low noise magnetic stirrer) and upon application of −60 mV bias potential (manual command). Channel insertion was recorded at 10 Hz sampling frequency and monitored from the stepwise variation of the ionic currents measured at constant voltage . After achieving a steady state of the macroscopic current in ~2 h, we proceeded with electrophysiology measurements before and after Cu 2+ additions. For this purpose, we used a 1 M CuSO 4 stock solution (Fisher Scientific) after proper serial dilution in buffered electrolyte solutions. After Cu 2+ addition to both sides of the membrane (4 µM final concentration for all experiments), we maintained stirring for several minutes to allow mixing and enable interactions between channels and ions. Linear voltage ramps were created with the digitizer by defining proper episodic stimulation protocols . For data recording, we used a 1 kHz low-pass hardware filter, 100 Hz low-pass software filter, and a sampling frequency of 1 Hz. To determine the status of the channels (i.e., open/closed) at any voltage (including zero voltage), we created an experimental setup that applied simultaneous AC and DC to the membrane. The DC voltage, utilized to control the status of the channels, was applied to the bilayer membrane with the manual command of the instrument, while the AC signals (10 Hz, 1 mV amplitude sinewave), utilized to estimate the status of the channels, were applied from the function generator via the external input of the electrophysiology amplifier . The unfiltered current included both AC and DC components, with each one dependent on the conductance of the channel population. The application of a high-pass 2 Hz filter suppressed the DC current component; the amplitude of the resulting AC current component was used to estimate the conductance status of the channels at any applied DC voltage. The amplitude of the AC signal was chosen to be very small (i.e., 1 mV) to prevent channel closing by voltage (which may not occur at very low positive voltages, or fast sweeps ), yet it was large enough to detect AC currents through open channels. Also, the low amplitude and frequency prevented attaining large values of the capacitive currents; for a 100 pF membrane, the amplitude of the capacitive current estimated for these experimental conditions would be less than 10 pA. To avoid excessive chopping or filtering of the AC signal, we utilized a sampling time of 2 msec, a 10 kHz low-pass hardware filter, and no other low-pass software filter. Our work showed that the addition of very small amounts of Cu 2+ ions to the bulk electrolyte solutions had multiple effects on the voltage gating profile of lysenin channels. Earlier reports showed that lysenin channels reconstituted into lipid membranes containing anionic lipids present a strong hysteresis in conductance manifested as a preference for the previously attained closed state. This hysteresis did not originate in the slow equilibration of the channels since it persisted for periods of the oscillatory stimulus that greatly exceeded the relaxation time of the channels. In this work, we utilized traditional I-V measurements together with an improved experimental setup and showed that Cu 2+ addition modifies the gating profile and substantially affects the response to oscillatory stimuli. Unlike other multivalent ions, Cu 2+ potentiated the voltage-induced gating and open-close channel transitions were attained at smaller potentials. These effects were more pronounced for the close–open transitions: during descending voltage ramps, the channels did not reopen even at 0 mV bias potential (such measurements were made possible by using a new experimental setup), and large hyperpolarizing voltages were needed to reinstate the open channel conductance. The history-dependent transitions between conformations led to bistability, caused a strong hysteresis in conductance, and endowed the channels with potential memory capabilities. An interesting fact is that this molecular memory can be fully addressed by employing electrical signals for writing, reading, and erasing it. Lysenin channels behave like true biological memristors but the molecular mechanisms by which such intricate functions are attained are yet to be deciphered. |
Diagnosis of visceral and cutaneous leishmaniasis using loop-mediated isothermal amplification (LAMP) protocols: a systematic review and meta-analysis | 8347d16b-e8b0-4f67-a4a2-44e4d9a9192a | 8785018 | Pathology[mh] | Leishmaniasis is a vector-borne disease caused by protozoan parasites of the genus Leishmania and transmitted by the females of phlebotomine sand flies . Factors such as proximity of animal reservoirs in the current model of peri-urban transmission, different susceptibilities of human populations and the environmental impact on vector distribution result in a complex interplay . There are various clinical manifestations, but a widely used classification differentiates between visceral leishmaniasis (VL), which is fatal if left untreated, cutaneous leishmaniasis (CL), mucocutaneous leishmaniasis (MCL), and a possible concurrent or late-term complication of VL which is called post kala-azar dermal leishmaniasis (PKDL) . Globally, there are about 12 million patients suffering from leishmaniasis, with more than 350 million people at risk in over 80 countries . The World Health Organization (WHO) estimates that around 0.7–1.0 million cases occur annually, 50,000 to 90,000 of which are VL cases and 0.6–1.0 million CL cases . Based on data from the Global Health Observatory data repository for 2018, 17,000 VL and 250,000 CL cases were reported to WHO by 53 countries. However, official numbers may be an underestimation for different reasons such as VL-related deaths outside of health care facilities . In addition, not all endemic countries reported data to WHO in 2018. Around 90% of VL cases occur in six countries: Bangladesh, Brazil, Ethiopia, India, South Sudan and Sudan. CL is distributed globally; the most affected countries are Afghanistan, Pakistan, Iran, Syria, Saudi Arabia, Algeria, Brazil, Colombia and Peru, with recent epidemics in Afghanistan and Syria . The majority of VL infections are caused by Leishmania donovani and Leishmania infantum . Several Leishmania species can cause CL; the most common causes of the infection are the species Leishmania major , Leishmania tropica , L. infantum (Mediterranean Basin, the Middle East, the Horn of Africa, Indian subcontinent), Leishmania aethiopica (in Ethiopia and Kenya), Leishmania braziliensis , Leishmania guyanensis (South America), and Leishmania mexicana (Mexico) . In some regions of the Southern Hemisphere, especially in South America, the areas endemic for Leishmania have been expanding in the recent past . In addition, due to climatic change, more habitats will become suitable for phlebotomine sand flies, resulting in a possible expansion of their geographic ranges and an establishment of endemic Leishmania transmission in more extreme latitudes throughout the world . Clinical symptoms of VL include fever, anaemia, leukopenia, hepatosplenomegaly, weight loss and diarrhoea. Most VL infections remain asymptomatic, but long incubation periods of up to 8 months are not uncommon, and symptomatic infections are often fatal if left untreated . The symptoms are similar to other diseases such as malaria and enteric fever, and a laboratory diagnosis is required for accurate diagnosis . Treatment recommendations for VL differ between regions but commonly used drugs are (liposomal) amphotericin B and pentavalent antimonials, both administered intravenously, or miltefosine, used orally . Therapeutic studies in the past focused mainly on monotherapy and the combination of existing drugs, but the Drugs for Neglected Diseases initiative (DND i ) has identified several candidates which might lead to innovative treatments for VL . The majority of CL cases manifest as chronic and normally painless skin lesions. These may heal spontaneously in response to development of cell-mediated immunity if untreated, although in most cases this process takes several months and up to years , with typically a low percentage of self-healing lesions for New World CL . Treatment of CL may include systemic therapy or local therapy such as heat or cryotherapy, topical creams (e.g. paromomycin) or intralesional injections of pentavalent antimonial derivatives . Lesions may leave disfiguring scars, possibly leading to stigmatization of recovered patients, having a long-term negative impact on psychological, social and economic well-being . In contrast to CL, MCL is potentially life-threatening if untreated. Ninety percent of MCL cases have a scar from a prior CL episode; depending on host cell-mediated immunity and parasite virulence, clinical progression to the mucosa may take place. Symptoms of an infection are progressive destruction of the oronasopharyngeal mucosa and cartilaginous facial and upper airway structures . The ratio of MCL to CL infections is low, and disease progression may be strongly dependent on the infecting species and possibly also on their infection with Leishmania RNA viruses . PKDL mostly occurs in eastern Africa and on the Indian subcontinent and is associated with a previous VL infection in most cases. It is manifested by mostly self-healing lesions which are only aesthetic problems in most infected individuals but are infectious to phlebotomine sand flies, possibly over decades . There are numerous different diagnostic test methods available for leishmaniasis, which can be divided into non-DNA-based and DNA-based methods . Among the non-DNA-based are serological methods detecting antibodies or antigens (such as proteins), and microscopic methods, which have long been regarded as the gold standard for VL and CL diagnosis . For VL diagnosis, the acquisition of tissue samples for microscopic methods is highly invasive, as spleen, lymph node or bone marrow aspirates are needed . For CL diagnosis, the sensitivity of microscopy is only moderate . Serological tests are less invasive and can be used in a near-PoC setting to support clinical VL diagnosis, as they generally have high sensitivity and low costs, and results can be determined in the field , but tests based on detection of antibodies largely cannot distinguish between current and past infections . Sensitivity is lower in immunocompromised individuals such as HIV-co-infected patients and in very young children . Furthermore, cross-reactivities are possible . Different from VL and partly also MCL, serological methods have low sensitivity in CL , as this disease usually only leads to a local immune response . Rapid diagnostic tests (RDTs) based on the detection of the rK39 antigen are widely used and reliable for diagnosis of VL . DNA-based test methods usually have high sensitivity and specificity, but require laboratory equipment such as a thermocycler and cold chain-kept reagents and are therefore difficult to implement in point-of-care (PoC) or near-PoC settings . In addition, laboratory staff need to be trained appropriately and there are concerns regarding the lack of standardization and quality control of molecular assays . However, they can also be applied to immunocompromised patients and, importantly, they do not require invasive sampling methods and can be performed with peripheral blood (VL) or lesion swab sampling (CL) . Polymerase chain reaction (PCR) and quantitative real-time PCR (qPCR) are among the most widely used DNA-based test methods . Nested PCR (LnPCR) increases the sensitivity in samples with low parasite density but is prone to contamination. Multiplex assays can detect several species (or species groups) at the same time but are also more expensive . Another promising molecular method for diagnosis of VL and CL is the loop-mediated isothermal amplification method (LAMP). LAMP uses a polymerase and typically four primers to amplify six target regions under isothermal conditions with high specificity. One of the inner forward and backward primers contains a complementary sequence which leads initially to a loop formation and in later amplification circles to dumbbell structures, forming continuously growing concatemers . LAMP has high specificity because amplification only occurs if all six target regions are correctly recognized by the primers . Since a large number of amplicons are produced and only a small quantity of sample is needed for successful amplification via LAMP, contamination of the workplace by amplicons of previous samples has been identified as a potential risk resulting in false-positive results . This risk can be reduced by using closed tubes which do not need to be opened to evaluate the result . Several methods for visual evaluation of amplification results have been developed. Pyrophosphate ions, which are reaction by-products, form a white precipitate with magnesium of the reaction buffer , and the addition of manganous ions and calcein leads to a visible colour change, enabling simple visual detection of positive samples without further equipment . SYBR Green, which is a DNA-binding dye that intercalates non-specifically into double-stranded DNA (dsDNA), can also be added to the tube initially blocked by a heat-sensitive capsule, as direct addition inhibits the amplification reaction . LAMP has been used in the diagnosis of a variety of diseases and detection of a whole spectrum of different pathogens in both humans and animals . LAMP has been established for various human pathogens, including Leishmania spp. , Trypanosoma brucei gambiense (human African trypanosomiasis) , Plasmodium falciparum (malaria) , Burkholderia pseudomallei (melioidosis) , Mycobacterium tuberculosis (tuberculosis) , Mycobacterium avium subsp. paratuberculosis (MAP, Johne's disease) and various Staphylococcus strains (food-borne infections) , among others. LAMP has also been used in combination with a reverse transcriptase enzyme (RT-LAMP) in order to amplify target RNA, making it a possible tool for detection of RNA viruses such as the Newcastle disease virus or SARS-CoV-2 (2019-nCoV) . RT-LAMP has been used to detect hepatitis B virus (hepatitis B) , H5N1 highly pathogenic avian influenza (HPAI, avian influenza) and classical swine fever virus (CSFV, swine fever) . To assess the performance of LAMP for CL and VL diagnoses, we conducted a systematic literature review, extracted data from eligible studies, and performed a qualitative and quantitative analysis, with a meta-analysis of selected datasets, to evaluate diagnostic test parameters compared to the well-established and commonly used reference standards microscopy and PCR-based methods (PCR, qPCR, LnPCR). Literature review protocol preparation The review protocol was registered in the International Prospective Register of Systematic Reviews (CRD42020150035) and can be accessed at https://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD42020150035 . Recommendations of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy and of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement were followed. Data sources and search strategy Structured searches were conducted by two reviewers on the PubMed and PubMed Central, Scopus, Web of Science, Cochrane Library, Embase, Epistemonikos and Global Index Medicus databases, using a comprehensive list of key terms including leishmania* AND (LAMP OR loop-mediated OR (isothermal AND amplification) but adapted to each database. Serological test methods were considered out of scope for the search strategy and the review overall, as they do not necessarily correlate with an active infection. A detailed description of the search strategy and search dates is available as supplementary information (see Additional file : Text S1). The initial search was complemented by a manual search of reference lists from retrieved articles and by citation tracking of review articles. If a study reported diagnostic performance values (e.g. specificity, sensitivity) but contained no individual sample data or information allowing completion of a 2 × 2 contingency table, further information was requested by mail from the corresponding and/or first author. If no further information was acquired, the respective study was included in the qualitative synthesis but not in the statistical analyses or in the meta-analysis. The literature search was conducted in July 2019 and repeated in July 2020 to include studies published up to the end of June 2020. Inclusion and exclusion criteria As inclusion criteria, studies were included if results for LAMP assays for diagnosis of leishmaniasis in clinical samples from humans or animals, with confirmation by microscopy, culture or molecular tests, were reported. No restrictions were made with respect to the publication language, date of publication or study design (consecutive or case–control) or data collection (prospective or retrospective). As exclusion criteria, studies were excluded in the case of lack of data regarding individual results reported, reference standard used or sample type. In addition, reviews and commentaries were excluded but references were analysed regarding potential further studies meeting the inclusion criteria. Selection process Deduplication of publications found in several databases was done manually and using Zotero 5.0.60 /5.0.84 . After removal of duplicates, each publication had its title and abstract reviewed based on the inclusion and exclusion criteria in a blinded manner by two independent reviewers, using Rayyan . After unblinding, discrepancies were resolved by discussion. In case an abstract did not contain enough information for rejection, the publication was automatically included for the full-text screening. Subsequently, the selected publications were read in full independently by both reviewers, either to confirm their eligibility and to extract the data or to exclude, again after unblinding and discussion with the second independent reviewer. Data extraction Data extraction was conducted by one reviewer and verified by a second reviewer based on a sample set of the included studies. We extracted data from primary studies to complete the four cell values of a diagnostic 2 × 2 table: true positives, false positives, true negatives, and false negatives. In addition, the following information was recorded: infecting species, sample type, reference test, LAMP target, country of patient’s origin, DNA extraction method, readout method of the LAMP and study design (consecutive or case–control). Study quality assessment The quality of included studies and risk of bias and applicability was assessed based on the QUADAS-2 tool . Data synthesis The accuracy measurements of interest for LAMP were sensitivity and specificity, which are defined as follows: sensitivity (S)—probability of a positive test in diseased individuals; specificity (E)—probability of a negative test in non-diseased individuals. In order to calculate S and E values for LAMP, we cross-tabulated each result against each one reference standard (microscopy and/or another molecular diagnostic method besides LAMP), stratified by each clinical condition (CL, VL or PKDL) and biological specimen used. Thus, for the same study, more than one analysis was possible: in general, each panel of samples extracted from a single study, tested with LAMP using the same sample type against the same reference standard test, was called “dataset”. For Schallig et al. and Vink et al. , two different datasets were created depending on the country where the panel of samples were analysed (see Additional file : Table S1, comments). Analysis Descriptive statistics as calculation of mean, median and test for normal distribution (Shapiro–Wilk) were also calculated in R version 3.6.2 . The accuracy measurements were calculated using R and the epiR package version 1.0.10 . For the subsequent meta-analysis, we were interested whether including studies with a sample size below 10 would introduce a bias and should be excluded, in line with previous publications . We therefore calculated Spearman’s rank correlation coefficient, including a 95% confidence interval (CI), in order to analyse the possible correlation between sample size and S or E using R. Forest plots showing S and E values for all datasets including a 95% CI were created using RevMan 5.3 . Subgroup 1 (“VL Microscopy LAMP: Blood” consisting of datasets 1, 8, 23, 38, 42, 44, 47, 59, 64), subgroup 2 (“VL PCR LAMP: Blood” consisting of datasets 11, 26, 30, 33, 35, 36, 43, 45, 48, 63, 65), subgroup 3 (“PKDL qPCR LAMP: Blood” consisting of datasets 28, 31 and 40), subgroup 4 (“CL Microscopy LAMP: Skin tissue” consisting of datasets 6, 52, 57, 60, 61, 80) and subgroup 5 (“CL PCR LAMP: Skin Tissue” consisting of datasets 4, 7, 29, 46, 53, 58, 62, 56) were used. For each subgroup of interest, diagnostic test results per patient tested were included more than once only if multiple samples of the same patients were taken at different time points (datasets 30, 31, 46), such as before and after treatment (at follow-up). For several included studies more than one diagnostic test result per patient was available, for example due to multiple LAMP tests with different primer pairs of the same patient sample set. The decision as to which datasets were included was based mainly on the aim of combining similar studies in the subgroups (e.g. same sample type). In addition, arbitrary reasons, such as which datasets best reflected target conditions, determined the choice of datasets. For example, an analysis of a panel of patient samples was conducted in both Suriname (datasets 52 and 53) and the Netherlands (datasets 54 and 55), but only datasets of the endemic country (Suriname) were used for the subgroup analysis. For different primer pairs (datasets 32 and 33) targeting the internal transcribed spacer 1 ( ITS1) sequence, the dataset with higher sensitivity was included for analysis. Pooled estimates for S and E of subgroups, I 2 and Tau-squared parameters were calculated using Comprehensive Meta-Analysis version 3.3.070 . Summary receiver operating characteristic (SROC) curves and area under the curve (AUC) of subgroups 1, 2, 4 and 5 were calculated using R using the mada package version 0.5.10 which is based on a bivariate random-effects model . For studies with 2 × 2 tables that contain entries of the value 0, in accordance with the package manual, continuity correction based on Haldane and Ascombe of adding 0.5 to all values of the affected tables was used . The review protocol was registered in the International Prospective Register of Systematic Reviews (CRD42020150035) and can be accessed at https://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD42020150035 . Recommendations of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy and of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement were followed. Structured searches were conducted by two reviewers on the PubMed and PubMed Central, Scopus, Web of Science, Cochrane Library, Embase, Epistemonikos and Global Index Medicus databases, using a comprehensive list of key terms including leishmania* AND (LAMP OR loop-mediated OR (isothermal AND amplification) but adapted to each database. Serological test methods were considered out of scope for the search strategy and the review overall, as they do not necessarily correlate with an active infection. A detailed description of the search strategy and search dates is available as supplementary information (see Additional file : Text S1). The initial search was complemented by a manual search of reference lists from retrieved articles and by citation tracking of review articles. If a study reported diagnostic performance values (e.g. specificity, sensitivity) but contained no individual sample data or information allowing completion of a 2 × 2 contingency table, further information was requested by mail from the corresponding and/or first author. If no further information was acquired, the respective study was included in the qualitative synthesis but not in the statistical analyses or in the meta-analysis. The literature search was conducted in July 2019 and repeated in July 2020 to include studies published up to the end of June 2020. As inclusion criteria, studies were included if results for LAMP assays for diagnosis of leishmaniasis in clinical samples from humans or animals, with confirmation by microscopy, culture or molecular tests, were reported. No restrictions were made with respect to the publication language, date of publication or study design (consecutive or case–control) or data collection (prospective or retrospective). As exclusion criteria, studies were excluded in the case of lack of data regarding individual results reported, reference standard used or sample type. In addition, reviews and commentaries were excluded but references were analysed regarding potential further studies meeting the inclusion criteria. Deduplication of publications found in several databases was done manually and using Zotero 5.0.60 /5.0.84 . After removal of duplicates, each publication had its title and abstract reviewed based on the inclusion and exclusion criteria in a blinded manner by two independent reviewers, using Rayyan . After unblinding, discrepancies were resolved by discussion. In case an abstract did not contain enough information for rejection, the publication was automatically included for the full-text screening. Subsequently, the selected publications were read in full independently by both reviewers, either to confirm their eligibility and to extract the data or to exclude, again after unblinding and discussion with the second independent reviewer. Data extraction was conducted by one reviewer and verified by a second reviewer based on a sample set of the included studies. We extracted data from primary studies to complete the four cell values of a diagnostic 2 × 2 table: true positives, false positives, true negatives, and false negatives. In addition, the following information was recorded: infecting species, sample type, reference test, LAMP target, country of patient’s origin, DNA extraction method, readout method of the LAMP and study design (consecutive or case–control). The quality of included studies and risk of bias and applicability was assessed based on the QUADAS-2 tool . The accuracy measurements of interest for LAMP were sensitivity and specificity, which are defined as follows: sensitivity (S)—probability of a positive test in diseased individuals; specificity (E)—probability of a negative test in non-diseased individuals. In order to calculate S and E values for LAMP, we cross-tabulated each result against each one reference standard (microscopy and/or another molecular diagnostic method besides LAMP), stratified by each clinical condition (CL, VL or PKDL) and biological specimen used. Thus, for the same study, more than one analysis was possible: in general, each panel of samples extracted from a single study, tested with LAMP using the same sample type against the same reference standard test, was called “dataset”. For Schallig et al. and Vink et al. , two different datasets were created depending on the country where the panel of samples were analysed (see Additional file : Table S1, comments). Descriptive statistics as calculation of mean, median and test for normal distribution (Shapiro–Wilk) were also calculated in R version 3.6.2 . The accuracy measurements were calculated using R and the epiR package version 1.0.10 . For the subsequent meta-analysis, we were interested whether including studies with a sample size below 10 would introduce a bias and should be excluded, in line with previous publications . We therefore calculated Spearman’s rank correlation coefficient, including a 95% confidence interval (CI), in order to analyse the possible correlation between sample size and S or E using R. Forest plots showing S and E values for all datasets including a 95% CI were created using RevMan 5.3 . Subgroup 1 (“VL Microscopy LAMP: Blood” consisting of datasets 1, 8, 23, 38, 42, 44, 47, 59, 64), subgroup 2 (“VL PCR LAMP: Blood” consisting of datasets 11, 26, 30, 33, 35, 36, 43, 45, 48, 63, 65), subgroup 3 (“PKDL qPCR LAMP: Blood” consisting of datasets 28, 31 and 40), subgroup 4 (“CL Microscopy LAMP: Skin tissue” consisting of datasets 6, 52, 57, 60, 61, 80) and subgroup 5 (“CL PCR LAMP: Skin Tissue” consisting of datasets 4, 7, 29, 46, 53, 58, 62, 56) were used. For each subgroup of interest, diagnostic test results per patient tested were included more than once only if multiple samples of the same patients were taken at different time points (datasets 30, 31, 46), such as before and after treatment (at follow-up). For several included studies more than one diagnostic test result per patient was available, for example due to multiple LAMP tests with different primer pairs of the same patient sample set. The decision as to which datasets were included was based mainly on the aim of combining similar studies in the subgroups (e.g. same sample type). In addition, arbitrary reasons, such as which datasets best reflected target conditions, determined the choice of datasets. For example, an analysis of a panel of patient samples was conducted in both Suriname (datasets 52 and 53) and the Netherlands (datasets 54 and 55), but only datasets of the endemic country (Suriname) were used for the subgroup analysis. For different primer pairs (datasets 32 and 33) targeting the internal transcribed spacer 1 ( ITS1) sequence, the dataset with higher sensitivity was included for analysis. Pooled estimates for S and E of subgroups, I 2 and Tau-squared parameters were calculated using Comprehensive Meta-Analysis version 3.3.070 . Summary receiver operating characteristic (SROC) curves and area under the curve (AUC) of subgroups 1, 2, 4 and 5 were calculated using R using the mada package version 0.5.10 which is based on a bivariate random-effects model . For studies with 2 × 2 tables that contain entries of the value 0, in accordance with the package manual, continuity correction based on Haldane and Ascombe of adding 0.5 to all values of the affected tables was used . Literature search The full workflow of the literature search, based on the principles of the PRISMA guidelines for systematic reviews , is shown in Fig. . A total of 394 publications were retrieved; after deduplication, 228 publications were screened by title and abstract and 50 by full text based on the inclusion/exclusion criteria as detailed in the “ ” section. Studies were excluded at the title/abstract screening stage for the following reasons: wrong pathogen (condition under investigation of the study was not caused by Leishmania sp.), no LAMP (LAMP was not used as a diagnostic test method) or wrong article type (reviews and commentaries were excluded, but references were screened for further studies). Studies were excluded at the full-text assessment stage for the following reasons: lack of data (inability to complete a 2 × 2 contingency table), duplicate study (the same clinical data were described in another study), no LAMP (LAMP was not used as a diagnostic test method), no paired samples (samples tested with LAMP and the reference standard were not from the same individuals) or promastigote form (test samples were derived from the promastigote form). Twenty-seven studies were accepted for further analysis and dataset extraction—22 studies regarding leishmaniasis diagnosis in humans (Additional file : Table S2) and five studies addressing diagnosis in animals (Table ). For the extracted variables infecting species and readout methods , we used ( indicated ) if not mentioned directly in the text. For the infecting species, this refers to identification through for example the use of specific primer pairs or epidemiological data without confirmation by further analysis. For the readout method, this refers to identification through specific reagents/kits used. Characteristics of included studies The datasets were stratified by clinical condition, sample type and reference test used. A full list of datasets per study is available as supplementary information (see Additional file : Table S1). Eighty-one and 12 datasets were constructed based on the included studies for LAMP diagnosis in humans and animals, respectively. In the case of missing data for completion of a 2 × 2 contingency table, or a need for clarification, the corresponding and/or first authors of 13 publications were contacted, enabling seven additional datasets to be constructed. The following descriptions are based on the included human studies, where the following studies are counted more than once as different indications are analysed: Adams et al. two studies (VL and CL), Verma et al. three studies (VL, PKDL and CL), Verma et al. two studies (VL, PKDL), and Sriworarat et al. two studies (VL and CL), resulting in 27 studies in total. In total, 2255 individuals and 6159 test results for diagnosis of leishmaniasis in humans are included in this review. Of the individual tests, 1453 are for diagnosis of VL and 650 of CL. The studies were performed from 2009 to 2019, and about half of them ( n = 14) during the past 4 years (2017–2020). Out of 27 studies, 21 (78%) evaluated the LAMP performance in the Old World, while four studies evaluated the LAMP performance in New World countries (Brazil, Colombia and Suriname) , and one study included a travel case from Venezuela . For two studies the origin of patients is not mentioned. Eighteen studies (67%) used a control group while nine (33%) were categorized as consecutive. Two studies included analysed LAMP performance in PKDL diagnosis. Twenty-three studies (85%) used a commercial kit for DNA extraction; in seven (26%) the kit used was the QIAamp ® DNA Blood mini kit (QIAGEN, Hilden, Germany), and six (22%) used a commercial kit for LAMP, which was the Loopamp™ Leishmania detection kit (Eiken Chemical, Tokyo, Japan). In 12 cases. L. donovani was found or indicated (e.g. through usage of species-specific primer pairs) as the infecting species, L. tropica was found in three studies, and L. infantum , L. major and L. guyanensis were found or indicated in two studies each. In 11 studies (40%) the target was kinetoplast DNA (kDNA); in seven (26%) the targets for LAMP were a combination of 18S ribosomal RNA (rRNA) and kDNA genes. The cysteine proteinase b ( cpb ) gene, ITS1 DNA sequences and k26 were used in one study each as the targets. In 23 studies (85%) a PCR method (PCR, qPCR or LnPCR) was used as a reference standard, and in 21 (78%) a microscopy method (microscopy or culture microscopy) was used as a reference standard. The sample size of the 27 included studies ranges from two to 274, with a median of 72 and an interquartile range from 38 (25th percentile) to 95.5 (75th percentile). QUADAS-2 based quality assessment The quality of included studies was analysed based on the QUADAS-2 tool ; the results separated by VL and CL diagnosis studies are shown as supplementary information (see Additional file : Figure S1). The risk regarding applicability of (1) reference standard, (2) index test and (3) patient selection were judged as low for the included studies. For index and reference test, the risk of bias is unclear in most included studies, with some having a high risk of bias regarding the categories flow and timing and patient selection. Performance of LAMP for the diagnosis of leishmaniasis The forest plots for the S and E of LAMP vs the reference test per dataset are given as supplementary information (see Additional file : Figure S2). Spearman’s rank correlation coefficient evaluating the correlation between S and sample size is r s (S, n ) = −0.45 (95% CI −0.67 to 0.24) including all studies, compared to r s (S, n ) = −0.02 (95% CI −0.31 to 0.29) excluding studies with a sample size ≤ 10, indicating a risk of moderate bias in the case of smaller sample sizes. For E, the correlation coefficient is r s (E, n ) = −0.13 (95% CI −0.41 to 0.14) if all studies are included and r s (E, n ) = −0.16 (95% CI −0.45 to 0.14) excluding studies with a sample size ≤ 10, indicating a low risk of bias in both cases . For the pooled estimates, we therefore excluded smaller studies with a sample size ≤ 10. Depending on the disease (VL, CL, PKDL), reference standard used (microscopy, PCR methods [PCR, qPCR, LnPCR were grouped together] and qPCR in the case of PKDL) and sample type for LAMP, datasets were combined and are shown under the respective heading. Pooled estimates for S and E of subgroups are shown in Fig. a–c. The pooled estimates for S are > 90% for all subgroups except subgroup 4 (LAMP compared with microscopy for CL diagnosis). For VL diagnosis compared to either of the two reference standards (microscopy, PCR) and PKDL diagnosis compared to qPCR, the pooled estimate for E is > 95%. The pooled estimate for subgroup 4 (specificity of LAMP for CL diagnosis compared to microscopy) is 67% (95% CI 45–84%), much lower than any other pooled estimate value. LAMP for diagnosis of VL Compared to microscopy as a reference standard using the sample types bone marrow aspirates (BMA), splenic aspirates (SA) or lymph node aspirates (LNA) for VL diagnosis (subgroup 1) (Fig. , S1), datasets ( n = 9) show S values for LAMP using blood as sample type ranging from 80 to 99% (pooled estimate 93.8%, 95% CI 87.8–96.9%) and E ( n = 7) from 72 to 100% (pooled estimate 97.2%, 95% CI 88.5–99.4%; two datasets did not contain values for E). Test results for 1141 individual tests are contained in subgroup 1, and the values for I 2 and Tau-squared are 67.78 and 0.76 for the S analysis and 86.55 and 3.22 for the E analysis. Compared to PCR methods (PCR, qPCR, LnPCR) as reference standards where both tests used blood samples for VL diagnosis (subgroup 2) (Fig. , S2), datasets ( n = 11) show an S ranging from 83 to 98% (pooled estimate 93.0%, 95% CI 89.5–95.5%) and E ranging from 66–99% (pooled estimate 96.4%, 95% CI 89.4–98.8%) for LAMP. Results of 1007 individual tests are contained in subgroup 2, and the values for I 2 and Tau-squared are 9.86 and 0.06 for the S analysis and 79.49 and 2.75 for the E analysis. LAMP for diagnosis of PKDL Compared to qPCR as a reference standard where both tests used tissue biopsy samples for PKDL diagnosis (subgroup 3) (Fig. , S3), datasets ( n = 3) show an S ranging from 83–97% (pooled estimate 96.3%, 95% CI 91.0–98.5%) and an E of 98% (pooled estimate 97.8%, 95% CI 90.0–99.6%) for LAMP. Test results of 198 individual tests are contained in subgroup 3, and the values for I 2 and Tau-squared are 0.00 and 0.00 for the S analysis and 0.00 and 0.00 for the E analysis. LAMP for diagnosis of CL Compared to microscopy as reference standard (subgroup 4) (Fig. , S4), datasets show an S ( n = 6) ranging from 83 to 99% (pooled estimate 89.2%, 95% CI 82.5–93.6%) and E ( n = 5) ranging from 31 to 94% (pooled estimate 64.0%, 95% CI 35.5–85.2%; one dataset did not contain values for E) for LAMP. Test results of 687 individual tests are contained in subgroup 4, and the values for I 2 and Tau-squared are 51.63 and 0.22 for the S analysis and 84.39 and 1.37 for the E analysis. Compared to PCR variations (PCR, qPCR, nested PCR) as a reference standard (subgroup 5) (Fig. , S5), datasets ( n = 8) show an S ranging from 80–99% (pooled estimate 91.6%, 95% CI 85.5–95.3%) and E ranging from 91–98% (pooled estimate 94.8%, 95% CI 87.6–97.9%) for LAMP. Test results of 672 individual tests are contained in subgroup 5, and the values for I 2 and Tau-squared are 57.73 and 0.38 for the S analysis and 0.00 and 0.00 for the E analysis. LAMP for diagnosis of CL and VL in animals In general, few studies reported data on leishmaniasis in animals. Compared to microscopy as a reference standard, datasets ( n = 3 , III, V and VIII) show an S ranging from 54 to 100% and an E ranging from 43%–77% for LAMP. Compared to PCR variations (qPCR, PCR-RFLP) as a reference standard, datasets ( n = 9, numbers I, II, IV, VI, VII and–IX-XII) show an S ranging from 0 to 100% and an E ranging from 50 to 100% for LAMP. In line with human studies, if datasets are derived from the same individuals within the same study, only those datasets with the reported higher S were considered. The three datasets comparing LAMP to microscopy are part of two separate studies investigating canine leishmaniasis, CL and VL, in 186 animals . Datasets III and VIII report an S of 100% (95% CI 74–100%) and 68% (95% CI 49–83%) and an E of 43% (95% CI 33–54%) and 77% (95% CI 61–89%), respectively. Six datasets (IV, VI, VII, X-XII), part of three studies , compare LAMP to PCR for investigation of canine leishmaniasis (CL and VL) in a total of 279 animals. Datasets IV, VI and XII report an S of 100% (95% CI 95–100%), 75% (95% CI 51–91%) and 91% (95% CI 59–100%), and an E of 91% (95% CI 77–98%), 78% (95% CI 65–88%) and 96% (95% CI 86–100%), respectively. One study (dataset IX, ) investigated VL in domestic cattle and only reported negative cases. Two datasets (I and II), part of one study , reported data from CL in Syrian hamsters, with a reported S of 89% (95% CI 65–99%) and an E of 100% (95% CI 40–100%) for dataset I and an S of and 100% (95% CI 59–100%) and an E of 50% (95% CI 1–99%) for dataset II, which only analysed seven samples. Due to the great heterogeneity with regard to animal species, forms of leishmaniasis (CL vs VL) and sample types, no pooled analysis was conducted. Analysis of LAMP performance using SROC curves Based on the subgroups, where similar studies such as LAMP used blood samples for diagnosis of VL compared to microscopy, analyses using SROC curves were performed. The SROC curves for different sample types comparing LAMP with microscopy and PCR are shown in Fig. . The AUC values are 0.973 (subgroup 1), 0.960 (subgroup 2), 0.881 (subgroup 4) and 0.964 (subgroup 5), indicating that LAMP is a highly sensitive and specific diagnostic test for VL, PKDL and CL. The full workflow of the literature search, based on the principles of the PRISMA guidelines for systematic reviews , is shown in Fig. . A total of 394 publications were retrieved; after deduplication, 228 publications were screened by title and abstract and 50 by full text based on the inclusion/exclusion criteria as detailed in the “ ” section. Studies were excluded at the title/abstract screening stage for the following reasons: wrong pathogen (condition under investigation of the study was not caused by Leishmania sp.), no LAMP (LAMP was not used as a diagnostic test method) or wrong article type (reviews and commentaries were excluded, but references were screened for further studies). Studies were excluded at the full-text assessment stage for the following reasons: lack of data (inability to complete a 2 × 2 contingency table), duplicate study (the same clinical data were described in another study), no LAMP (LAMP was not used as a diagnostic test method), no paired samples (samples tested with LAMP and the reference standard were not from the same individuals) or promastigote form (test samples were derived from the promastigote form). Twenty-seven studies were accepted for further analysis and dataset extraction—22 studies regarding leishmaniasis diagnosis in humans (Additional file : Table S2) and five studies addressing diagnosis in animals (Table ). For the extracted variables infecting species and readout methods , we used ( indicated ) if not mentioned directly in the text. For the infecting species, this refers to identification through for example the use of specific primer pairs or epidemiological data without confirmation by further analysis. For the readout method, this refers to identification through specific reagents/kits used. The datasets were stratified by clinical condition, sample type and reference test used. A full list of datasets per study is available as supplementary information (see Additional file : Table S1). Eighty-one and 12 datasets were constructed based on the included studies for LAMP diagnosis in humans and animals, respectively. In the case of missing data for completion of a 2 × 2 contingency table, or a need for clarification, the corresponding and/or first authors of 13 publications were contacted, enabling seven additional datasets to be constructed. The following descriptions are based on the included human studies, where the following studies are counted more than once as different indications are analysed: Adams et al. two studies (VL and CL), Verma et al. three studies (VL, PKDL and CL), Verma et al. two studies (VL, PKDL), and Sriworarat et al. two studies (VL and CL), resulting in 27 studies in total. In total, 2255 individuals and 6159 test results for diagnosis of leishmaniasis in humans are included in this review. Of the individual tests, 1453 are for diagnosis of VL and 650 of CL. The studies were performed from 2009 to 2019, and about half of them ( n = 14) during the past 4 years (2017–2020). Out of 27 studies, 21 (78%) evaluated the LAMP performance in the Old World, while four studies evaluated the LAMP performance in New World countries (Brazil, Colombia and Suriname) , and one study included a travel case from Venezuela . For two studies the origin of patients is not mentioned. Eighteen studies (67%) used a control group while nine (33%) were categorized as consecutive. Two studies included analysed LAMP performance in PKDL diagnosis. Twenty-three studies (85%) used a commercial kit for DNA extraction; in seven (26%) the kit used was the QIAamp ® DNA Blood mini kit (QIAGEN, Hilden, Germany), and six (22%) used a commercial kit for LAMP, which was the Loopamp™ Leishmania detection kit (Eiken Chemical, Tokyo, Japan). In 12 cases. L. donovani was found or indicated (e.g. through usage of species-specific primer pairs) as the infecting species, L. tropica was found in three studies, and L. infantum , L. major and L. guyanensis were found or indicated in two studies each. In 11 studies (40%) the target was kinetoplast DNA (kDNA); in seven (26%) the targets for LAMP were a combination of 18S ribosomal RNA (rRNA) and kDNA genes. The cysteine proteinase b ( cpb ) gene, ITS1 DNA sequences and k26 were used in one study each as the targets. In 23 studies (85%) a PCR method (PCR, qPCR or LnPCR) was used as a reference standard, and in 21 (78%) a microscopy method (microscopy or culture microscopy) was used as a reference standard. The sample size of the 27 included studies ranges from two to 274, with a median of 72 and an interquartile range from 38 (25th percentile) to 95.5 (75th percentile). The quality of included studies was analysed based on the QUADAS-2 tool ; the results separated by VL and CL diagnosis studies are shown as supplementary information (see Additional file : Figure S1). The risk regarding applicability of (1) reference standard, (2) index test and (3) patient selection were judged as low for the included studies. For index and reference test, the risk of bias is unclear in most included studies, with some having a high risk of bias regarding the categories flow and timing and patient selection. The forest plots for the S and E of LAMP vs the reference test per dataset are given as supplementary information (see Additional file : Figure S2). Spearman’s rank correlation coefficient evaluating the correlation between S and sample size is r s (S, n ) = −0.45 (95% CI −0.67 to 0.24) including all studies, compared to r s (S, n ) = −0.02 (95% CI −0.31 to 0.29) excluding studies with a sample size ≤ 10, indicating a risk of moderate bias in the case of smaller sample sizes. For E, the correlation coefficient is r s (E, n ) = −0.13 (95% CI −0.41 to 0.14) if all studies are included and r s (E, n ) = −0.16 (95% CI −0.45 to 0.14) excluding studies with a sample size ≤ 10, indicating a low risk of bias in both cases . For the pooled estimates, we therefore excluded smaller studies with a sample size ≤ 10. Depending on the disease (VL, CL, PKDL), reference standard used (microscopy, PCR methods [PCR, qPCR, LnPCR were grouped together] and qPCR in the case of PKDL) and sample type for LAMP, datasets were combined and are shown under the respective heading. Pooled estimates for S and E of subgroups are shown in Fig. a–c. The pooled estimates for S are > 90% for all subgroups except subgroup 4 (LAMP compared with microscopy for CL diagnosis). For VL diagnosis compared to either of the two reference standards (microscopy, PCR) and PKDL diagnosis compared to qPCR, the pooled estimate for E is > 95%. The pooled estimate for subgroup 4 (specificity of LAMP for CL diagnosis compared to microscopy) is 67% (95% CI 45–84%), much lower than any other pooled estimate value. Compared to microscopy as a reference standard using the sample types bone marrow aspirates (BMA), splenic aspirates (SA) or lymph node aspirates (LNA) for VL diagnosis (subgroup 1) (Fig. , S1), datasets ( n = 9) show S values for LAMP using blood as sample type ranging from 80 to 99% (pooled estimate 93.8%, 95% CI 87.8–96.9%) and E ( n = 7) from 72 to 100% (pooled estimate 97.2%, 95% CI 88.5–99.4%; two datasets did not contain values for E). Test results for 1141 individual tests are contained in subgroup 1, and the values for I 2 and Tau-squared are 67.78 and 0.76 for the S analysis and 86.55 and 3.22 for the E analysis. Compared to PCR methods (PCR, qPCR, LnPCR) as reference standards where both tests used blood samples for VL diagnosis (subgroup 2) (Fig. , S2), datasets ( n = 11) show an S ranging from 83 to 98% (pooled estimate 93.0%, 95% CI 89.5–95.5%) and E ranging from 66–99% (pooled estimate 96.4%, 95% CI 89.4–98.8%) for LAMP. Results of 1007 individual tests are contained in subgroup 2, and the values for I 2 and Tau-squared are 9.86 and 0.06 for the S analysis and 79.49 and 2.75 for the E analysis. Compared to qPCR as a reference standard where both tests used tissue biopsy samples for PKDL diagnosis (subgroup 3) (Fig. , S3), datasets ( n = 3) show an S ranging from 83–97% (pooled estimate 96.3%, 95% CI 91.0–98.5%) and an E of 98% (pooled estimate 97.8%, 95% CI 90.0–99.6%) for LAMP. Test results of 198 individual tests are contained in subgroup 3, and the values for I 2 and Tau-squared are 0.00 and 0.00 for the S analysis and 0.00 and 0.00 for the E analysis. Compared to microscopy as reference standard (subgroup 4) (Fig. , S4), datasets show an S ( n = 6) ranging from 83 to 99% (pooled estimate 89.2%, 95% CI 82.5–93.6%) and E ( n = 5) ranging from 31 to 94% (pooled estimate 64.0%, 95% CI 35.5–85.2%; one dataset did not contain values for E) for LAMP. Test results of 687 individual tests are contained in subgroup 4, and the values for I 2 and Tau-squared are 51.63 and 0.22 for the S analysis and 84.39 and 1.37 for the E analysis. Compared to PCR variations (PCR, qPCR, nested PCR) as a reference standard (subgroup 5) (Fig. , S5), datasets ( n = 8) show an S ranging from 80–99% (pooled estimate 91.6%, 95% CI 85.5–95.3%) and E ranging from 91–98% (pooled estimate 94.8%, 95% CI 87.6–97.9%) for LAMP. Test results of 672 individual tests are contained in subgroup 5, and the values for I 2 and Tau-squared are 57.73 and 0.38 for the S analysis and 0.00 and 0.00 for the E analysis. In general, few studies reported data on leishmaniasis in animals. Compared to microscopy as a reference standard, datasets ( n = 3 , III, V and VIII) show an S ranging from 54 to 100% and an E ranging from 43%–77% for LAMP. Compared to PCR variations (qPCR, PCR-RFLP) as a reference standard, datasets ( n = 9, numbers I, II, IV, VI, VII and–IX-XII) show an S ranging from 0 to 100% and an E ranging from 50 to 100% for LAMP. In line with human studies, if datasets are derived from the same individuals within the same study, only those datasets with the reported higher S were considered. The three datasets comparing LAMP to microscopy are part of two separate studies investigating canine leishmaniasis, CL and VL, in 186 animals . Datasets III and VIII report an S of 100% (95% CI 74–100%) and 68% (95% CI 49–83%) and an E of 43% (95% CI 33–54%) and 77% (95% CI 61–89%), respectively. Six datasets (IV, VI, VII, X-XII), part of three studies , compare LAMP to PCR for investigation of canine leishmaniasis (CL and VL) in a total of 279 animals. Datasets IV, VI and XII report an S of 100% (95% CI 95–100%), 75% (95% CI 51–91%) and 91% (95% CI 59–100%), and an E of 91% (95% CI 77–98%), 78% (95% CI 65–88%) and 96% (95% CI 86–100%), respectively. One study (dataset IX, ) investigated VL in domestic cattle and only reported negative cases. Two datasets (I and II), part of one study , reported data from CL in Syrian hamsters, with a reported S of 89% (95% CI 65–99%) and an E of 100% (95% CI 40–100%) for dataset I and an S of and 100% (95% CI 59–100%) and an E of 50% (95% CI 1–99%) for dataset II, which only analysed seven samples. Due to the great heterogeneity with regard to animal species, forms of leishmaniasis (CL vs VL) and sample types, no pooled analysis was conducted. Based on the subgroups, where similar studies such as LAMP used blood samples for diagnosis of VL compared to microscopy, analyses using SROC curves were performed. The SROC curves for different sample types comparing LAMP with microscopy and PCR are shown in Fig. . The AUC values are 0.973 (subgroup 1), 0.960 (subgroup 2), 0.881 (subgroup 4) and 0.964 (subgroup 5), indicating that LAMP is a highly sensitive and specific diagnostic test for VL, PKDL and CL. Leishmaniasis is considered a neglected tropical disease with various clinical manifestations endemic in more than 80 countries. Early diagnosis and treatment is not only of utmost importance for the individual but also for the community as key components of leishmaniasis control . Since its invention, LAMP, a modification of the PCR protocol, has been described as a very robust and specific molecular diagnostic method due to the primer and amplification structure used . General advantages further include easy readout methods through visibility of reaction by-products such as turbidity , or addition of different dyes . In this section, we will discuss characteristics of the included studies and the performance of LAMP for the diagnosis of VL and CL, as well as the observed heterogeneity among the datasets. This is followed by an assessment of the implementability of LAMP in the diagnostic workflow, a brief discussion of the importance of diagnosis of leishmaniasis in animal hosts, and concluded by the study’s strengths and limitations. The pooled estimates of the subgroups comparing LAMP with microscopy/PCR for VL/CL diagnosis were > 90% for sensitivity and > 95% for specificity, except for LAMP compared to microscopy for CL diagnosis (subgroup 4), where specificity was found to be 64%, therefore only moderate. These results correspond to the calculated AUC values which are > 0.96, except for the same subgroup 4, where an AUC value of 0.881 was found. This subgroup 4 consisted of six studies for a total of 687 individual tests performed, giving a broad 95% confidence interval from 35 to 85% for specificity. This result deserves reflection. Considering the known low sensitivity of the direct microscopic test, this low specificity may demonstrate not a failure but a superior performance of the LAMP, capable of identifying true cases which are erroneously counted as false positives due to the reference test being microscopy. To overcome this issue, a composite reference standard could be used, such as that by Vink et al. . In this study, considerably more positive cases were detected by the molecular method than by microscopy (out of the 257 considered true cases, 252 were positive by qPCR and 204 by microscopy). Alternatively, statistical methods such as latent class modelling have been used in the absence of a gold standard for diagnosis . We found that for most subgroups the observed heterogeneity can be attributed to differences between the studies rather than sampling error only . The calculated I 2 values were > 0.1 for most subgroup analyses, except for subgroup 3 (LAMP compared with qPCR for PKDL diagnosis) and subgroup 2 (LAMP compared with PCR for VL diagnosis) with regard to sensitivity, and subgroup 5 (LAMP compared with PCR for CL diagnosis) with regard to specificity. Heterogeneity in the subgroups may be due to several factors potentially influencing the results of an analytical method. We found little data dedicated to the study of robustness of LAMP in the context of leishmaniasis diagnosis , and some parameters, such as stability of DNA contained in clinical samples, inter-operator reliability or operator training (e.g. new method vs a method well established in the conducting laboratory), were rarely reported in studies. Further validation studies using standardized protocols and conducted in endemic countries would enable better comparisons and support decision-making in relation to diagnostic algorithms in different scenarios. We further recommend including individual sample data for publication, in order to allow statistical meta-analyses. Parameters possibly influencing LAMP performance are sample type, DNA extraction method, target sequence and readout method (see Table ). Molecular targets, and the variety of suitable markers, for Leishmania species have been discussed in detail in Akhoundi et al. . The most frequently used targets in the studies included were kDNA and 18S rRNA, the structural RNA of the ribosomal small subunit. 18S rRNA has the advantage of being a candidate for pan- Leishmania assays due to sections of high sequence conservation between species . To a lesser extent, ITS1 , cpb , k26 and L151 were also used. In general, primers must be designed carefully and, if possible, tested in silico and in vitro, as cross-reactivity with other closely related genera such as Trypanosoma has been observed in some studies . The impact of this cross-reactivity could be reduced by taking into account different clinical presentations of patients . Special consideration should be applied to endemic areas of South America, where co-infections of leishmaniasis and Chagas disease infections caused by Trypanosoma cruzi are possible, as endemic areas of the respective pathogens overlap . An overview and evaluation of different readout methods can be found in Nzelu et al. . LAMP results can be interpreted visually by turbidity or colour change, which is used in the majority of studies. In some studies, positive samples are confirmed by gel electrophoresis . However, opening of tubes after the reaction bears the risk of introducing amplicon contamination and should therefore be conducted only with caution and suitable internal quality controls . In most studies included (85%), commercial kits were used for DNA extraction, which offer the advantage of better reproducibility, but could be less suitable for a PoC setting due to equipment requirements. Some studies also used a “direct boil-and-spin” approach : whole blood was centrifuged after addition of a lysis agent and heating. The results were found to be comparable to other LAMP protocols involving more sophisticated DNA extraction and purification (Figs. , , ), and are also in line with studies such as Nzelu et al. , but further studies using clinical samples would be needed for confirmation. Depending on the desired level of implementation, an evaluation of a “LAMP near-PoC” method focusing on using as little equipment as possible, for example the usage of electricity-free heat sources (such as the non-instrumented nucleic acid amplification [NINA] device or commercial pocket warmers ), might provide valuable insights. Protocols without kits and low laboratory equipment requirements favour the cost–benefit ratio compared to other molecular methods, making LAMP a cost-effective diagnostic method . The desired parameters of a diagnostic test strongly depend on the intended usage . As molecular diagnostic tests can have very high analytical sensitivity, they correlate better with infection status than actual disease . There are several possible reasons that the identification of asymptomatic individuals might also be desired. First of all, epidemiological prevalence studies allow for effective regional disease monitoring, and might support related decisions, for example the identification of areas where prophylactic measures (such as the usage of bed nets or insecticide-impregnated fly screens) should be promoted . Furthermore, in the context of blood donations, a method with high analytical sensitivity is desired. Contaminated blood products pose a potential risk of transmission, particularly for immunocompromised blood recipients . Related to epidemiological prevalence studies in humans, another possible area of applicability includes xenomonitoring, where a large quantity of samples can be analysed in a short time using a pooling approach . A guideline to aid in selecting the optimal diagnostic test for an intended purpose was published by WHO, reporting the ASSURED criteria (Affordable, Sensitive, Specific, User-friendly, Rapid and robust, Equipment-free and Deliverable) and their adaptation to fit each diagnostic need, also taking into account special requirements for PoC diagnostic tests . This guideline suggests six evaluation steps, starting with defining the test purpose, comparing characteristics of available products, reviewing the regulatory approval, obtaining data under first, ideal, and second, real conditions and finally, monitoring the test performance in routine use. Unfortunately, we were only able to report a limited number of studies using LAMP for the diagnosis of CL and VL in animals, and due to the heterogeneity in terms of species, forms of leishmaniasis and sample types, no pooled analysis was conducted. This is particularly disappointing, since the failure of leishmaniasis control is partially associated with a failure of control of infected animal hosts, such as dogs in domestic settings . Taking Brazil as an example, high costs for control and prevention of canine leishmaniasis have been reported previously, which are in contrast to the limited financial resources for control programmes in endemic areas . In addition, current available serological screening tests for canine leishmaniasis present a certain level of disagreement . Therefore, research into highly sensitive and specific as well as affordable methods for diagnosis of leishmaniasis in animal hosts, most importantly dogs, is very much needed and crucial for control efforts. In summary, our results show LAMP to be a suitable candidate for a PoC-test in human patients, but further research and matching against actual requirements is needed. For example, we found LAMP to only partly cover the requirements for a PoC test for CL, such as minimum sensitivity of 85% and minimum specificity of 90%, and other parameters covered in a comprehensive target product profile developed by the Foundation for Innovative New Diagnostics (FIND) . In our opinion, the strengths of this literature review and meta-analysis are the comprehensive search strategy and the number of databases included in the literature search. In addition, we aimed to include unpublished data (e.g. conference abstracts) and contacted authors; thus, a number of additional datasets could be collected. The most important limitation of this literature review and meta-analysis is the heterogeneity for most analyses based on our results; consequently, the results have to be interpreted with caution . In addition, the risk of bias was evaluated, and many of the included studies have unclear and/or high risk of bias for the evaluated parameters of “patient selection” and “flow and timing”. Moreover, although we aimed to exclude patient samples that were used in several studies, we were unable to do so and therefore decided to include a subset of VL and PKDL samples that were analysed in two studies by Verma et al. . In summary, LAMP has high sensitivity and specificity compared to microscopy and PCR methods for diagnosis of CL, PKDL and VL. An advantage of LAMP which is shared by other molecular methods is the possibility to use minimally and non-invasive sample types, such as whole blood for VL and swabs for CL diagnosis. Advantages more specific to LAMP are the high robustness and isothermal amplification, so LAMP could be conducted with unpurified or minimally purified samples and with heat sources not relying on electricity, which could be interesting in a (near-)PoC setting. Currently, LAMP seems to be a suitable diagnostic test in prevalence studies, epidemiological studies (in humans and animals) and diagnosis in a diagnostic algorithm, especially for immunocompromised patients, or possibly for monitoring therapeutic success. Our findings are limited by the rather low number of studies available; thus, further large-scale studies evaluating LAMP in field settings, complemented by cost-effectiveness analyses, are recommended to gain further insights. Additional file 1: Text S1. Search strategy and results per database. Shows the specific search strategy used for the databases included in the review, the search dates and the number of results per database. Additional file 2: Table S1. Datasets of included studies addressing the diagnosis of leishmaniasis in humans and animals. Additional file 3: Table S2. Main methodological characteristics of studies addressing the diagnosis of leishmaniasis in humans. Additional file 4: Figure S1. QUADAS-2 based quality risk assessment. Additional file 5: Figure S2. Forest plots for sensitivity and specificity for all identified datasets. |
European expert consensus on a structured approach to circular stapling anastomosis in minimally invasive left‐sided colorectal resection | 4c9e8fdd-98bc-45fd-a0e1-101c3bb3d738 | 11842941 | Surgical Procedures, Operative[mh] | The performance of intestinal anastomosis is one of the most critical steps in colorectal surgery, and complications associated with anastomosis can have devastating consequences for the patient's clinical, functional and oncological outcomes. Complications also create a significant burden on the healthcare system. Circular stapling is commonly performed in left‐sided colorectal anastomosis (sigmoid colectomy, high and low anterior resection) for benign and malignant conditions. It is used in open, laparoscopic and robotic surgeries. A recent review of a healthcare database with 13 167 patients who underwent left‐sided colorectal resection showed that 22.7% of patients had circular anastomotic complications . In another study, knowledge gaps in many surgeons' understandings of the safe use of various commonly used medical devices, including stapling knowledge, were reported . A high incidence of technical errors involving the use of circular staplers has also been reported . Consequently, there is a need for surgical strategies and technologies to standardize and quality assure anastomotic techniques to lower the risk of anastomotic complications . Emerging evidence has shown a strong relationship between the intraoperative performance of the surgeon operator and patient outcomes . Our endeavour from a surgical community is to improve intraoperative performance , which we believe will have a considerable impact on patient safety and operative outcomes. One scientific approach to improving intraoperative performance is proficiency‐based progression (PBP) simulation training. PBP begins by deconstructing the procedure or skill being focused on into explicitly defined (binary) performance metrics, which are then validated . The PBP approach to training makes skill acquisition more objective, transparent and fair. During training, trainees are given metric‐based feedback on their performance, which is explicit, constructive and formative . In a recent systematic review of 12 prospective randomized and blinded clinical studies (PBP‐trained versus traditionally trained surgeons), PBP‐trained surgeons demonstrated significantly fewer performance errors (a 60% reduction) . Our overarching goal was to improve training in circular stapling devices in minimally invasive left‐sided colorectal anastomosis using PBP methodology, and this first part of our project was to develop and objectively define performance metrics that characterize a reference approach to the application of circular stapling devices in left‐sided colorectal anastomosis during minimally invasive operations (i.e. laparoscopic and robotic) and to obtain face and content validity through a consensus meeting (i.e. with a Delphi panel) of very experienced and expert colorectal surgeons (senior consultant >10 years’ colorectal practice). The principle of metric development and stress testing (face and content validation) for PBP training has been described in detail previously . This approach was applied when developing the circular stapling anastomosis metrics for minimally invasive left‐sided colorectal anastomosis and is described below. Metrics Group The Metrics Group consists of three experienced colorectal surgeons (AW, GB, ST) with a special interest in minimally invasive surgery, a senior behavioural scientist and an education–training expert (AGG), and a research fellow who is specialized in metrics development for surgical procedures (RF). Input was sought from device engineers who specialize in circular stapling devices. Circular stapling anastomosis metrics development A detailed task analysis and deconstruction process was used to deconstruct a reference approach to the use of circular stapling anastomosis for minimally invasive left‐sided colorectal procedures in small, nonoverlapping performance units . Published written guidelines, video teaching materials, manufacturer's instructions for use and access to 10 anonymized unedited minimally invasive left‐sided colorectal operations using circular stapling anastomosis performed by surgeons with different levels of experience supported the metrics development and procedure characterization process. The goal was to characterize a ‘reference’ approach to circular stapling anastomosis used in minimally invasive left‐sided colorectal operations. A reference procedure is assumed to be a straightforward and uncomplicated guide for trainees in learning the optimum performance of these procedures. The phases and steps are the same for female and male patients undergoing the anastomosis part of the minimally invasive left‐sided colorectal resection. For the ‘reference procedure’ there are agreed criteria for patient selection and procedure‐specific factors (Table ). A one‐day preliminary face‐to‐face planning meeting, three face‐to‐face meetings for metrics identification and definition and the metric stress test were conducted. Videoconferences (a total of 5 h) using Zoom (San Jose, CA, USA) and email exchanges were used to complement face‐to‐face meetings for further clarification and definition of the metrics. At the beginning of the metrics development the Metrics Group agreed on the following definitions: Performance metrics: units of observable behaviour which together constitute a stepwise description of a reference approach to a procedure. Procedural phase: a group or series of integrally related events or actions that, when combined with other phases, make up or constitute a complete operative procedure. Step: a component task, the series aggregate of which forms the completion of a specific procedure. Error: a deviation from optimal performance. Critical error: a major deviation from optimal performance, which is likely to cause harm to the patient or compromise the safe completion of the procedure . The metrics, therefore, consist of procedural phases involved in a minimally invasive left‐sided colorectal anastomosis. Each phase comprises specific steps required for accomplishment. The importance of the metrics approach in defining these phases and steps is that these are explicit and unambiguous. The procedural step either occurred or did not occur and can be scored as such by an external reviewer with high reliability . Similarly, procedural errors and critical errors were defined associated with particular steps within different phases of the procedure. For errors, behaviours exhibited by the operator may not necessarily in and of themselves lead to a bad outcome or an event with more serious consequences, but their enactment sets the stage or increases the probability for a more serious event to occur or detracts from the efficient and possibly safe execution of the desired procedure. In contrast, a ‘critical error’ is a more serious occurrence and represents operative performance that could either jeopardize the outcome of the procedure or lead to significant iatrogenic damage . Figure illustrates an example of a procedural phase characterized by circular stapling anastomosis in minimally invasive left‐sided colorectal procedures. In addition to the metrics, valuable knowledge and principles of the operation were compiled, such as the mechanics and science of anastomosis, to facilitate the learning process; these formed the didactic component for the learner during the training process. Once the Metrics Group had defined the metrics they were then used to score five unedited anonymized circular stapler anastomosis parts of the minimally invasive approach for left‐sided colorectal resection performed by different surgeons with various levels of experience. Scoring was performed by the members of the Metrics Group independently. Any difference in the scoring was discussed in order to identify discrepancies in interpretation or ambiguities in the metric definition. Based on this process, and if agreed upon, changes were made in the metrics, which facilitated the scoring agreement. This process was repeated for each video until the Metrics Group was satisfied with the metrics and they could be scored with a high degree of reliability (i.e. inter‐rater reliability >0.8, which is the internationally agreed gold standard) . Metrics stress testing (face and content validation) with a modified Delphi approach Once the metrics for the circular stapling anastomosis for minimally invasive left‐sided colorectal resection had been defined and characterized, face validity and content were verified by a group of experienced colorectal surgeons. An international panel of expert colorectal surgeons was invited to join the Delphi panel to provide a more objective and independent assessment of the metrics. Informed consent was obtained from the Delphi panel members. The panel was chosen for their colorectal surgical experience and their demonstrated educational interests and commitment. The equality, diversity and inclusion principle was adhered to when selecting the Delphi panel members . Sixteen expert colorectal surgeons, including the Metrics Group members from nine countries, a nonvoting behavioural scientist and a nonvoting fellow who is familiar with metrics development in surgical procedures, attended a consensus meeting in Dublin, Ireland on 23 September 2022 (Table ). A brief overview of the project and meeting objectives was presented. Background information regarding PBP training methodology, prior literature demonstrating the validity of this training approach for procedural specialties and the specific objectives of the current Delphi panel were reviewed . Each phase of the procedure, the procedural steps that were included in that phase, and the potential errors were presented. It was also explained that the associated metrics had been developed by the Metrics Group for a reference approach to circular stapling anastomosis for minimally invasive left‐sided colorectal resections. It was acknowledged that the designated reference procedure might not reflect the exact techniques employed by individual Delphi panellists, but that the operative steps presented accurately embodied the essential and key components of the procedure and ‘were not wrong’ . To assess the correlation of the procedural steps, errors and critical errors before and after the Delphi process, changes were analysed with the Pearson chi‐square test (IPM SPSS Statistics for Windows, version 26; IBM Corp., Armonk, NY, USA). A p ‐value of <0.05 was considered statistically significant. The Metrics Group consists of three experienced colorectal surgeons (AW, GB, ST) with a special interest in minimally invasive surgery, a senior behavioural scientist and an education–training expert (AGG), and a research fellow who is specialized in metrics development for surgical procedures (RF). Input was sought from device engineers who specialize in circular stapling devices. A detailed task analysis and deconstruction process was used to deconstruct a reference approach to the use of circular stapling anastomosis for minimally invasive left‐sided colorectal procedures in small, nonoverlapping performance units . Published written guidelines, video teaching materials, manufacturer's instructions for use and access to 10 anonymized unedited minimally invasive left‐sided colorectal operations using circular stapling anastomosis performed by surgeons with different levels of experience supported the metrics development and procedure characterization process. The goal was to characterize a ‘reference’ approach to circular stapling anastomosis used in minimally invasive left‐sided colorectal operations. A reference procedure is assumed to be a straightforward and uncomplicated guide for trainees in learning the optimum performance of these procedures. The phases and steps are the same for female and male patients undergoing the anastomosis part of the minimally invasive left‐sided colorectal resection. For the ‘reference procedure’ there are agreed criteria for patient selection and procedure‐specific factors (Table ). A one‐day preliminary face‐to‐face planning meeting, three face‐to‐face meetings for metrics identification and definition and the metric stress test were conducted. Videoconferences (a total of 5 h) using Zoom (San Jose, CA, USA) and email exchanges were used to complement face‐to‐face meetings for further clarification and definition of the metrics. At the beginning of the metrics development the Metrics Group agreed on the following definitions: Performance metrics: units of observable behaviour which together constitute a stepwise description of a reference approach to a procedure. Procedural phase: a group or series of integrally related events or actions that, when combined with other phases, make up or constitute a complete operative procedure. Step: a component task, the series aggregate of which forms the completion of a specific procedure. Error: a deviation from optimal performance. Critical error: a major deviation from optimal performance, which is likely to cause harm to the patient or compromise the safe completion of the procedure . The metrics, therefore, consist of procedural phases involved in a minimally invasive left‐sided colorectal anastomosis. Each phase comprises specific steps required for accomplishment. The importance of the metrics approach in defining these phases and steps is that these are explicit and unambiguous. The procedural step either occurred or did not occur and can be scored as such by an external reviewer with high reliability . Similarly, procedural errors and critical errors were defined associated with particular steps within different phases of the procedure. For errors, behaviours exhibited by the operator may not necessarily in and of themselves lead to a bad outcome or an event with more serious consequences, but their enactment sets the stage or increases the probability for a more serious event to occur or detracts from the efficient and possibly safe execution of the desired procedure. In contrast, a ‘critical error’ is a more serious occurrence and represents operative performance that could either jeopardize the outcome of the procedure or lead to significant iatrogenic damage . Figure illustrates an example of a procedural phase characterized by circular stapling anastomosis in minimally invasive left‐sided colorectal procedures. In addition to the metrics, valuable knowledge and principles of the operation were compiled, such as the mechanics and science of anastomosis, to facilitate the learning process; these formed the didactic component for the learner during the training process. Once the Metrics Group had defined the metrics they were then used to score five unedited anonymized circular stapler anastomosis parts of the minimally invasive approach for left‐sided colorectal resection performed by different surgeons with various levels of experience. Scoring was performed by the members of the Metrics Group independently. Any difference in the scoring was discussed in order to identify discrepancies in interpretation or ambiguities in the metric definition. Based on this process, and if agreed upon, changes were made in the metrics, which facilitated the scoring agreement. This process was repeated for each video until the Metrics Group was satisfied with the metrics and they could be scored with a high degree of reliability (i.e. inter‐rater reliability >0.8, which is the internationally agreed gold standard) . Once the metrics for the circular stapling anastomosis for minimally invasive left‐sided colorectal resection had been defined and characterized, face validity and content were verified by a group of experienced colorectal surgeons. An international panel of expert colorectal surgeons was invited to join the Delphi panel to provide a more objective and independent assessment of the metrics. Informed consent was obtained from the Delphi panel members. The panel was chosen for their colorectal surgical experience and their demonstrated educational interests and commitment. The equality, diversity and inclusion principle was adhered to when selecting the Delphi panel members . Sixteen expert colorectal surgeons, including the Metrics Group members from nine countries, a nonvoting behavioural scientist and a nonvoting fellow who is familiar with metrics development in surgical procedures, attended a consensus meeting in Dublin, Ireland on 23 September 2022 (Table ). A brief overview of the project and meeting objectives was presented. Background information regarding PBP training methodology, prior literature demonstrating the validity of this training approach for procedural specialties and the specific objectives of the current Delphi panel were reviewed . Each phase of the procedure, the procedural steps that were included in that phase, and the potential errors were presented. It was also explained that the associated metrics had been developed by the Metrics Group for a reference approach to circular stapling anastomosis for minimally invasive left‐sided colorectal resections. It was acknowledged that the designated reference procedure might not reflect the exact techniques employed by individual Delphi panellists, but that the operative steps presented accurately embodied the essential and key components of the procedure and ‘were not wrong’ . To assess the correlation of the procedural steps, errors and critical errors before and after the Delphi process, changes were analysed with the Pearson chi‐square test (IPM SPSS Statistics for Windows, version 26; IBM Corp., Armonk, NY, USA). A p ‐value of <0.05 was considered statistically significant. The ages of the panel members ranged from 34 to 65 years, and there were five female surgeons. Six panel members were heads of their respective departments and four were full professors affiliated with universities. The combined number of colorectal resections performed or supervised by the Delphi panel was more than 1500 per annum. The Metrics Group proposed three phases for the circular stapling anastomosis in minimally invasive left‐sided colorectal resection, each with a defined beginning and end (Table ). Some criteria needed to be fulfilled before the circular stapling anastomosis stage. During the Delphi meeting, the Delphi panel suggested and agreed upon two additional conditions (see section): the rectal stump should be clean and the surgeon should (have) read the instructions for use for the circular stapling device. During the Delphi meeting, four steps were added, making a total of 36 steps for the three phases of the circular stapling anastomosis (Table ). The added steps were ‘Surgeons request the correct staple length and height’ when using a linear stapler in the transection of the rectum (Phase I), ‘Surgeons request for the correct stapler and stapler size’ when using a circular stapler in the preparation of the proximal colon for anastomosis (phase II), ‘Verify verbal communication between the surgical team members before firing the stapler’, ‘Surgeon fire the stapler in a standing position (to stabilize during firing) during anastomosis’ (Phase III). Modifications were made in four steps (Phases I and II) to make the steps more explicit and instructive. The Metrics Group identified 40 procedural errors in the three phases, and after the Delphi meeting the total number of procedural errors was 42 (Table ). There were 38 procedural critical errors before and 39 after the Delphi meeting (Table ). Furthermore, the number of procedural steps, errors and critical errors before and after the Delphi changes were highly correlated [Pearson correlation coefficient r = 0.974 (95% CI r = 0.861–0.994) p < 0.001]. On average, there were more procedural steps [before 10.7 (SD = 5.9); after 12 (SD = 6.2)] at the end of the Delphi meeting. The same was observed for errors [before 13.3 (SD = 9.3); after 14 (SD = 9.2)] and critical errors [before 12.7 (SD = 9); after 13 (SD = 8)]. When we compared these differences with Wilcoxon sign rank (two‐tail) tests none of the differences were found to be statistically significant (steps, Z = −1.633, p = 0.102; errors, Z = −0.447, p = 0.665; critical errors, Z = 0, p = 1.0). After discussion and changes to the metrics incorporated during the meeting, the metrics for circular stapling anastomosis in minimally invasive left‐sided colorectal resection received 100% consensus from the Delphi panel. Anastomotic complications are common following left‐sided colorectal resection. Among these complications, an anastomotic leak can have devastating consequences for patients' outcomes, including survival rate, cancer recurrence, permanent stoma, negative impact on the bowel and sexual function and long‐term quality of life . Complications also increase the length of hospital stay and place a significant extra resource burden on healthcare institutions . Researchers have been studying the factors associated with anastomotic complications and identifying management strategies to reduce the burden caused by these complications . The circular stapling device is commonly used in left‐sided colorectal anastomosis, in both cancer and benign conditions, but this crucial step of the procedure has not been taught in surgical training. Given that evidence suggests there are gaps in stapling knowledge and a high incidence of technical errors when using a circular stapler, there is an imperative to standardize and define structured training for this critical part of the procedure . More focus is now placed on the surgeon's skill, as evidence now shows that it is strongly linked with patient outcomes . The Metrics Group has identified one scientific approach to structured training in circular stapling anastomosis in minimally invasive left‐sided colorectal resection, namely PBP simulation training. This method makes skill acquisition more objective, transparent and fair. Based on Level 1a evidence, use of the PBP method significantly reduced performance errors by 60% . Using the PBP method, we characterized the performance metrics (procedural phases, steps, errors, critical errors) for circular stapling anastomosis for minimally invasive left‐side colorectal resection. A minimally invasive approach for left‐sided resection is widely practised, but practitioners would find the metrics useful for the open approach. The performance metrics development process was robust and has been used with success in other disciplines . The Metrics Group consisted of three expert colorectal surgeons and individuals who specialize in the PBP methodology, including a senior behavioural scientist with more than two decades of experience in surgical training. Expert engineers working with the circular stapling device were consulted, specifically in relation to instructions for use and technical device handling. These performance metrics were scrutinized by a panel of expert colorectal surgeons from different European countries and a renowned minimally invasive expert surgeon from an academic centre in Malaysia. During a minimally invasive approach to left‐sided colorectal resection, surgeons have variations of practice when performing circular stapling anastomosis. The performance metrics presented in the Delphi meeting aimed to outline a standardized approach suitable for learners. Minor modifications were made during the Delphi meeting to make the performance metrics more explicit and instructive. Some general principles, for example stapling technologies, will be provided as didactic to the trainees in addition to the metrics. The pre‐ and post‐Delphi metrics were highly correlated (Tables , , , ). After incorporating the changes suggested by the Delphi panel, voting was obtained at the end of the discussion of each phase. All of the procedural phases received unanimous agreement. Anastomotic complications, particularly leaks, are among the most feared complications in colorectal surgery. The anastomotic part of the procedure is performed towards the end of an operation; potentially, issues of fatigue and concentration may be introduced at this crucial part of the operation. A successful operation also depends on the skills of the operating team, not only the lead surgeon. This is important, as often the introduction of the circular stapling device is performed by more junior surgical team members. During the Delphi meeting, the panel members recognized the knowledge gap and training needed in the use of the circular stapling device. Some valuable additional comments were made and incorporated into the performance metrics, such as ‘Surgeons request for the correct stapler and stapler size’ and ‘Verify verbal communication between the surgical team members before firing the stapler’. The PBP approach to characterize these three phases of circular stapling anastomosis during a crucial part of a minimally invasive approach to left‐sided colorectal resection allows surgeons to learn the steps with explicit performance instructions about what to do and, possibly more importantly, what not to do. The PBP method affords performance assessments where the metrics are used to provide feedback to learners that are objective, transparent, explicit, constructive and formative. The errors and critical errors that were described would further enhance training. The proposed metrics are for a standard and straightforward procedure. The aim is to provide a structured stepwise approach to use of the device during this segment of the procedure. We do, however, appreciate the variety of practices; for example, when making the purse‐string for the proximal end of the colon, a purse‐string applicator can be used instead of a manual purse‐string, as detailed in our metrics. During a minimally invasive approach to left‐sided colorectal resection, circular stapling anastomosis can be broken down into procedural phases and steps, with errors and critical errors known as performance metrics. Data from a large group of expert colorectal surgeons from Europe provided evidence to support the face and content of these metrics. We consider the metrics essential for developing structured training using circular stapling anastomosis in a minimally invasive approach to left‐sided colorectal resection. Further development of these metrics is vital to guide the training curriculum and assessment. Samson Tou: Conceptualization; investigation; funding acquisition; writing – original draft; methodology; validation; writing – review and editing; project administration; data curation; supervision; resources; visualization; formal analysis. Anthony G. Gallagher: Conceptualization; investigation; funding acquisition; methodology; validation; writing – review and editing; visualization; formal analysis; project administration; data curation; supervision; resources. Gabriele Bislenghi: Investigation; writing – review and editing. Rui Farinha: Investigation; methodology; writing – review and editing. Albert Wolthuis: Conceptualization; investigation; funding acquisition; methodology; validation; visualization; writing – review and editing; project administration; supervision; resources. Medtronic (Surgical Division) provided the educational grant for this study but did not influence the selection of the experts, the design and conduct of the research, data collection, analysis or the preparation of the manuscript. ST received education grants from Intuitive Foundation and Medtronic. AGG holds education research grants from Medtronic (Dublin, Ireland), AO Education Institute (Davos, Switzerland), and the Arthroscopic Association of North America (Chicago, USA) to investigate metric‐based education and training. All participants provided informed consent prior to participating in the study, and the study protocol was approved by the institutional review board at the University of Leuven. |
What did the pandemic teach us about effective health communication? Unpacking the COVID-19 infodemic | 2b1c2cab-34f0-4723-a414-32f121081549 | 9747260 | Health Communication[mh] | According to the Protective Action Decision Model (PADM) , behavioral response to health risks depends partly on information exposure. By receiving timely and accurate information about COVID-19, individuals are better equipped to formulate accurate risk perceptions and engage in preventive steps . Following this logic, it is essential that evidence-based COVID-19 information be translated in a manner that meets the needs of diverse stakeholder groups by understanding the factors associated with health information-seeking behavior (HISB). One strategy for understanding HISBs during the pandemic is to explore preferences for the first sought information source as an indicator of persuasiveness , and for sources deemed most trustworthy as a proxy for credibility . The concept of uncertainty is important to HISB. The novelty of COVID-19 and lack of societal preparedness increased uncertainty in how to respond , increasing the likelihood that individuals will seek to manage this uncertainty by searching for relevant information . Related to this idea of HISB as a tool to manage uncertainty is self-efficacy, which refers to the extent to which one believes in their ability to successfully perform a behavior . Prior to engaging in a HISB, individuals have a tendency to first develop outcome expectations and evaluate whether they possess the ability to enact this search . HISB during the pandemic operates within the context of advances in mass media technology, with increased use of digital media platforms (e.g., Internet search engines, social media) and the associated concerns regarding false information . In addition to information received from interpersonal sources (e.g., friends, family, health care providers), information consumers now have diverse opportunities to seek and obtain health-related information with platforms such as Facebook, YouTube, and Google providing 24/7 access to information of varying quality . The affordances of these platforms (e.g., sharing, liking, commenting) allow for an enhanced ability to create, receive, and disseminate health information. This increased media choice also allows for selective exposure to like-minded voices, which can lead to increased perceptions of bias within the general media . Further, media slant towards a specific political ideology or issue position can be extreme within these mediated settings, with exposure having an influence on COVID-19 incidence . All told, preferences for health information in the current media landscape warrant exploration to assess how audience factors are related to HISB. While technological advancements have placed a wealth of information at our fingertips, there are disparities in who utilizes and benefits from these technologies based in part on longstanding social and digital inequalities . While it has been argued that groups often marginalized by society (e.g., inequality based on age, gender, educational attainment, etc.) are simply lagging behind the curve in uptake of these technologies and will eventually bridge the gap, many of these groups often require government intervention to stimulate use and are more likely to discontinue use once begun . Also, for these marginalized groups, self-efficacy in the use of technology is likely to be lower compared to those with more capital . Therefore, while mass media technology provides wide reach and convenience to many, the associated inequities in use suggest that health communication campaigns seeking to tailor dissemination strategies should attend to audience features that may point to source preferences. The COVID-19 pandemic illuminated how racial and ethnic discrimination can be amplified via the media, making HISB difficult for some groups. Anti-Asian sentiment, fueled in part by social media, has seen a dramatic increase during the pandemic, and politicians have used this crisis to propagate stereotyping and discriminatory policies against racial and ethnic minorities . Black Americans who have historically been confronted with significant racism and discrimination in the US also report that their experiences with discrimination have increased during COVID-19 . This lived discrimination can act as a biological stressor for which individuals must develop coping strategies, such as information-seeking . Given that uncertainty about COVID-19 has increased HISB , focused effort is needed to deliver evidence-based health information through preferred sources in order to combat mis/disinformation and improve population health. Prior work has demonstrated that self-efficacy is positively associated with the frequency of HISB during the pandemic . However additional work is needed to explore the sources people seek out first and which ones they trust the most, particularly in relation to confidence in information-seeking ability. In some cases, first sought and most trusted sources may be the same. In other cases, the sources that are most readily available to an individual may not be the most trusted. For example, some individuals may find health care providers to be highly trustworthy, but they are unavailable 24/7 to meet information needs. Digital inequalities have also likely been exacerbated during the pandemic as marginalized groups are unable to offset the loss of in-person communication ; these factors may contribute to differences in HISB . Further, increases in perceived discrimination may be associated with information-seeking strategies during the pandemic . Building on previous work related to HISB during COVID-19 , the aim of this study was to investigate individual preferences for the first sought out and most trusted sources of COVID-19 information to guide tailored campaign development. Research questions RQ1: Are sociodemographic characteristics associated with preferences for (a) first sought and (b) most trusted source of COVID-19 information? RQ2: Are discrimination and self-efficacy associated with preferences for (a) first sought and (b) most trusted source of COVID-19 information? RQ1: Are sociodemographic characteristics associated with preferences for (a) first sought and (b) most trusted source of COVID-19 information? RQ2: Are discrimination and self-efficacy associated with preferences for (a) first sought and (b) most trusted source of COVID-19 information? Study design and participant recruitment Using a cross-sectional study design, between September and November 2020, a period that saw approximately 94,000 deaths from COVID-19 in the US (Johns Hopkins COVID-19 Tracker https://coronavirus.jhu.edu/us-map ), US adults aged ≥ 18 years ( N = 1800) recruited through a panel owned by a cloud-based survey platform completed the online Florida Health Ancestry Study survey (FHAS). The sampling framework was specified so that quotas would represent the general US adult population (see Table ). Participants meeting these inclusion criteria received an electronic link to the survey. Partial responses were not recorded, but all participants were given one week to complete the survey. The “Forced Response” validation was used for all items, although participants could select “prefer not to answer.“ A $15.00 incentive was mailed to participants who completed the survey. The University of Florida Institutional Review Board (IRB201901264) approved this study with a waiver of documentation of informed consent. Instrument Participants completed the 48-item FHAS survey developed using the behavioral core measures from NCI-designated cancer center catchment area supplements . The FHAS includes investigator-derived measures related to COVID-19, perceived discrimination, and self-efficacy in obtaining health information (see supplement “additional_file_ ” for more information on items used in this analysis). For all items, responses of “Don’t know” and “Prefer not to answer” were treated as missing. Measures COVID-19 information-seeking To measure the first sought and most trusted sources of information about COVID-19, participants responded to two items (the COVID-19 questions in this study were adapted from a Palliative Care & Supportive Oncology Workgroup Survey and the eHealth Literacy Scale ), (“When you had a strong need to get information about COVID-19, where did you FIRST go to get information?“; “When you had a strong need to get information about COVID-19, which of the following did you find to be the MOST trusted as a source of information about coronavirus or COVID-19?“). For the univariable and multivariable analyses, response options for both items were dichotomized into the following sources: “Mass media” (Internet: Google or another search engine/WebMD or another medical website; Printed materials: newspapers, magazines; Social media: Facebook, Instagram, Twitter; Television) and “Interpersonal” (Conversations with people you trust: friends, relatives, or co-workers; Health care provider: doctor, nurse, social worker). Responses of “Other (Please specify:)” were treated as missing. Self-efficacy On a 5-point scale where 1 = “Not confident at all” and 5 = “Completely confident,“ self-efficacy was measured as confidence in obtaining general health information using a single item , “Overall, how confident are you that you could get advice or information about health or medical topics if you needed it?“ (M = 4.1, SD = 1.0). Perceived discrimination Experiences with everyday discrimination were assessed with a five-item measure on a four-point scale where 0 = “Never,“ 1 = “Rarely,“ and 2 = “Sometimes”; responses of Often,“ “At least once a week,“ and “Almost every day” were categorized as 3. Participants were asked how often they are treated with less courtesy or respect than others, how often they receive poorer services at restaurants or stores, how often people act as if they are afraid of them, how often people act as if they are not smart, and how often they are threatened or harassed. Perceived discrimination was calculated as the mean score of these items (α = 0.91, M = 1.3, SD = 1.0). Sociodemographic characteristics Participant information about age, gender, race, education, marital status, living situation (live alone/live with someone), income, and overall health status was also obtained. COVID-19 mitigation beliefs On a 5-point scale where 1 = “Strongly disagree” and 5 = “Strongly agree,“ participants responded to two items asking how important they thought it was to wear a mask and maintain social distance when going out in public. These two items were combined for a mean score (α = 0.81, M = 4.5, SD = 1.0). Analysis plan Multivariable logistic regression models were fitted for the first source of COVID-19 information (mass media vs. interpersonal) and the most trusted source (mass media vs. interpersonal), respectively. Specifically, an odds ratio (OR) larger than 1 indicated higher odds of choosing a mass media source, and an OR smaller than 1 showed higher odds of selecting an interpersonal information source. Univariable logistic regressions were fitted first with factors identified as potentially relevant to COVID-19 information-seeking based on previous research (e.g., ), and factors with p -values less than 0.15 were then considered for multivariable logistic regressions. Backward selection was used to build final multivariable models. Age, race, gender, education, marital status, and overall health status were kept in the multivariable model of the first source of COVID-19 information, while living situation, income, and marital status were kept in the multivariable model of most trusted source of COVID-19 information regardless of their p -values. Multivariable multinomial logistic regression models were also fitted for the first sought and most trusted source of COVID-19 information to look at specific associations between source types, but in a non-aggregated fashion: comparing trusted individuals vs. Internet vs. printed materials vs. social media vs. Television vs. health care providers. Using a cross-sectional study design, between September and November 2020, a period that saw approximately 94,000 deaths from COVID-19 in the US (Johns Hopkins COVID-19 Tracker https://coronavirus.jhu.edu/us-map ), US adults aged ≥ 18 years ( N = 1800) recruited through a panel owned by a cloud-based survey platform completed the online Florida Health Ancestry Study survey (FHAS). The sampling framework was specified so that quotas would represent the general US adult population (see Table ). Participants meeting these inclusion criteria received an electronic link to the survey. Partial responses were not recorded, but all participants were given one week to complete the survey. The “Forced Response” validation was used for all items, although participants could select “prefer not to answer.“ A $15.00 incentive was mailed to participants who completed the survey. The University of Florida Institutional Review Board (IRB201901264) approved this study with a waiver of documentation of informed consent. Participants completed the 48-item FHAS survey developed using the behavioral core measures from NCI-designated cancer center catchment area supplements . The FHAS includes investigator-derived measures related to COVID-19, perceived discrimination, and self-efficacy in obtaining health information (see supplement “additional_file_ ” for more information on items used in this analysis). For all items, responses of “Don’t know” and “Prefer not to answer” were treated as missing. COVID-19 information-seeking To measure the first sought and most trusted sources of information about COVID-19, participants responded to two items (the COVID-19 questions in this study were adapted from a Palliative Care & Supportive Oncology Workgroup Survey and the eHealth Literacy Scale ), (“When you had a strong need to get information about COVID-19, where did you FIRST go to get information?“; “When you had a strong need to get information about COVID-19, which of the following did you find to be the MOST trusted as a source of information about coronavirus or COVID-19?“). For the univariable and multivariable analyses, response options for both items were dichotomized into the following sources: “Mass media” (Internet: Google or another search engine/WebMD or another medical website; Printed materials: newspapers, magazines; Social media: Facebook, Instagram, Twitter; Television) and “Interpersonal” (Conversations with people you trust: friends, relatives, or co-workers; Health care provider: doctor, nurse, social worker). Responses of “Other (Please specify:)” were treated as missing. Self-efficacy On a 5-point scale where 1 = “Not confident at all” and 5 = “Completely confident,“ self-efficacy was measured as confidence in obtaining general health information using a single item , “Overall, how confident are you that you could get advice or information about health or medical topics if you needed it?“ (M = 4.1, SD = 1.0). Perceived discrimination Experiences with everyday discrimination were assessed with a five-item measure on a four-point scale where 0 = “Never,“ 1 = “Rarely,“ and 2 = “Sometimes”; responses of Often,“ “At least once a week,“ and “Almost every day” were categorized as 3. Participants were asked how often they are treated with less courtesy or respect than others, how often they receive poorer services at restaurants or stores, how often people act as if they are afraid of them, how often people act as if they are not smart, and how often they are threatened or harassed. Perceived discrimination was calculated as the mean score of these items (α = 0.91, M = 1.3, SD = 1.0). Sociodemographic characteristics Participant information about age, gender, race, education, marital status, living situation (live alone/live with someone), income, and overall health status was also obtained. COVID-19 mitigation beliefs On a 5-point scale where 1 = “Strongly disagree” and 5 = “Strongly agree,“ participants responded to two items asking how important they thought it was to wear a mask and maintain social distance when going out in public. These two items were combined for a mean score (α = 0.81, M = 4.5, SD = 1.0). To measure the first sought and most trusted sources of information about COVID-19, participants responded to two items (the COVID-19 questions in this study were adapted from a Palliative Care & Supportive Oncology Workgroup Survey and the eHealth Literacy Scale ), (“When you had a strong need to get information about COVID-19, where did you FIRST go to get information?“; “When you had a strong need to get information about COVID-19, which of the following did you find to be the MOST trusted as a source of information about coronavirus or COVID-19?“). For the univariable and multivariable analyses, response options for both items were dichotomized into the following sources: “Mass media” (Internet: Google or another search engine/WebMD or another medical website; Printed materials: newspapers, magazines; Social media: Facebook, Instagram, Twitter; Television) and “Interpersonal” (Conversations with people you trust: friends, relatives, or co-workers; Health care provider: doctor, nurse, social worker). Responses of “Other (Please specify:)” were treated as missing. On a 5-point scale where 1 = “Not confident at all” and 5 = “Completely confident,“ self-efficacy was measured as confidence in obtaining general health information using a single item , “Overall, how confident are you that you could get advice or information about health or medical topics if you needed it?“ (M = 4.1, SD = 1.0). Experiences with everyday discrimination were assessed with a five-item measure on a four-point scale where 0 = “Never,“ 1 = “Rarely,“ and 2 = “Sometimes”; responses of Often,“ “At least once a week,“ and “Almost every day” were categorized as 3. Participants were asked how often they are treated with less courtesy or respect than others, how often they receive poorer services at restaurants or stores, how often people act as if they are afraid of them, how often people act as if they are not smart, and how often they are threatened or harassed. Perceived discrimination was calculated as the mean score of these items (α = 0.91, M = 1.3, SD = 1.0). Participant information about age, gender, race, education, marital status, living situation (live alone/live with someone), income, and overall health status was also obtained. On a 5-point scale where 1 = “Strongly disagree” and 5 = “Strongly agree,“ participants responded to two items asking how important they thought it was to wear a mask and maintain social distance when going out in public. These two items were combined for a mean score (α = 0.81, M = 4.5, SD = 1.0). Multivariable logistic regression models were fitted for the first source of COVID-19 information (mass media vs. interpersonal) and the most trusted source (mass media vs. interpersonal), respectively. Specifically, an odds ratio (OR) larger than 1 indicated higher odds of choosing a mass media source, and an OR smaller than 1 showed higher odds of selecting an interpersonal information source. Univariable logistic regressions were fitted first with factors identified as potentially relevant to COVID-19 information-seeking based on previous research (e.g., ), and factors with p -values less than 0.15 were then considered for multivariable logistic regressions. Backward selection was used to build final multivariable models. Age, race, gender, education, marital status, and overall health status were kept in the multivariable model of the first source of COVID-19 information, while living situation, income, and marital status were kept in the multivariable model of most trusted source of COVID-19 information regardless of their p -values. Multivariable multinomial logistic regression models were also fitted for the first sought and most trusted source of COVID-19 information to look at specific associations between source types, but in a non-aggregated fashion: comparing trusted individuals vs. Internet vs. printed materials vs. social media vs. Television vs. health care providers. Participant characteristics are presented in Table . Average age was about 47 years (M = 46.6, SD = 17.5) with slightly more females (51.1%) than males (48.3%). Participants were primarily White (75.5%), followed by Black (14.8%) and Asian (5.8%). Most participants were college-educated (72.4%). In addition, a majority of participants were non-Hispanic (82.4%). Over half of the participants reported an income of $50,000 or greater (56%). Most participants were married (56.7%), living with someone else (77.6%), and did not live in a rural area (69.2%). Among the overall sample, 61.4% of participants preferred mass media as the first source of COVID-19 information, while the most trusted source was evenly split. RQ1: How are sociodemographic characteristics associated with information seeking about COVID-19? Table presents univariable and multivariable logistic regression estimates for the association between individual characteristics and COVID-19 information-seeking behavior (See Additional file : Appendix for boxplots and bar graphs of significant predictors). Tables and present multivariable multinomial logistic regression estimates that provide a more granulated analysis of information-seeking across source category. Tables , and also present univariable and multivariable logistic regression estimates along with findings from the multinomial analysis to evaluate the association between individual characteristics and COVID-19 information-seeking behavior. Univariable/multivariable logistic model On univariable analysis, characteristics associated with a preference for mass media as the first source of information rather than interpersonal connections were older age (OR: 1.02, p < .01), poor health status (OR: 1.99, p = .05), and stronger beliefs in the importance of masking and social distancing (OR: 1.27, p < .01). Conversely, factors related to a preference for interpersonal communication as an initial source were self-identifying as Black or African American (OR: 0.63, p = < 0.01), self-identifying as male (OR: 0.73, p = < 0.01), and high school education or less (OR: 0.79, p = .04). Further, related to trustworthiness, living with someone else (OR: 0.74, p = .02). Having a higher income level (see Table ) was associated with greater trust in interpersonal sources of COVID-19 information in the univariable model. In the multivariable model, older age and stronger beliefs in the importance of masking and social distancing were independently associated with a preference for mass media as the first source of COVID-19 information. Self-identifying as male and less educational attainment were independently related to increased odds of seeking COVID-19 information first from interpersonal sources. Living with someone else was independently associated with trust in interpersonal rather than mass media sources. Multivariable multinomial logistic model Findings from the multivariable multinomial analysis suggest the preference of older adults for mass media as a first source of information was only significant for printed materials (e.g., newspapers, magazines) (OR: 1.02, p = .04) and television (OR: 1.04, p < .01) when compared to health care providers. There was no specific preference for mass media type based on mitigation beliefs. Also, while there was not a reported preference for interpersonal source based on educational attainment, the Internet (e.g., Google, WebMD) was less preferred as an initial source of COVID-19 information by participants with less formal educational attainment when compared to health care providers (OR: 0.50, p < .01). Similarly, while males were inclined towards interpersonal sources first, there was not a meaningful difference in the preferred interpersonal source type based on gender. However, male participants did report less preference for Internet (OR: 0.70, p = .02) and television (OR: 0.69, p = .04) sources when compared to their health care providers. Regarding the sources most trusted for COVID-19 information, living with someone else was not found to have a significant relationship with a preferred interpersonal source, but printed materials were considered a less trustworthy source of information compared to health care providers for individuals living with another person (OR: 0.47, p = .02). RQ2: How are discrimination and self-efficacy associated with information-seeking about COVID-19? Tables , and also present univariable and multivariable logistic regression estimates for the relationship between self-efficacy, perceived discrimination, and COVID-19 information-seeking behavior. Univariable/multivariable logistic model On univariable analysis, experiences with discrimination (OR: 0.73, p < .01) were related to a preference for interpersonal sources of COVID-19 information. Further, greater confidence in personal health information-seeking ability (self-efficacy) was associated with seeking out interpersonal sources first (OR: 0.87, p = .01) and regarding these sources as more trustworthy (OR: 0.79, p < .001) compared to mass media sources. In the multivariable model, having more experiences with discrimination was independently related to an increased odds of seeking COVID-19 information first from interpersonal sources. Increased self-efficacy was also an independent correlate of both increased preference and trust in interpersonal sources for COVID-19 information compared to mass media. Multivariable multinomial logistic model Results of the multivariable multinomial analysis suggest that individuals with stronger experiences with discrimination preferred to seek out COVID-19 information first from trusted family, relatives, or coworkers (OR: 1.30, p = .05) and printed materials (OR: 1.5, p < .01), but were less likely to seek information first from the Internet (OR: 0.70, p < .01) and television (OR: 0.63, p < .01) compared to their health care provider. There was not a meaningful difference in which interpersonal source participants preferred based on self-efficacy; however, greater efficacy was associated with less preference for the Internet (OR: 0.71, p < .01), social media (OR: 0.66, p = .02), and television (OR: 0.73, p < .01) compared to health care providers. Regarding the most trusted source of COVID-19 information, individuals with greater efficacy had smaller odds of viewing their family, relatives, or coworkers (OR: 0.76, p = .01), Internet (OR: 0.72, p < .01), and social media (OR:0.67, p < .01) as a trustworthy source of information compared to health care providers. Table provides a summary of the study findings. Table presents univariable and multivariable logistic regression estimates for the association between individual characteristics and COVID-19 information-seeking behavior (See Additional file : Appendix for boxplots and bar graphs of significant predictors). Tables and present multivariable multinomial logistic regression estimates that provide a more granulated analysis of information-seeking across source category. Tables , and also present univariable and multivariable logistic regression estimates along with findings from the multinomial analysis to evaluate the association between individual characteristics and COVID-19 information-seeking behavior. Univariable/multivariable logistic model On univariable analysis, characteristics associated with a preference for mass media as the first source of information rather than interpersonal connections were older age (OR: 1.02, p < .01), poor health status (OR: 1.99, p = .05), and stronger beliefs in the importance of masking and social distancing (OR: 1.27, p < .01). Conversely, factors related to a preference for interpersonal communication as an initial source were self-identifying as Black or African American (OR: 0.63, p = < 0.01), self-identifying as male (OR: 0.73, p = < 0.01), and high school education or less (OR: 0.79, p = .04). Further, related to trustworthiness, living with someone else (OR: 0.74, p = .02). Having a higher income level (see Table ) was associated with greater trust in interpersonal sources of COVID-19 information in the univariable model. In the multivariable model, older age and stronger beliefs in the importance of masking and social distancing were independently associated with a preference for mass media as the first source of COVID-19 information. Self-identifying as male and less educational attainment were independently related to increased odds of seeking COVID-19 information first from interpersonal sources. Living with someone else was independently associated with trust in interpersonal rather than mass media sources. Multivariable multinomial logistic model Findings from the multivariable multinomial analysis suggest the preference of older adults for mass media as a first source of information was only significant for printed materials (e.g., newspapers, magazines) (OR: 1.02, p = .04) and television (OR: 1.04, p < .01) when compared to health care providers. There was no specific preference for mass media type based on mitigation beliefs. Also, while there was not a reported preference for interpersonal source based on educational attainment, the Internet (e.g., Google, WebMD) was less preferred as an initial source of COVID-19 information by participants with less formal educational attainment when compared to health care providers (OR: 0.50, p < .01). Similarly, while males were inclined towards interpersonal sources first, there was not a meaningful difference in the preferred interpersonal source type based on gender. However, male participants did report less preference for Internet (OR: 0.70, p = .02) and television (OR: 0.69, p = .04) sources when compared to their health care providers. Regarding the sources most trusted for COVID-19 information, living with someone else was not found to have a significant relationship with a preferred interpersonal source, but printed materials were considered a less trustworthy source of information compared to health care providers for individuals living with another person (OR: 0.47, p = .02). On univariable analysis, characteristics associated with a preference for mass media as the first source of information rather than interpersonal connections were older age (OR: 1.02, p < .01), poor health status (OR: 1.99, p = .05), and stronger beliefs in the importance of masking and social distancing (OR: 1.27, p < .01). Conversely, factors related to a preference for interpersonal communication as an initial source were self-identifying as Black or African American (OR: 0.63, p = < 0.01), self-identifying as male (OR: 0.73, p = < 0.01), and high school education or less (OR: 0.79, p = .04). Further, related to trustworthiness, living with someone else (OR: 0.74, p = .02). Having a higher income level (see Table ) was associated with greater trust in interpersonal sources of COVID-19 information in the univariable model. In the multivariable model, older age and stronger beliefs in the importance of masking and social distancing were independently associated with a preference for mass media as the first source of COVID-19 information. Self-identifying as male and less educational attainment were independently related to increased odds of seeking COVID-19 information first from interpersonal sources. Living with someone else was independently associated with trust in interpersonal rather than mass media sources. Findings from the multivariable multinomial analysis suggest the preference of older adults for mass media as a first source of information was only significant for printed materials (e.g., newspapers, magazines) (OR: 1.02, p = .04) and television (OR: 1.04, p < .01) when compared to health care providers. There was no specific preference for mass media type based on mitigation beliefs. Also, while there was not a reported preference for interpersonal source based on educational attainment, the Internet (e.g., Google, WebMD) was less preferred as an initial source of COVID-19 information by participants with less formal educational attainment when compared to health care providers (OR: 0.50, p < .01). Similarly, while males were inclined towards interpersonal sources first, there was not a meaningful difference in the preferred interpersonal source type based on gender. However, male participants did report less preference for Internet (OR: 0.70, p = .02) and television (OR: 0.69, p = .04) sources when compared to their health care providers. Regarding the sources most trusted for COVID-19 information, living with someone else was not found to have a significant relationship with a preferred interpersonal source, but printed materials were considered a less trustworthy source of information compared to health care providers for individuals living with another person (OR: 0.47, p = .02). Tables , and also present univariable and multivariable logistic regression estimates for the relationship between self-efficacy, perceived discrimination, and COVID-19 information-seeking behavior. Univariable/multivariable logistic model On univariable analysis, experiences with discrimination (OR: 0.73, p < .01) were related to a preference for interpersonal sources of COVID-19 information. Further, greater confidence in personal health information-seeking ability (self-efficacy) was associated with seeking out interpersonal sources first (OR: 0.87, p = .01) and regarding these sources as more trustworthy (OR: 0.79, p < .001) compared to mass media sources. In the multivariable model, having more experiences with discrimination was independently related to an increased odds of seeking COVID-19 information first from interpersonal sources. Increased self-efficacy was also an independent correlate of both increased preference and trust in interpersonal sources for COVID-19 information compared to mass media. Multivariable multinomial logistic model Results of the multivariable multinomial analysis suggest that individuals with stronger experiences with discrimination preferred to seek out COVID-19 information first from trusted family, relatives, or coworkers (OR: 1.30, p = .05) and printed materials (OR: 1.5, p < .01), but were less likely to seek information first from the Internet (OR: 0.70, p < .01) and television (OR: 0.63, p < .01) compared to their health care provider. There was not a meaningful difference in which interpersonal source participants preferred based on self-efficacy; however, greater efficacy was associated with less preference for the Internet (OR: 0.71, p < .01), social media (OR: 0.66, p = .02), and television (OR: 0.73, p < .01) compared to health care providers. Regarding the most trusted source of COVID-19 information, individuals with greater efficacy had smaller odds of viewing their family, relatives, or coworkers (OR: 0.76, p = .01), Internet (OR: 0.72, p < .01), and social media (OR:0.67, p < .01) as a trustworthy source of information compared to health care providers. Table provides a summary of the study findings. On univariable analysis, experiences with discrimination (OR: 0.73, p < .01) were related to a preference for interpersonal sources of COVID-19 information. Further, greater confidence in personal health information-seeking ability (self-efficacy) was associated with seeking out interpersonal sources first (OR: 0.87, p = .01) and regarding these sources as more trustworthy (OR: 0.79, p < .001) compared to mass media sources. In the multivariable model, having more experiences with discrimination was independently related to an increased odds of seeking COVID-19 information first from interpersonal sources. Increased self-efficacy was also an independent correlate of both increased preference and trust in interpersonal sources for COVID-19 information compared to mass media. Results of the multivariable multinomial analysis suggest that individuals with stronger experiences with discrimination preferred to seek out COVID-19 information first from trusted family, relatives, or coworkers (OR: 1.30, p = .05) and printed materials (OR: 1.5, p < .01), but were less likely to seek information first from the Internet (OR: 0.70, p < .01) and television (OR: 0.63, p < .01) compared to their health care provider. There was not a meaningful difference in which interpersonal source participants preferred based on self-efficacy; however, greater efficacy was associated with less preference for the Internet (OR: 0.71, p < .01), social media (OR: 0.66, p = .02), and television (OR: 0.73, p < .01) compared to health care providers. Regarding the most trusted source of COVID-19 information, individuals with greater efficacy had smaller odds of viewing their family, relatives, or coworkers (OR: 0.76, p = .01), Internet (OR: 0.72, p < .01), and social media (OR:0.67, p < .01) as a trustworthy source of information compared to health care providers. Table provides a summary of the study findings. The purpose of this study was to explore factors associated with audience preferences (first sought, most trusted) for COVID-19 information to inform the development of tailored health communication strategies. The current work adds to literature on HISB during the COVID-19 pandemic by providing evidence for the relationship between sociodemographics and source trust first proposed by Ali et al. , and extends by demonstrating how information sources, notably those first sought, are related to discrimination and information efficacy. Sociodemographics driving COVID-19 information-seeking Age One key finding is that mass media outlets, specifically print materials (e.g., newspapers, magazines) and TV, were preferred as initial sources for COVID-19 information for older participants. The elderly are particularly vulnerable to becoming severely ill from COVID-19, increasing the urgency for tailored communication strategies . This preference for mass media as initial sources of information conflicts with previous findings suggesting that older adults rely on interpersonal sources such as health care providers and family members, not only for information but also to satisfy emotional needs stemming from social isolation during the pandemic . One rationale for this inconsistency might be that the COVID-19 pandemic morphed into a political wedge issue in which risk perceptions, conspiracy beliefs, and responses to government recommendations were demarcated along partisan lines . As a result, older adults might have sought information from their political echo chambers (e.g., cable news networks) rather than other sources such as government websites or health care providers . Another explanation is that the novelty of the SARS-CoV-2 virus and the associated uncertainty, fear, and confusion limited the value of interpersonal discussions, prompting information to be sought elsewhere. It is worth noting that a large portion of this sample was college-educated, and other factors including health status may have contributed to this finding; individuals with chronic conditions may access COVID-19 information more often through the mass media but have less trust in these sources . The interaction of age and health status on COVID-19 information-seeking is an area of future study. Education Another key finding was that communication with interpersonal sources was preferred as a primary resource for information by those with lower levels of educational attainment. Further analysis revealed that there was not a significant difference in preference of first information source for participants with less formal education between preferring friends/relatives/co-workers or health care providers. However, the Internet was a less preferred source compared to health care providers for these participants, suggesting that providers can be targeted for campaigns aimed at this group. Studies of education level and COVID-19 misinformation have reported relationships with a multitude of factors, including lower confidence in government and scientific institutions as well as lower perceived infection risk . However, previous research suggests that those with less formal education may perceive a greater risk of dying from COVID-19 and experience greater economic consequences because of the pandemic ; it is possible that this increased risk prompts information-seeking from professional sources. Further, individuals with lower levels of educational attainment are more likely to have reduced health literacy, and these individuals may instead turn to their doctors for information . Educational attainment has been found to positively correlate with a diversity of sources , furthering the argument that education level is a barrier to information-seeking through mass media. Mitigation beliefs Participants with weaker beliefs in the importance of masking and social distancing when in public were more likely to seek out COVID-19 information through their interpersonal contacts first, regardless of the source. Individuals with strong doubts about the effectiveness of masking and social distancing are less prone to seek knowledge through external mass media channels, particularly when there is evolving information . This finding offers confirming evidence for previous research demonstrating a significant relationship between COVID-19 information-seeking and adherence to mitigation strategies . Given the politicization and polarization of the pandemic, those more skeptical of mitigation strategies would be more likely to look for information within their interpersonal networks rather than a media system that is viewed as biased . These individuals may be challenging to target with health communication campaigns. However, given the demonstrated direct relationship between COVID-19 information seeking and preventive behavior , there is a pressing need for evidence-based efforts. Discrimination and self-efficacy driving COVID-19 information-seeking Discrimination Individuals reporting more common experiences with discrimination also described a greater preference for interpersonal contacts as an initial source for COVID-19 information, specifically friends, relatives, and co-workers. Discrimination can cause a delay in seeking medical care, including cancer screenings , and significantly increases stress response . One speculation for this finding is that while mass media may be used as a means of coping with the stress that comes along with mistreatment, information exposure during a health crisis such as COVID-19 can intensify feelings of stress, leading to avoidance . Given the high levels of discrimination reported during the pandemic , these groups may find it less distressing to receive information from trusted interpersonal sources, particularly those that share similar demographic backgrounds . Additional research is needed to disentangle the effect of different sources of discrimination (e.g., gender, race, ethnicity) on information-seeking about COVID-19 . Self-efficacy Finally, this work also found that individuals with greater confidence in their ability to obtain health information preferred to seek out interpersonal sources first, with a particularly lower preference for the Internet, social media, and TV compared to their health care provider. Participants with greater efficacy also found interpersonal sources to be more trustworthy, yet maintained a lower perception of trustworthiness for friends, relatives, and family compared to health care providers. Individuals tend to make determinations on whether to engage in information-seeking by evaluating three types of efficacies: communication efficacy (whether the individual has the skill to seek information), target efficacy (whether their interpersonal source has the knowledge and is willing to share it), and coping efficacy (whether the individual can emotionally deal with the information) . Thus, individuals with greater efficacy may feel more confidence in their ability to seek information from interpersonal sources based on their communication skills, beliefs that their interpersonal sources have reliable information, and beliefs that they can cope with the information potentially shared. Interpersonal sources may also help calm the often overwhelming “noise” of competing and emerging information shared by media channels. Individuals who are confident in obtaining health information are also more likely to experience feelings such as fatalism when they experience challenges and frustrations in seeking this information . Therefore, individuals with increased self-efficacy in their HISB may prefer to engage with interpersonal sources rather than mass media to attenuate the uncertainty associated with this massive influx of information. Age One key finding is that mass media outlets, specifically print materials (e.g., newspapers, magazines) and TV, were preferred as initial sources for COVID-19 information for older participants. The elderly are particularly vulnerable to becoming severely ill from COVID-19, increasing the urgency for tailored communication strategies . This preference for mass media as initial sources of information conflicts with previous findings suggesting that older adults rely on interpersonal sources such as health care providers and family members, not only for information but also to satisfy emotional needs stemming from social isolation during the pandemic . One rationale for this inconsistency might be that the COVID-19 pandemic morphed into a political wedge issue in which risk perceptions, conspiracy beliefs, and responses to government recommendations were demarcated along partisan lines . As a result, older adults might have sought information from their political echo chambers (e.g., cable news networks) rather than other sources such as government websites or health care providers . Another explanation is that the novelty of the SARS-CoV-2 virus and the associated uncertainty, fear, and confusion limited the value of interpersonal discussions, prompting information to be sought elsewhere. It is worth noting that a large portion of this sample was college-educated, and other factors including health status may have contributed to this finding; individuals with chronic conditions may access COVID-19 information more often through the mass media but have less trust in these sources . The interaction of age and health status on COVID-19 information-seeking is an area of future study. Education Another key finding was that communication with interpersonal sources was preferred as a primary resource for information by those with lower levels of educational attainment. Further analysis revealed that there was not a significant difference in preference of first information source for participants with less formal education between preferring friends/relatives/co-workers or health care providers. However, the Internet was a less preferred source compared to health care providers for these participants, suggesting that providers can be targeted for campaigns aimed at this group. Studies of education level and COVID-19 misinformation have reported relationships with a multitude of factors, including lower confidence in government and scientific institutions as well as lower perceived infection risk . However, previous research suggests that those with less formal education may perceive a greater risk of dying from COVID-19 and experience greater economic consequences because of the pandemic ; it is possible that this increased risk prompts information-seeking from professional sources. Further, individuals with lower levels of educational attainment are more likely to have reduced health literacy, and these individuals may instead turn to their doctors for information . Educational attainment has been found to positively correlate with a diversity of sources , furthering the argument that education level is a barrier to information-seeking through mass media. Mitigation beliefs Participants with weaker beliefs in the importance of masking and social distancing when in public were more likely to seek out COVID-19 information through their interpersonal contacts first, regardless of the source. Individuals with strong doubts about the effectiveness of masking and social distancing are less prone to seek knowledge through external mass media channels, particularly when there is evolving information . This finding offers confirming evidence for previous research demonstrating a significant relationship between COVID-19 information-seeking and adherence to mitigation strategies . Given the politicization and polarization of the pandemic, those more skeptical of mitigation strategies would be more likely to look for information within their interpersonal networks rather than a media system that is viewed as biased . These individuals may be challenging to target with health communication campaigns. However, given the demonstrated direct relationship between COVID-19 information seeking and preventive behavior , there is a pressing need for evidence-based efforts. One key finding is that mass media outlets, specifically print materials (e.g., newspapers, magazines) and TV, were preferred as initial sources for COVID-19 information for older participants. The elderly are particularly vulnerable to becoming severely ill from COVID-19, increasing the urgency for tailored communication strategies . This preference for mass media as initial sources of information conflicts with previous findings suggesting that older adults rely on interpersonal sources such as health care providers and family members, not only for information but also to satisfy emotional needs stemming from social isolation during the pandemic . One rationale for this inconsistency might be that the COVID-19 pandemic morphed into a political wedge issue in which risk perceptions, conspiracy beliefs, and responses to government recommendations were demarcated along partisan lines . As a result, older adults might have sought information from their political echo chambers (e.g., cable news networks) rather than other sources such as government websites or health care providers . Another explanation is that the novelty of the SARS-CoV-2 virus and the associated uncertainty, fear, and confusion limited the value of interpersonal discussions, prompting information to be sought elsewhere. It is worth noting that a large portion of this sample was college-educated, and other factors including health status may have contributed to this finding; individuals with chronic conditions may access COVID-19 information more often through the mass media but have less trust in these sources . The interaction of age and health status on COVID-19 information-seeking is an area of future study. Another key finding was that communication with interpersonal sources was preferred as a primary resource for information by those with lower levels of educational attainment. Further analysis revealed that there was not a significant difference in preference of first information source for participants with less formal education between preferring friends/relatives/co-workers or health care providers. However, the Internet was a less preferred source compared to health care providers for these participants, suggesting that providers can be targeted for campaigns aimed at this group. Studies of education level and COVID-19 misinformation have reported relationships with a multitude of factors, including lower confidence in government and scientific institutions as well as lower perceived infection risk . However, previous research suggests that those with less formal education may perceive a greater risk of dying from COVID-19 and experience greater economic consequences because of the pandemic ; it is possible that this increased risk prompts information-seeking from professional sources. Further, individuals with lower levels of educational attainment are more likely to have reduced health literacy, and these individuals may instead turn to their doctors for information . Educational attainment has been found to positively correlate with a diversity of sources , furthering the argument that education level is a barrier to information-seeking through mass media. Participants with weaker beliefs in the importance of masking and social distancing when in public were more likely to seek out COVID-19 information through their interpersonal contacts first, regardless of the source. Individuals with strong doubts about the effectiveness of masking and social distancing are less prone to seek knowledge through external mass media channels, particularly when there is evolving information . This finding offers confirming evidence for previous research demonstrating a significant relationship between COVID-19 information-seeking and adherence to mitigation strategies . Given the politicization and polarization of the pandemic, those more skeptical of mitigation strategies would be more likely to look for information within their interpersonal networks rather than a media system that is viewed as biased . These individuals may be challenging to target with health communication campaigns. However, given the demonstrated direct relationship between COVID-19 information seeking and preventive behavior , there is a pressing need for evidence-based efforts. Discrimination Individuals reporting more common experiences with discrimination also described a greater preference for interpersonal contacts as an initial source for COVID-19 information, specifically friends, relatives, and co-workers. Discrimination can cause a delay in seeking medical care, including cancer screenings , and significantly increases stress response . One speculation for this finding is that while mass media may be used as a means of coping with the stress that comes along with mistreatment, information exposure during a health crisis such as COVID-19 can intensify feelings of stress, leading to avoidance . Given the high levels of discrimination reported during the pandemic , these groups may find it less distressing to receive information from trusted interpersonal sources, particularly those that share similar demographic backgrounds . Additional research is needed to disentangle the effect of different sources of discrimination (e.g., gender, race, ethnicity) on information-seeking about COVID-19 . Self-efficacy Finally, this work also found that individuals with greater confidence in their ability to obtain health information preferred to seek out interpersonal sources first, with a particularly lower preference for the Internet, social media, and TV compared to their health care provider. Participants with greater efficacy also found interpersonal sources to be more trustworthy, yet maintained a lower perception of trustworthiness for friends, relatives, and family compared to health care providers. Individuals tend to make determinations on whether to engage in information-seeking by evaluating three types of efficacies: communication efficacy (whether the individual has the skill to seek information), target efficacy (whether their interpersonal source has the knowledge and is willing to share it), and coping efficacy (whether the individual can emotionally deal with the information) . Thus, individuals with greater efficacy may feel more confidence in their ability to seek information from interpersonal sources based on their communication skills, beliefs that their interpersonal sources have reliable information, and beliefs that they can cope with the information potentially shared. Interpersonal sources may also help calm the often overwhelming “noise” of competing and emerging information shared by media channels. Individuals who are confident in obtaining health information are also more likely to experience feelings such as fatalism when they experience challenges and frustrations in seeking this information . Therefore, individuals with increased self-efficacy in their HISB may prefer to engage with interpersonal sources rather than mass media to attenuate the uncertainty associated with this massive influx of information. Individuals reporting more common experiences with discrimination also described a greater preference for interpersonal contacts as an initial source for COVID-19 information, specifically friends, relatives, and co-workers. Discrimination can cause a delay in seeking medical care, including cancer screenings , and significantly increases stress response . One speculation for this finding is that while mass media may be used as a means of coping with the stress that comes along with mistreatment, information exposure during a health crisis such as COVID-19 can intensify feelings of stress, leading to avoidance . Given the high levels of discrimination reported during the pandemic , these groups may find it less distressing to receive information from trusted interpersonal sources, particularly those that share similar demographic backgrounds . Additional research is needed to disentangle the effect of different sources of discrimination (e.g., gender, race, ethnicity) on information-seeking about COVID-19 . Finally, this work also found that individuals with greater confidence in their ability to obtain health information preferred to seek out interpersonal sources first, with a particularly lower preference for the Internet, social media, and TV compared to their health care provider. Participants with greater efficacy also found interpersonal sources to be more trustworthy, yet maintained a lower perception of trustworthiness for friends, relatives, and family compared to health care providers. Individuals tend to make determinations on whether to engage in information-seeking by evaluating three types of efficacies: communication efficacy (whether the individual has the skill to seek information), target efficacy (whether their interpersonal source has the knowledge and is willing to share it), and coping efficacy (whether the individual can emotionally deal with the information) . Thus, individuals with greater efficacy may feel more confidence in their ability to seek information from interpersonal sources based on their communication skills, beliefs that their interpersonal sources have reliable information, and beliefs that they can cope with the information potentially shared. Interpersonal sources may also help calm the often overwhelming “noise” of competing and emerging information shared by media channels. Individuals who are confident in obtaining health information are also more likely to experience feelings such as fatalism when they experience challenges and frustrations in seeking this information . Therefore, individuals with increased self-efficacy in their HISB may prefer to engage with interpersonal sources rather than mass media to attenuate the uncertainty associated with this massive influx of information. Audience segmentation refers to the process of dividing an audience into definable, measurable groups to create messaging that is responsive to specific population needs . This approach to message design can significantly impact engagement, as well as attitude and behavior change and is thus considered an essential piece of tailored communication strategies already applied to COVID-19 messaging . Findings from this study have meaningful implications for future practice through the identified audience variations regarding information-seeking preferences. These results can be leveraged to enhance the capability of specific target audiences to engage with evidence-based COVID-19 information. The politicization of COVID-19 and its influence on health inequalities, along with the rapid and uneven pace of information dissemination on COVID-19 guidelines, has been a challenge for effective health communication . Thus, health communication campaigns that can efficiently identify strategies to reach various audiences in a targeted manner will have increased effectiveness. The following guidelines should be priority considerations when developing audience-focused COVID-19 information campaigns: Understand the unique contexts of the intended audience, including the influence of societal inequalities on information-seeking behavior. Taking a user-centered approach to campaign design that actively seeks out and incorporates feedback will ensure that the preferences, needs, and values of the target audience are fully understood. This approach will also enhance campaign acceptability while reducing the effort required to engage with its components, all of which will increase efficacy. We offer the following specific recommendations for campaign tailoring based on the findings of this study: Campaigns targeting older adults should develop materials for dissemination through television and print. When developing campaigns targeting individuals with less formal educational attainment, include medical professionals. Incorporating close social ties (i.e., friends, relatives, and co-workers) may increase the effectiveness of campaigns targeting groups experiencing discrimination. Audiences with greater efficacy can be effectively targeted through their health care provider, whereas those with weaker beliefs in their ability to obtain health information can be better reached through the Internet (e.g., WebMD) and social media. 2. In addition to examining the “what” and “how” of message dissemination, the “where” and “who” should also be carefully considered. Theoretical frameworks such as diffusion of innovations and social influence can serve as starting points to further understand the influence of social networks and source credibility in information-seeking. Building capacity to bring these campaigns to scale will also be required and can be facilitated through the development of diverse collaborations that include community members and other stakeholders. 3. Lastly, consider the context of the topic and understand that source preferences for information may vary when the topics change, particularly given the political climate (e.g., COVID-19 information seeking may be very different than cancer screening). Campaign development should be iterative and agile in order to adapt to the fluidity inherent to these politically charged health topics, with systems in place for ongoing evaluation. Strengths and limitations This study adds to the literature on information-seeking about COVID-19 through the examination of sources of COVID-19 information most likely to be sought first and the exploration of the role of discrimination and self-efficacy on source preference (i.e., first sought and most trusted). The findings also offer support for previous research on the influence of sociodemographic factors in HISB. This study is not without limitations. The measures of information sources may contain within-group differences (e.g., different social media platforms such as Facebook and Twitter are often used in different ways). Yet, this study provides compelling evidence for HISB during the pandemic and how individuals can be targeted with persuasive messaging. Also, while this online survey asked only for the FIRST preferred source or the MOST trusted, communication does not occur in a vacuum. Mass and interpersonal methods of communication are becoming increasingly intermingled , and factors such as authority (e.g., government websites and health care providers) might play a role . Additional research is needed to build information-seeking models of increasing complexity surrounding the interplay of these factors. This study adds to the literature on information-seeking about COVID-19 through the examination of sources of COVID-19 information most likely to be sought first and the exploration of the role of discrimination and self-efficacy on source preference (i.e., first sought and most trusted). The findings also offer support for previous research on the influence of sociodemographic factors in HISB. This study is not without limitations. The measures of information sources may contain within-group differences (e.g., different social media platforms such as Facebook and Twitter are often used in different ways). Yet, this study provides compelling evidence for HISB during the pandemic and how individuals can be targeted with persuasive messaging. Also, while this online survey asked only for the FIRST preferred source or the MOST trusted, communication does not occur in a vacuum. Mass and interpersonal methods of communication are becoming increasingly intermingled , and factors such as authority (e.g., government websites and health care providers) might play a role . Additional research is needed to build information-seeking models of increasing complexity surrounding the interplay of these factors. The COVID-19 pandemic has weakened the US economy and led to tremendous life loss, and the uptake of protective measures is lacking due in part to false information being circulated within the media and personal networks. The findings of this study contribute to our understanding of how people are seeking out information about COVID-19 during the pandemic, which will allow for the development of evidence-based dissemination strategies. As information-seeking increases during the pandemic, exposure to risk information can have a direct tie to behavior, and the results of this study suggest that even with such a wide diversity of digital information sources and the capacity for scalable health communication campaigns that maximizing efforts to involve interpersonal connections may be preferable for some individuals. This idea is even more relevant during the current infodemic, where mass media channels have, in many ways, been corrupted by misinformation. By considering the audience factors illuminated in this study, researchers and practitioners become better equipped to deliver messaging through the sources and channels that are highly sought and trusted. Additional file 1. Study questionnaire. Additional file 2: Appendix 1. Boxplots and bar graphs for predictors of COVID-19 information-seeking. |
Facing Osteoporosis: Is Hormonal Therapy Losing an Opportunity to be Used? The Role of Gynecologists | 23e95e6f-d0d9-4469-8275-3fe73d569bc3 | 9800066 | Gynaecology[mh] | |
Analyzing Pathophysiology and Immune Cells and Their Cytokines and Mediators in Precision‐Cut Slices of the Murine Lung | 302af829-4e3d-4645-8334-d24b5e5f7c1c | 11773432 | Anatomy[mh] | Monitoring immune responses of the murine airway is an essential step in immunological research, showing the dynamic changes that occur in disease. Precision‐cut lung slices (PCLSs) provide a technique for studying local immune function and also preserving the architecture of the lung tissue. The injection of low‐melting‐point agarose into the murine lung enables precise cutting into thin reproducible slices. Use of PCLSs is a sustainable and replicable method to study dimensions of both healthy and diseased states. The consistent product of this approach has led to the standardization of the protocol in our laboratory and has helped us better study and understand models of asthma, fibrosis, respiratory virus infection, and lung cancer (Cervenkova et al., ; Koziol‐White et al., ; Richter et al., ; Sewald & Danov, ; Springer & Fischer, ; Stegmayr & Wagner, ). Use of PCLSs supports research in more pathophysiological conditions and allows the observation of therapeutic effects on an intact lung and immune cells and their cytokines and mediators, which could improve preclinical investigations. A murine model of asthma includes sensitization and contact with allergens, and in a lung cancer model, LL/2‐luc‐M38 cells containing the luciferin promoter are injected into the tail vein of the mouse. In both of these models, different therapeutic settings can be analyzed and compared directly in the same lung at the histological, cellular, and molecular levels. These analyses encompass both dynamic observations, such as observation of interactions among different cell types and their cytokine and mediator release, as well as static evaluations, including analysis of pathological changes like collagen deposition and cellular damage within the intact lung architecture. Using PCLSs allows both powerful ex vivo and in vivo analyses of a selected therapy. A critical advantage is the possibility to generate several different read‐outs from a single mouse, significantly reducing the number of animals required, which is consistent with the Reduction principle of the 3Rs (Replacement, Reduction, Refinement) in animal research. Thus, PCLSs are a more ethical and efficient method of analysis in comparison to other conventional methods, such as preparation of a whole‐lung cell suspension from a single murine lung, used typically for flow cytometry, cell culture, and RNA analysis, and also use of organoids. In this article, we describe how to obtain PCLSs from a murine model of lung cancer, starting from LL/2‐luc‐M38 cell pretreatment, as indicated in Support Protocol followed by Support Protocol . At the end of Support Protocol , PCLSs can be generated as described in the Basic Protocol. After that, PCLSs can be cultured with different stimuli. At the end of the PCLS culture, the slices can be analyzed histologically as described in Support Protocol and supernatants analyzed for cytokines as described in Support Protocol . Supernatants can be analyzed for the presence of cytotoxic mediators as described in Support Protocol . Support Protocol details how to analyze the regulation of all genes at the RNA level in the PCLSs. Moreover, to analyze immune cells modulating and infiltrating the PCLSs, Support Protocol should be followed after the Basic Protocol. NOTE : All protocols involving animals must be reviewed and approved by the appropriate Animal Care and Use Committee and must follow regulations for the care and use of laboratory animals. Here, all experiments were undertaken with an approved license (no. Az 55.2‐2532‐2‐1286‐20) from the Ethical Committee of the Regierung Oberfranken (Erlangen, Germany). NOTE : All solutions and equipment coming into contact with cells must be sterile, and proper sterile technique should be used accordingly. NOTE : All culture incubations are performed in a humidified 37°C, 5% CO 2 incubator unless otherwise specified This protocol allows preparation of PCLSs for histopathologic and physiologic analysis of the murine lung after exposure to different stimuli. Materials Low‐melting‐point agarose (Roth, cat. no. 6351.1) Phosphate‐buffered saline (PBS), without calcium or magnesium (Gibco‐Invitrogen, cat. no. 14190), room temperature and 4°C Pentobarbital sodium (WDT, Garbsen, ZI‐N: 400883.00.00, PZN: 4955173; Release: 300 mg/ml pentobarbital sodium) 70% (v/v) ethanol, 99.8% pure (AppliChem, cat. no. 64‐17‐5) Mice PCLS medium (see recipe), 37°C Clinically relevant stimuli (e.g., anti‐CD3/CD28 or anti‐PD1 antibody stimulation) 50‐ml conical tubes Glass beaker Microwave 37°C water bath 1‐ml syringes (Syringe, Omnican ® ‐F, 1 ml, B. Braun, cat. no. 9161502) Tissue slicer (Alabama R&D, TSE, model no. 6000; Fig. ) Blades Surgical instruments (including anatomical scissors, spatula, scalpel, and tweezers; see Fig. and ) Medical yarn Dissecting board with pins IV catheter (i.v. Catheter 24G, Jelco ® , Smiths Medical, cat. no. 4013, lot no. 6027133) Petri dishes Tissue coring plunger, 8 mm (TSE) Tissue coring press (TSE, model no. MD5000; optional) Cell culture plates Shaker (IKA Rocker 3D digital, model no. 100101072) CAUTION : Pentobarbital sodium is a potent anesthetic drug with strong effects on the human central nervous system. At elevated doses, pentobarbital functions as an anticonvulsant for emergent seizure control and for inducing medically induced comas. Thus, all laboratory personnel must be informed of its impact and about safety precautions, such as wearing protective eyewear, gloves, and a face mask. See the hazard sheet for further details. Low‐melting‐point agarose gel preparation (5 min; Fig. and ) 1 Place the low‐melting‐point agarose on ice before preparation. 2 Transfer 250 mg low‐melting‐point agarose into a 50‐ml conical tube. 3 Dissolve into 10 ml PBS to make a 2.5% (w/v) solution. 4 Place the tube with agarose solution in a glass beaker and carefully heat up in a microwave. 5 Let the agarose solution cool down to 37°C. 6 Place in a 37°C water bath during the preparation of the lung and take the agarose out of the water bath before injection. Tissue slicer setup (10 min) 7 Secure the tissue slicer on a stable surface. 8 Insert a new clean, sharp blade into the blade holder. 9 Adjust the blade angle according to the manufacturer's guidelines (typically a slight angle to reduce tissue tearing). 10 Fill the chamber with a bottle (500 ml) of cold PBS. 11 Ensure the chamber is securely attached to the tissue slicer. 12 Set the cutting speed (usually low to moderate). 13 Adjust slice thickness (180 to 250 µm for lung slices). Preparation of other equipment and reagents (5 min) 14 Prepare several small surgical instruments, as shown in Fig. and . 15 Label 50‐ml conical tubes corresponding to each mouse, fill each with 20 ml cold PBS, and place on ice (Fig. and ). 16 Cut medical yarn into pieces of 20 cm each per mouse. 17 Prepare 300 mg/ml pentobarbital sodium working solution in PBS for anesthetizing the mice. 18 Clean the table where the surgery will be performed with a soap solution and sterilize it with 70% ethanol. Induction of anesthesia (5 min) 19 Inject each mouse intraperitoneally (i.p.) with 200 µl of 300 mg/ml pentobarbital sodium (22.5 mg pentobarbital; see step 17), depending on the weight of the mouse (Fig. ). The 200 µl of 300 mg/ml working solution combines 75 µl of an 800 mg/ml stock solution and 125 µl PBS. The exact mass of substance (600 mg/kg body weight) injected is adapted depending on the weight of the mouse (Maxeiner et al., ). The amount of 300 mg/ml pentobarbital sodium that we apply in this phase is sufficient to perform the surgical dissection. Pinch the foot of the mouse with tweezers to check the surgical tolerance of the mouse after injection. Dissection (10 min) 20 Place a mouse on its back and secure it to a dissecting board with pins with its limbs spread out. Disinfect the fur of the mouse with 70% ethanol. 21 Make an incision onto the lower abdomen of the mouse and pull the fur with the skin up above the head until the organs are revealed. Secure the skin to the dissecting board. 22 Carefully remove the glandular tissue by pulling it to the head until the trachea is seen. Free the trachea by removing any connective tissue. Dissection of the trachea should be started immediately after surgery begins to avoid changes in the airway environment (Fig. ). 23 Fix the trachea on a flat surface by carefully pushing a spatula under the trachea. Prepare the medical yarn (see step 16) by wrapping it around the trachea loosely, so it can be tightened rapidly after injection. 24 Cut open the chest by carefully holding the sternum with a pair of tweezers and cutting along the outer edge of the rib cage, trying not to damage the lung. IMPORTANT NOTE : Do not remove the rib cage entirely because the lung can be easily injured. Insertion of IV catheter and injection of agarose (10 min) 25 CRITICAL STEP: Carefully insert the IV catheter into the trachea. When the catheter is in place, delicately pull the needle out and slide the cannula into the trachea until resistance is sensed. Then, pull it a few millimeters back out to ensure its correct position and to avoid the backflow of the agarose gel (Fig. and ). 26 Fill a 1‐ml syringe with 1 ml of 2.5% low‐melting‐point agarose from step 6, carefully insert it into the cannula, and slowly inject the solution. Carefully observe if the agarose gel is entering the lung or not. If it flows outside, stop and arrange the equipment again (Fig. ). Surgical removal of the lungs (20 min) 27 Rapidly bind the trachea at the insertion point with the medical yarn, but not tightly (Fig. ), and place a glove filled with ice on the lungs for 10 to 15 min to allow the agar to solidify (Fig. ). 28 Cut the rib cage along the sides to allow it to be lifted or removed and then carefully dissect the lung lobes. 29 Transfer the entire set of lungs into a tube of cold PBS (see step 15). Now the lung is ready to be subjected to tissue slicing (Fig. ). Tissue slicing (30 min) 30 Place an agarose‐embedded lung lobe into a petri dish filled with cold PBS (Fig. ). 31 Adjust the settings of the tissue slicer (prepared in steps 7 to 13) for optimal cutting (e.g., 180‐ to 250‐µm section thickness by using a core unit of 0.5‐ or 0.8‐mm diameter). 32 Punch the biopsy (Fig. ), adjust the tissue coring plunger/insert set (Fig. ), and insert the tissue block (Fig. ). 33 If available, fix the tissue coring plunger into the tissue coring press and make PCLSs (see Video ). Otherwise, install the plunger/insert set into the tissue slicer chamber (Fig. and ) and start the slicing process, ensuring the blade moves smoothly through the tissue. Collect the slices in cold PBS to maintain tissue viability (Fig. and ). Handling of lung slices (1 to 2 hr) 34 Transfer the lung slices to cell culture plates containing PBS or PCLS medium. 35 Remove excessive agarose by incubating the PCLSs in PBS or PCLS medium three times for 30 min each on a shaker in a humidified 37°C, 5% CO 2 cell incubator. 36 After the washing process, proceed with treatment of the lung slices with the chosen clinically relevant stimuli (e.g., anti‐CD3/CD28 or anti‐PD1 antibody stimulation) in the PCLS medium and maintain the PCLS culture in the humidified 37°C, 5% CO 2 cell incubator until experiment is completed. Low‐melting‐point agarose (Roth, cat. no. 6351.1) Phosphate‐buffered saline (PBS), without calcium or magnesium (Gibco‐Invitrogen, cat. no. 14190), room temperature and 4°C Pentobarbital sodium (WDT, Garbsen, ZI‐N: 400883.00.00, PZN: 4955173; Release: 300 mg/ml pentobarbital sodium) 70% (v/v) ethanol, 99.8% pure (AppliChem, cat. no. 64‐17‐5) Mice PCLS medium (see recipe), 37°C Clinically relevant stimuli (e.g., anti‐CD3/CD28 or anti‐PD1 antibody stimulation) 50‐ml conical tubes Glass beaker Microwave 37°C water bath 1‐ml syringes (Syringe, Omnican ® ‐F, 1 ml, B. Braun, cat. no. 9161502) Tissue slicer (Alabama R&D, TSE, model no. 6000; Fig. ) Blades Surgical instruments (including anatomical scissors, spatula, scalpel, and tweezers; see Fig. and ) Medical yarn Dissecting board with pins IV catheter (i.v. Catheter 24G, Jelco ® , Smiths Medical, cat. no. 4013, lot no. 6027133) Petri dishes Tissue coring plunger, 8 mm (TSE) Tissue coring press (TSE, model no. MD5000; optional) Cell culture plates Shaker (IKA Rocker 3D digital, model no. 100101072) CAUTION : Pentobarbital sodium is a potent anesthetic drug with strong effects on the human central nervous system. At elevated doses, pentobarbital functions as an anticonvulsant for emergent seizure control and for inducing medically induced comas. Thus, all laboratory personnel must be informed of its impact and about safety precautions, such as wearing protective eyewear, gloves, and a face mask. See the hazard sheet for further details. and ) 1 Place the low‐melting‐point agarose on ice before preparation. 2 Transfer 250 mg low‐melting‐point agarose into a 50‐ml conical tube. 3 Dissolve into 10 ml PBS to make a 2.5% (w/v) solution. 4 Place the tube with agarose solution in a glass beaker and carefully heat up in a microwave. 5 Let the agarose solution cool down to 37°C. 6 Place in a 37°C water bath during the preparation of the lung and take the agarose out of the water bath before injection. 7 Secure the tissue slicer on a stable surface. 8 Insert a new clean, sharp blade into the blade holder. 9 Adjust the blade angle according to the manufacturer's guidelines (typically a slight angle to reduce tissue tearing). 10 Fill the chamber with a bottle (500 ml) of cold PBS. 11 Ensure the chamber is securely attached to the tissue slicer. 12 Set the cutting speed (usually low to moderate). 13 Adjust slice thickness (180 to 250 µm for lung slices). 14 Prepare several small surgical instruments, as shown in Fig. and . 15 Label 50‐ml conical tubes corresponding to each mouse, fill each with 20 ml cold PBS, and place on ice (Fig. and ). 16 Cut medical yarn into pieces of 20 cm each per mouse. 17 Prepare 300 mg/ml pentobarbital sodium working solution in PBS for anesthetizing the mice. 18 Clean the table where the surgery will be performed with a soap solution and sterilize it with 70% ethanol. 19 Inject each mouse intraperitoneally (i.p.) with 200 µl of 300 mg/ml pentobarbital sodium (22.5 mg pentobarbital; see step 17), depending on the weight of the mouse (Fig. ). The 200 µl of 300 mg/ml working solution combines 75 µl of an 800 mg/ml stock solution and 125 µl PBS. The exact mass of substance (600 mg/kg body weight) injected is adapted depending on the weight of the mouse (Maxeiner et al., ). The amount of 300 mg/ml pentobarbital sodium that we apply in this phase is sufficient to perform the surgical dissection. Pinch the foot of the mouse with tweezers to check the surgical tolerance of the mouse after injection. 20 Place a mouse on its back and secure it to a dissecting board with pins with its limbs spread out. Disinfect the fur of the mouse with 70% ethanol. 21 Make an incision onto the lower abdomen of the mouse and pull the fur with the skin up above the head until the organs are revealed. Secure the skin to the dissecting board. 22 Carefully remove the glandular tissue by pulling it to the head until the trachea is seen. Free the trachea by removing any connective tissue. Dissection of the trachea should be started immediately after surgery begins to avoid changes in the airway environment (Fig. ). 23 Fix the trachea on a flat surface by carefully pushing a spatula under the trachea. Prepare the medical yarn (see step 16) by wrapping it around the trachea loosely, so it can be tightened rapidly after injection. 24 Cut open the chest by carefully holding the sternum with a pair of tweezers and cutting along the outer edge of the rib cage, trying not to damage the lung. IMPORTANT NOTE : Do not remove the rib cage entirely because the lung can be easily injured. 25 CRITICAL STEP: Carefully insert the IV catheter into the trachea. When the catheter is in place, delicately pull the needle out and slide the cannula into the trachea until resistance is sensed. Then, pull it a few millimeters back out to ensure its correct position and to avoid the backflow of the agarose gel (Fig. and ). 26 Fill a 1‐ml syringe with 1 ml of 2.5% low‐melting‐point agarose from step 6, carefully insert it into the cannula, and slowly inject the solution. Carefully observe if the agarose gel is entering the lung or not. If it flows outside, stop and arrange the equipment again (Fig. ). 27 Rapidly bind the trachea at the insertion point with the medical yarn, but not tightly (Fig. ), and place a glove filled with ice on the lungs for 10 to 15 min to allow the agar to solidify (Fig. ). 28 Cut the rib cage along the sides to allow it to be lifted or removed and then carefully dissect the lung lobes. 29 Transfer the entire set of lungs into a tube of cold PBS (see step 15). Now the lung is ready to be subjected to tissue slicing (Fig. ). 30 Place an agarose‐embedded lung lobe into a petri dish filled with cold PBS (Fig. ). 31 Adjust the settings of the tissue slicer (prepared in steps 7 to 13) for optimal cutting (e.g., 180‐ to 250‐µm section thickness by using a core unit of 0.5‐ or 0.8‐mm diameter). 32 Punch the biopsy (Fig. ), adjust the tissue coring plunger/insert set (Fig. ), and insert the tissue block (Fig. ). 33 If available, fix the tissue coring plunger into the tissue coring press and make PCLSs (see Video ). Otherwise, install the plunger/insert set into the tissue slicer chamber (Fig. and ) and start the slicing process, ensuring the blade moves smoothly through the tissue. Collect the slices in cold PBS to maintain tissue viability (Fig. and ). 34 Transfer the lung slices to cell culture plates containing PBS or PCLS medium. 35 Remove excessive agarose by incubating the PCLSs in PBS or PCLS medium three times for 30 min each on a shaker in a humidified 37°C, 5% CO 2 cell incubator. 36 After the washing process, proceed with treatment of the lung slices with the chosen clinically relevant stimuli (e.g., anti‐CD3/CD28 or anti‐PD1 antibody stimulation) in the PCLS medium and maintain the PCLS culture in the humidified 37°C, 5% CO 2 cell incubator until experiment is completed. LL/2 CELL TREATMENT WITH G‐418 SOLUTION To make sure that LL/2‐luc‐M38 cells will be expressing luciferase and thus visible with a luminometer, this protocol should be performed. Materials Murine LL/2‐luc‐M38 cell line (LL/2 cells; Bioware cell line, Caliper Life Sciences) LL/2 cell medium (see recipe) containing 500 µg/ml G‐418 solution (Sigma‐Aldrich), 37°C Cell culture flask, T‐75 (Sarstedt, cat. no. 83.3911) NOTE : The murine LL/2‐luc‐M38 (LL/2) cell line was purchased from and authenticated by Caliper Life Sciences. All Caliper Life Sciences cell lines were confirmed to be pathogen free by IMPACT profile I (PCR) at the University of Missouri Research Animal Diagnostic and Investigative Laboratory. In these cells, luciferase expression is coupled to a neomycin‐resistance gene, which renders the cell resistant to Geneticin (G‐418). 1 To see if the luciferase vector is still present intracellularly, treat LL/2 cells in a cell culture flask with LL/2 cell medium containing 500 µg/ml G‐418 solution for 3 days to select luciferase‐expressing cells. 2 Perform three cell passages between thawing and usage (see Support Protocol ). Mycoplasma contamination is checked using the Mycoplasma Detection Kit (Absource Diagnostics) according to the manufacturer's protocol. Murine LL/2‐luc‐M38 cell line (LL/2 cells; Bioware cell line, Caliper Life Sciences) LL/2 cell medium (see recipe) containing 500 µg/ml G‐418 solution (Sigma‐Aldrich), 37°C Cell culture flask, T‐75 (Sarstedt, cat. no. 83.3911) NOTE : The murine LL/2‐luc‐M38 (LL/2) cell line was purchased from and authenticated by Caliper Life Sciences. All Caliper Life Sciences cell lines were confirmed to be pathogen free by IMPACT profile I (PCR) at the University of Missouri Research Animal Diagnostic and Investigative Laboratory. In these cells, luciferase expression is coupled to a neomycin‐resistance gene, which renders the cell resistant to Geneticin (G‐418). MURINE MODEL OF LUNG ADENOCARCINOMA AND IN VIVO IMAGING After performing Support Protocol , LL/2 cell culture can be used to expand the cell number before their in vivo injection. Additional Materials (also see Support Protocol ) LL/2 cell medium (see recipe), 37°C DMEM (without supplements) 6‐ to 8‐week‐old female mice Anesthetic chamber connected to isoflurane vaporizer Luciferin (Promega) Scale 1‐ml syringes (Omnican‐F, 0.30 × 12 mm/ G30×1/2, Braun, cat. no. 9161502S) IVIS Spectrum In Vivo Imaging System (PerkinElmer) NOTE : All mice are housed individually ventilated cages (IVCs) equipped with micro‐isolator lids. 1 Culture LL/2 cells in LL/2 cell medium for 1 week before injection day. 2 Resuspend 500,000 LL/2 cells in 200 µl DMEM (without supplements) into the tail vein of 6‐ to 8‐week‐old female mice. 3 To track tumor development, weigh the mice every 2 days until the end of the experiment. 4 One day before the experiment ends, weigh the mice and measure luciferase activity using the IVIS Spectrum In Vivo Imaging System. 5 Anesthetize mice using an anesthetic chamber connected to an isoflurane vaporizer. 6 Inject the mice i.p. with luciferin (0.10 mg/g body weight) using a 1‐ml syringe. 7 After 20 min, measure luciferase activity with the IVIS Spectrum In Vivo Imaging System by detecting luminescence intensity (photons per second). 8 Perform lung tumor load analysis in logarithmic scale mode and determine the total flux (photons per second) as described previously (Heim et al., ). ) LL/2 cell medium (see recipe), 37°C DMEM (without supplements) 6‐ to 8‐week‐old female mice Anesthetic chamber connected to isoflurane vaporizer Luciferin (Promega) Scale 1‐ml syringes (Omnican‐F, 0.30 × 12 mm/ G30×1/2, Braun, cat. no. 9161502S) IVIS Spectrum In Vivo Imaging System (PerkinElmer) NOTE : All mice are housed individually ventilated cages (IVCs) equipped with micro‐isolator lids. H&E STAINING OF PCLS SECTIONS At the end of the experiment described in Support Protocol followed by the Basic Protocol, the PCLSs generated can be prepared for histology. Materials PCLSs (see Basic Protocol) 10% (w/v) formalin in PBS 80%, 90%, 95%, and 100% (v/v) ethanol Xylene Paraffin Hematoxylin and eosin (H&E) Non‐aqueous mounting medium Histology cassettes Glass container Embedding unit (Leica) Microtome (Leica) Glass slides and coverslips 1 Fix the PCLSs by placing in histology cassettes in a glass container filled with 10% formalin in PBS. 2 Dehydrate using a gradient series of ethanol (80%, 90%, 95%, and 100% ethanol). 3 Incubate with 100% xylene. 4 Place the slices in an embedding unit and fill with paraffin. 5 Wait for the embedding unit to cool down and then take the paraffin block out. 6 Cut 4‐µm‐thick slices from the PCLS paraffin blocks on a microtome. 7 Stain slices with H&E for visualization of lung tumor, inflammation, and histology (Fig. ). 8 Seal the sections on a glass slide with a glass coverslip by using a drop of non‐aqueous mounting medium. PCLSs (see Basic Protocol) 10% (w/v) formalin in PBS 80%, 90%, 95%, and 100% (v/v) ethanol Xylene Paraffin Hematoxylin and eosin (H&E) Non‐aqueous mounting medium Histology cassettes Glass container Embedding unit (Leica) Microtome (Leica) Glass slides and coverslips MEASURING CYTOKINES BY ELISA This protocol detects cytokines and other proteins or peptides present in PCLS culture supernatants by binding to a specific antibody. There are different categories of mediators that can be measured in the PCLS supernatants depending on the focus of the research group, and the chosen pretreatment of the PCLSs and storage conditions must be taken into consideration. At the end of PCLS culture, supernatants are collected in properly labeled sterile Eppendorf tubes and stored frozen until ELISAs are performed in 96‐well plates, which permits high‐throughput results. On the left‐hand side of the plate, the standard curve for the cytokine to be detected is set up in duplicate. Briefly, a series of cytokine dilutions is prepared, including a negative control without cytokine, as indicated by the manufacturer. The standard curve wells are treated the same as the other wells except for the addition of the cytokine instead of the samples. For further protocol details, manufacturers' instructions and specific protocols (see Current Protocols article: Hornbeck, ) should be analyzed. Materials Capture antibody diluted in reagent diluent [0.1% (w/v) bovine serum albumin (BSA) and 0.05% (w/v) Tween in Tris‐buffered saline] Wash buffer: PBS/0.05% (w/v) Tween Blocking solution [e.g., 1% (w/v) BSA in PBS with 0.05% (w/v) NaN 3 ] Standard specific for cytokine to be detected PCLS supernatants Secondary biotin‐conjugated antibody diluted in reagent diluent Streptavidin conjugated to horseradish peroxidase diluted in reagent diluent Substrate Stop solution (2 N H 2 SO 4 ) 96‐well plate Adhesive foil (Berthold Technologies, cat. no. 55590) Plate reader NOTE : To prepare the PCLS supernatants, culture PCLSs (see Basic Protocol) for 48 hr in PCLS medium (see recipe). 1 Coat the bottom of each well of a 96‐well plate with a capture antibody (diluted in reagent diluent) that will specifically bind the anticipated antigen or standard (cytokine) present in the PCLS supernatants, seal the plate with adhesive foil, and incubate overnight at 4°C. Following incubation, perform three washes with wash buffer (PBS/0.05% Tween). 2 To prevent nonspecific binding, incubate the wells with blocking solution for 1 hr at room temperature, followed by three washes with PBS/0.05% Tween. 3 Add the standard specific for the cytokine to be detected and the PCLS supernatants obtained from the PCLS cultures. 4 To detect the bound antigens (cytokines), add a secondary biotin‐conjugated antibody diluted in reagent diluent to each well (detection antibody) at the concentration indicated by the manufacturer. After an incubation of 1 hr at room temperature, discard the secondary antibody solution and wash three times with PBS/0.05% Tween. Then, add streptavidin conjugated to horseradish ‐peroxidase diluted in reagent diluent to each well and cover the plate with foil. Incubate 20 min at room temperature in the dark. At the end of this incubation, perform three washes with PBS/0.05% Tween. Enzymes like peroxidase or alkaline phosphatase can metabolize colorless substrates (chromogens) into colored products. 5 Add the substrate to be used by the enzyme linked to the secondary antibody. For example, 3,3′,5,5′‐tetramethylbenzidin (TMB) can be used for peroxidase, resulting in the appearance of a colored product in the presence of specific secondary antibody binding. 6 When the enzyme reaction is completed, add stop‐solution (2 N H 2 SO 4 ), place the entire plate into a plate reader, and determine the optical density (i.e., the amount of colored product) for each well at 450 nm. The amount of color produced is proportional to the amount of proteins (cytokines) bound to the capture antibody on the bottom of the wells. Generally, the measurements of cytokines in PCLSs are achieved by ELISA. In our laboratory, we use ELISA kits from two different manufacturers, BD Bioscience and R&D Systems. The assays are performed in accordance with the manufacturers’ protocols. Another assay for cytokine release is the cytometric bead array (CBA) assay, as we previously described (Doganci et al., ). Capture antibody diluted in reagent diluent [0.1% (w/v) bovine serum albumin (BSA) and 0.05% (w/v) Tween in Tris‐buffered saline] Wash buffer: PBS/0.05% (w/v) Tween Blocking solution [e.g., 1% (w/v) BSA in PBS with 0.05% (w/v) NaN 3 ] Standard specific for cytokine to be detected PCLS supernatants Secondary biotin‐conjugated antibody diluted in reagent diluent Streptavidin conjugated to horseradish peroxidase diluted in reagent diluent Substrate Stop solution (2 N H 2 SO 4 ) 96‐well plate Adhesive foil (Berthold Technologies, cat. no. 55590) Plate reader NOTE : To prepare the PCLS supernatants, culture PCLSs (see Basic Protocol) for 48 hr in PCLS medium (see recipe). MEASURING CYTOTOXIC ACTIVITY IN PCLS CONDITIONED MEDIUM This protocol is used to analyze the cytotoxic activity of T cells or other cells against the murine tumor cell line LL/2, used to induce tumors in vivo . Materials Murine LL/2‐luc‐M38 cell line (LL/2 cells; Bioware cell line, Caliper Life Sciences) LL/2 cell medium (see recipe), 37°C PBS, without calcium or magnesium (Gibco‐Invitrogen, cat. no. 14190) 30% to 50% (v/v) PCLS conditioned medium (PCLS‐CM), 37°C 30% to 50% (v/v) CTLL2 conditioned medium (CTLL2‐CM), 37°C Luciferin (Promega) 96‐well white‐walled plates, sterile (Berthold Technologies, cat. no. 24910) Adhesive foil (Berthold Technologies, cat. no. 55590) Centro XS³ LB 960 Microplate Luminometer (Berthold Technologies) NOTE : To prepare the PCLS‐CM, first culture PCLSs (see Basic Protocol) for 48 hr in PCLS medium (see recipe). CM is prepared by combining 30% to 50% PCLS supernatant with 70% to 50% cell culture medium. The CM percentage is selected based on the cytotoxicity anticipated. NOTE : To prepare the CTLL2‐CM, the murine CTLL2 cell line was kindly provided by Dr. Ulrike Schleicher of the Institute of Microbiology, University Hospital Erlangen (Erlangen, Germany). CTTL2 cells are cultured in RPMI1640 medium (Gibco, Thermo Fisher Scientific), supplemented with 10% (v/v) fetal bovine serum (FBS), 1% (v/v) penicillin/streptomycin, 1% (v/v) l ‐glutamine, and 4 ng/ml IL2 (PeproTech). 1 Seed LL/2 cells in LL/2 cell medium at a density of ∼7 × 10³ cells per well in 96‐well white‐walled plates. In addition, set up a standard curve in quadruplicates with eight increasing concentrations of LL/2 cells on the left side of the plate. Culture for 24 hr in a humidified 37°C, 5% CO 2 cell incubator. 2 After 24 hr, discard supernatants, wash cells in PBS, and incubate cells with fresh LL/2 cell medium, 30% to 50% PCLS conditioned medium (PCLS‐CM), or 30% to 50% CTLL2‐CM as a positive control for 24 hr in a humidified 37°C, 5% CO 2 cell incubator. CM is prepared by adding the 30‐50 % supernatants of the PCLS to 70‐50% cell culture medium, respectively. The CM % is selected based on the cytotoxicity anticipated. 3 After 24‐hr incubation, remove the medium, treat the LL/2 cells with Luciferin/PBS 150 µg/ml for 25 min, 37°C, 5% CO 2 , and detect the fluorescence intensity using the Centro XS³ LB 960 MicroplateLuminometer. Before analysis stick the adhesive foil to the bottom of the plate. 4 Calculate respective cell numbers by using the values from the standard curve (Table ). For calculation of the results, a log‐log regression analysis is used. The linear regression (R‐squared) is drawn, and the coefficient should be >0.99 for optimal results. Murine LL/2‐luc‐M38 cell line (LL/2 cells; Bioware cell line, Caliper Life Sciences) LL/2 cell medium (see recipe), 37°C PBS, without calcium or magnesium (Gibco‐Invitrogen, cat. no. 14190) 30% to 50% (v/v) PCLS conditioned medium (PCLS‐CM), 37°C 30% to 50% (v/v) CTLL2 conditioned medium (CTLL2‐CM), 37°C Luciferin (Promega) 96‐well white‐walled plates, sterile (Berthold Technologies, cat. no. 24910) Adhesive foil (Berthold Technologies, cat. no. 55590) Centro XS³ LB 960 Microplate Luminometer (Berthold Technologies) NOTE : To prepare the PCLS‐CM, first culture PCLSs (see Basic Protocol) for 48 hr in PCLS medium (see recipe). CM is prepared by combining 30% to 50% PCLS supernatant with 70% to 50% cell culture medium. The CM percentage is selected based on the cytotoxicity anticipated. NOTE : To prepare the CTLL2‐CM, the murine CTLL2 cell line was kindly provided by Dr. Ulrike Schleicher of the Institute of Microbiology, University Hospital Erlangen (Erlangen, Germany). CTTL2 cells are cultured in RPMI1640 medium (Gibco, Thermo Fisher Scientific), supplemented with 10% (v/v) fetal bovine serum (FBS), 1% (v/v) penicillin/streptomycin, 1% (v/v) l ‐glutamine, and 4 ng/ml IL2 (PeproTech). ANALYSIS OF RNA FROM PCLSs BY REAL‐TIME PCR To analyze total gene trascription in the PCLSs after cell culture, total RNA is extracted. Gene transcription in response to given stimuli can be compared to untreated PCLS gene transcription by using qPCR or gene or RNA sequencing. Materials PCLSs [see Basic Protocol; stored at –80°C in 300 µl RNAprotect (Qiagen, lot no. 160027782) until RNA is extracted] QIAzol (Qiagen) Chloroform Glycogen Isopropanol 70% (v/v) ethanol RNase‐free water, sterile RevertAid First Strand cDNA Synthesis Kit (Fermentas) iTaq Universal SYBR Green Supermix (Bio‐Rad) Quantitative real‐time PCR (qPCR) primers (Eurofins MWG Operon) 1.5‐ and 2‐ml Eppendorf tubes Small metallic beads, sterilized (Innuscreen, cat. no. 845‐CS‐11400050) Parafilm Tissue homogenizer (Minilys, Bertin Technologies, or SpeedMill Plus, Analytik Jena) Vortex Microcentrifuge, 4°C Spectrophotometer CFX‐96 Real‐Time PCR Detection System (Bio‐Rad) CFX Manager Software (Bio‐Rad) RNA extraction from PCLSs 1 Transfer the PCLSs from the RNAprotect solution to a new 2‐ml Eppendorf tube containing 500 µl QIAzol and two sterilized small metallic beads. Wrap the cap of the Eppendorf tube with Parafilm to ensure that the tube does not open during the homogenization step. 2 Carefully place the Eppendorf tube in the tissue homogenizer SpeedMill Plus and homogenize for 1 min. 3 Transfer the homogenized solution into a new Eppendorf tube. 4 Add 500 µl QIAzol to the homogenized solution. 5 To perform the segregation phase: Incubate the samples for 5 min at room temperature. Add chloroform at 100 µl/ml. Vortex the samples for 20 s. Incubate for an additional 3 min at room temperature. Centrifuge 15 min at 12,000 × g , 4°C. 6 To perform RNA precipitation: Transfer the upper aqueous phase containing the RNA into a new 1.5‐ml Eppendorf tube. Add 3 µl glycogen and 350 µl isopropanol. Vortex for 15 s. Incubate for 15 min at room temperature. Centrifuge 10 min at 12,000 × g , 4°C. 7 To perform RNA wash: Carefully discard the supernatant. Add 1 ml of 70% ethanol to the RNA pellet. Centrifuge 5 min at 12,000 × g , 4°C. Discard the ethanol. 8 To perform RNA drying and resuspension: Leave the Eppendorf tube open to allow the RNA pellet to air‐dry. Add 21.5 µl sterile RNase‐free water to dissolve the RNA. Measure RNA concentration and purity with a spectrophotometer. Store the RNA at –80°C until further analysis. The RNA obtained is reported in Table . Quantitative real‐time PCR 9 Reverse‐transcribe RNA into cDNA using the RevertAid First Strand cDNA Synthesis Kit according to the manufacturer's instructions. 10 Perform qPCR with synthesized cDNA using iTaq Universal SYBR Green Supermix and appropriate primers in a total volume of 20 µl under the following thermal cycling conditions using the CFX‐96 Real‐Time PCR Detection System: 11 Perform relative quantification using the 2−ΔΔ C t method using CFX Manager Software (Bio‐Rad). Hypoxanthine‐guanine‐phosphoribosyltransferase (HPRT) can be used as the internal standard. PCLSs [see Basic Protocol; stored at –80°C in 300 µl RNAprotect (Qiagen, lot no. 160027782) until RNA is extracted] QIAzol (Qiagen) Chloroform Glycogen Isopropanol 70% (v/v) ethanol RNase‐free water, sterile RevertAid First Strand cDNA Synthesis Kit (Fermentas) iTaq Universal SYBR Green Supermix (Bio‐Rad) Quantitative real‐time PCR (qPCR) primers (Eurofins MWG Operon) 1.5‐ and 2‐ml Eppendorf tubes Small metallic beads, sterilized (Innuscreen, cat. no. 845‐CS‐11400050) Parafilm Tissue homogenizer (Minilys, Bertin Technologies, or SpeedMill Plus, Analytik Jena) Vortex Microcentrifuge, 4°C Spectrophotometer CFX‐96 Real‐Time PCR Detection System (Bio‐Rad) CFX Manager Software (Bio‐Rad) 1 Transfer the PCLSs from the RNAprotect solution to a new 2‐ml Eppendorf tube containing 500 µl QIAzol and two sterilized small metallic beads. Wrap the cap of the Eppendorf tube with Parafilm to ensure that the tube does not open during the homogenization step. 2 Carefully place the Eppendorf tube in the tissue homogenizer SpeedMill Plus and homogenize for 1 min. 3 Transfer the homogenized solution into a new Eppendorf tube. 4 Add 500 µl QIAzol to the homogenized solution. 5 To perform the segregation phase: Incubate the samples for 5 min at room temperature. Add chloroform at 100 µl/ml. Vortex the samples for 20 s. Incubate for an additional 3 min at room temperature. Centrifuge 15 min at 12,000 × g , 4°C. 6 To perform RNA precipitation: Transfer the upper aqueous phase containing the RNA into a new 1.5‐ml Eppendorf tube. Add 3 µl glycogen and 350 µl isopropanol. Vortex for 15 s. Incubate for 15 min at room temperature. Centrifuge 10 min at 12,000 × g , 4°C. 7 To perform RNA wash: Carefully discard the supernatant. Add 1 ml of 70% ethanol to the RNA pellet. Centrifuge 5 min at 12,000 × g , 4°C. Discard the ethanol. 8 To perform RNA drying and resuspension: Leave the Eppendorf tube open to allow the RNA pellet to air‐dry. Add 21.5 µl sterile RNase‐free water to dissolve the RNA. Measure RNA concentration and purity with a spectrophotometer. Store the RNA at –80°C until further analysis. The RNA obtained is reported in Table . 9 Reverse‐transcribe RNA into cDNA using the RevertAid First Strand cDNA Synthesis Kit according to the manufacturer's instructions. 10 Perform qPCR with synthesized cDNA using iTaq Universal SYBR Green Supermix and appropriate primers in a total volume of 20 µl under the following thermal cycling conditions using the CFX‐96 Real‐Time PCR Detection System: 11 Perform relative quantification using the 2−ΔΔ C t method using CFX Manager Software (Bio‐Rad). Hypoxanthine‐guanine‐phosphoribosyltransferase (HPRT) can be used as the internal standard. FLOW CYTOMETRY ANALYSIS OF CELLS ISOLATED FROM PCLSs At the end of the Basic Protocol, the PCLSs can be dissociated into a single‐cell suspension, and the single cells can be phenotyped by FACS by using fluorochrome‐labeled antibodies. Materials PCLSs (see Basic Protocol) Cell digestion solution (see recipe; make fresh) PBS (Anprotec) Zombie Aqua Fixable Viability Kit (BioLegend) FACS buffer (see recipe) FACS antibodies 15‐ml tubes 37°C incubator Shaker 40‐µm cell strainers Microcentrifuge, 4°C Flow cytometer 1 Place the PCLSs in 15‐ml tubes and digest the PCLSs enzymatically with fresh cell digestion solution at 37°C for 45 min under shaking (300 rpm) in an incubator. 2 Push the mixture through a 40‐µm cell strainer and into a fresh 15‐ml tube to obtain a single‐cell suspension. 3 Centrifuge the cells for 5 min at 1500 × g , 4°C, and resuspend in PBS. 4 Incubate the cells for 15 min with fixable dye from the Zombie Aqua Fixable Viability Kit to identify the living cells. 5 Stop the staining with 200 µl FACS buffer. 6 Centrifuge the cells again (see step 3) and stain with specific FACS antibodies for T‐cell analysis (Table ). 7 Incubate for 25 min in the dark at room temperature for surface staining. 8 Stop the staining with 200 µl FACS buffer. 9 Centrifuge the cells (see step 3). 10 Resuspend in 200 µl FACS buffer and analyze using a flow cytometer. PCLSs (see Basic Protocol) Cell digestion solution (see recipe; make fresh) PBS (Anprotec) Zombie Aqua Fixable Viability Kit (BioLegend) FACS buffer (see recipe) FACS antibodies 15‐ml tubes 37°C incubator Shaker 40‐µm cell strainers Microcentrifuge, 4°C Flow cytometer Cell digestion solution PBS (Anprotec) 10 mg/ml DNAse I (Roche Diagnostics) 300 U/ml collagenase type IA (Sigma‐Aldrich) Prepare fresh immediately before use and store at 4°C until needed FACS buffer PBS (Anprotec) 2% (v/v) FBS (Sigma‐Aldrich) 1.4 mM EDTA (from 1% Na2‐EDTA solution, Bio&SELL) Store ≤1 week at 4°C LL/2 cell medium DMEM 10% (v/v) FBS 1% (v/v) penicillin/streptomycin 1% (v/v) l ‐glutamine Store ≤1 week at 4°C PCLS medium 50 ml high‐glucose DMEM 0.5 ml 100× MEM non‐essential amino acid mixture (1× final) 0.5 ml 100× penicillin/streptomycin (1000 U/ml final) 0.5 ml 100 mM sodium pyruvate (1 mM final) 0.225 ml 10 mg/ml gentamycin (45 µg/ml final) 300 µl 250 µg/ml amphotericin B (1.5 µg/ml final) Store ≤1 week at 4°C PBS (Anprotec) 10 mg/ml DNAse I (Roche Diagnostics) 300 U/ml collagenase type IA (Sigma‐Aldrich) Prepare fresh immediately before use and store at 4°C until needed PBS (Anprotec) 2% (v/v) FBS (Sigma‐Aldrich) 1.4 mM EDTA (from 1% Na2‐EDTA solution, Bio&SELL) Store ≤1 week at 4°C DMEM 10% (v/v) FBS 1% (v/v) penicillin/streptomycin 1% (v/v) l ‐glutamine Store ≤1 week at 4°C 50 ml high‐glucose DMEM 0.5 ml 100× MEM non‐essential amino acid mixture (1× final) 0.5 ml 100× penicillin/streptomycin (1000 U/ml final) 0.5 ml 100 mM sodium pyruvate (1 mM final) 0.225 ml 10 mg/ml gentamycin (45 µg/ml final) 300 µl 250 µg/ml amphotericin B (1.5 µg/ml final) Store ≤1 week at 4°C Critical Parameters This set of protocols requires pre‐training in and standardization of technical procedures that are quite difficult for the beginner. A critical step in the Basic Protocol (step 25) is the implantation of the cannula into the murine trachea. Once proficient in performing the surgical procedure and injecting the low‐melting‐point agarose solution described here, you will obtain accurate information about the current state of the immune system in the airways of your mouse model. Troubleshooting Correctly performing the tracheostomy determines the success of the Basic Protocol. Therefore, one should strictly avoid cutting more than one third of the trachea's perimeter to avoid a complete disruption of the trachea. Please see Table for further advice. Understanding Results This article provides a reproducible method for observing the dynamics of immunologic processes in the lung during disease. The entire analysis of the cell phenotype and structure, as well as the released cytokines and mediators, only from one mouse is achieved using this technique (Basic Protocol and Support Protocols to ). This approach is completed in 2 days. After cell culture, we successfully obtained histological sections using standard histopathological protocols and stained them with H&E (Support Protocol and Fig. ). Remarkably, approximately 30 PCLSs can be obtained from a single murine lung, and approximately 40 histologic slides (each 4‐µm thick) per PCLS can be mounted. Additionally, the molecular processes in the cells can be examined by RNA isolation executed by RNA sequencing analysis and/or reverse transcription followed by qPCR for targeted gene expression (Support Protocol , Fig. , and Table ). The identification of cell types with increased frequency during disease can be carried out through flow cytometric analysis (Support Protocol and Fig. ), allowing the characterization of relevant immune cell populations and their phenotypes. Therapeutic success can also be assessed by studying apoptosis or intracellular mechanisms. However, the limitation of using FACS as an analysis method is the small yield of cells that can be secured after cell culture and digestion. Still, CD4+ and CD8+ T‐cell populations in samples from a wild‐type mouse can be retrieved using this method. Time Considerations The time required varies depending on the specific mouse model used in the laboratory. The disease induction and exposure to allergens determine the duration of the experiment. The following are the time requirements for the Basic Protocol: Low‐melting‐point agarose gel preparation – 5 min: prepare a 2.5% low‐melting‐point agarose solution for lung infusion. Tissue slicer setup – 10 min: ensure that the tissue slicer is calibrated and ready for use. Preparation of other equipment and reagents – 5 min: prepare all necessary reagents (e.g., culture media, buffers, staining solutions). Induction of anesthesia – 5 min: administer the anesthetic drug to the mouse via i.p. injection. Dissection – 10 min: dissect the trachea of the mouse. Insertion of IV catheter and injection of agarose – 10 min: secure the IV catheter for lung infusion of agarose. Surgical removal of the lungs – 20 min: remove the lungs of the mouse. Tissue slicing – 30 min: prepare lung slices using a tissue slicer. Handling of lung slices – 1 to 2 hr: after slicing, handle and culture the PCLSs, ensuring proper incubation and washing steps. This phase includes initial handling and setup for downstream analyses (e.g., cytokine analysis, histology, or flow cytometry). The following are the time requirements for the support protocols: Support Protocol – 3 days. Support Protocol – 14 to 21 days after intravenous LL/2 cell injection. Support Protocol – ≥3 days. Support Protocol – 2 days. Support Protocol – 3 days. Support Protocol – 2 days. Support Protocol – 1 day. Author Contributions M.T. and C.C. contributed to the establishment of the methods1, RNA extraction, composition of the three figures, and the writing of the manuscript. In addition, M.T. generated the video and contributed to the revision of the manuscript. J.W. contributed to the establishment of the protocols, edited the manuscript and figure legends, and contributed to the revision of the manuscript. E.N. performed the histology and contributed to the PCLS method establishment. S.K. contributed to the flow cytometry protocol and provided the FACS data related to this protocol. M.T.C. injected the LL/2 cells into the mice to generate the murine lung adenocarcinoma model. S.Z. contributed to the RNA part. C.I.G. supervised the pathology and histology analysis. S.F. conceived and supervised the project and contributed to the revision of the manuscript. Conflict of Interest The authors declare no conflict of interest. This set of protocols requires pre‐training in and standardization of technical procedures that are quite difficult for the beginner. A critical step in the Basic Protocol (step 25) is the implantation of the cannula into the murine trachea. Once proficient in performing the surgical procedure and injecting the low‐melting‐point agarose solution described here, you will obtain accurate information about the current state of the immune system in the airways of your mouse model. Correctly performing the tracheostomy determines the success of the Basic Protocol. Therefore, one should strictly avoid cutting more than one third of the trachea's perimeter to avoid a complete disruption of the trachea. Please see Table for further advice. This article provides a reproducible method for observing the dynamics of immunologic processes in the lung during disease. The entire analysis of the cell phenotype and structure, as well as the released cytokines and mediators, only from one mouse is achieved using this technique (Basic Protocol and Support Protocols to ). This approach is completed in 2 days. After cell culture, we successfully obtained histological sections using standard histopathological protocols and stained them with H&E (Support Protocol and Fig. ). Remarkably, approximately 30 PCLSs can be obtained from a single murine lung, and approximately 40 histologic slides (each 4‐µm thick) per PCLS can be mounted. Additionally, the molecular processes in the cells can be examined by RNA isolation executed by RNA sequencing analysis and/or reverse transcription followed by qPCR for targeted gene expression (Support Protocol , Fig. , and Table ). The identification of cell types with increased frequency during disease can be carried out through flow cytometric analysis (Support Protocol and Fig. ), allowing the characterization of relevant immune cell populations and their phenotypes. Therapeutic success can also be assessed by studying apoptosis or intracellular mechanisms. However, the limitation of using FACS as an analysis method is the small yield of cells that can be secured after cell culture and digestion. Still, CD4+ and CD8+ T‐cell populations in samples from a wild‐type mouse can be retrieved using this method. The time required varies depending on the specific mouse model used in the laboratory. The disease induction and exposure to allergens determine the duration of the experiment. The following are the time requirements for the Basic Protocol: Low‐melting‐point agarose gel preparation – 5 min: prepare a 2.5% low‐melting‐point agarose solution for lung infusion. Tissue slicer setup – 10 min: ensure that the tissue slicer is calibrated and ready for use. Preparation of other equipment and reagents – 5 min: prepare all necessary reagents (e.g., culture media, buffers, staining solutions). Induction of anesthesia – 5 min: administer the anesthetic drug to the mouse via i.p. injection. Dissection – 10 min: dissect the trachea of the mouse. Insertion of IV catheter and injection of agarose – 10 min: secure the IV catheter for lung infusion of agarose. Surgical removal of the lungs – 20 min: remove the lungs of the mouse. Tissue slicing – 30 min: prepare lung slices using a tissue slicer. Handling of lung slices – 1 to 2 hr: after slicing, handle and culture the PCLSs, ensuring proper incubation and washing steps. This phase includes initial handling and setup for downstream analyses (e.g., cytokine analysis, histology, or flow cytometry). The following are the time requirements for the support protocols: Support Protocol – 3 days. Support Protocol – 14 to 21 days after intravenous LL/2 cell injection. Support Protocol – ≥3 days. Support Protocol – 2 days. Support Protocol – 3 days. Support Protocol – 2 days. Support Protocol – 1 day. M.T. and C.C. contributed to the establishment of the methods1, RNA extraction, composition of the three figures, and the writing of the manuscript. In addition, M.T. generated the video and contributed to the revision of the manuscript. J.W. contributed to the establishment of the protocols, edited the manuscript and figure legends, and contributed to the revision of the manuscript. E.N. performed the histology and contributed to the PCLS method establishment. S.K. contributed to the flow cytometry protocol and provided the FACS data related to this protocol. M.T.C. injected the LL/2 cells into the mice to generate the murine lung adenocarcinoma model. S.Z. contributed to the RNA part. C.I.G. supervised the pathology and histology analysis. S.F. conceived and supervised the project and contributed to the revision of the manuscript. The authors declare no conflict of interest. |
Effect of exogenous treatment with zaxinone and its mimics on rice root microbiota across different growth stages | a6aaa982-3a64-41d4-a81f-4bfb8933f3c8 | 11682185 | Microbiology[mh] | Ensuring food security for the world population is challenged by a variety of factors, including climate change, environmental pollution, and, in particular, the rapidly escalating demand to address the expanding human population . According to the United Nations Food and Agriculture Organization (FAO) estimates, agriculture will need to increase food production by nearly 70% by 2050. There is a projected need for a 112% increase in food production to meet anticipated caloric requirements in specific regions such as South Asia and sub-Saharan Africa . Rice ( Oryza sativa L.), a member of the Poaceae family, is a global major food crop, supplying staple sustenance to nearly half of the world’s population . Approximately 60% of the world’s rice is cultivated in Southeast Asia, and its production and consumption are on the rise in Africa as well . Currently, the productivity of rice is threatened by pests, soil degradation , , diminishing water , and environmental pollution . Weeds (37.02%), insects (27.9%), and fungal pests (15.6%) are recognized as primary contributors to yield losses . In sub-Saharan Africa, parasitic weeds of the genus Striga cause significant yield reductions , a situation expected to worsen with the influence of climate change . Given that modeling projections for rice production indicate an emerging constraint on yields , global adaptation and mitigation strategies are imperative. A promising solution is the use of growth-promoting biostimulants that include molecules and/or microorganisms enhancing plant fitness in terms of plant growth, productivity, and nutrient utilization efficiency. Additionally, biostimulants may have the capacity to bolster tolerance against a broad spectrum of abiotic and biotic stresses – . The employment of plant-associated microorganisms (plant microbiota) represents a particularly promising, long-term solution to the challenges of attaining food security and preserving the environment . The plant microbiota is shaped by eco-evolutionary processes driven by the metabolic affinity between partners and chemical signals released by plant roots into the rhizosphere to screen the microbial community , . Besides beneficial microbes, the application of phytohormones or hormone-like compounds, such as auxins and cytokinins or sterols and polyamines, respectively, showed positive effects on plants, including the promotion of plant growth and productivity . The plant pigments carotenoids are a source of a series of regulatory metabolites. Oxidative cleavage of these carotenoids leads to a class of compounds called apocarotenoids that encompass precursors for the phytohormones abscisic acid (ABA) and strigolactones (SLs), as well as bioactive metabolites and growth regulators, such as β-cyclocitral, anchorene, and zaxinone . Indeed, apocarotenoids play a role in nearly all aspects of plant physiology and development, contribute to the plant response to both abiotic and biotic stresses, and mediate plant-plant and plant-microbe interactions . The carotenoid-derived hormone SLs regulates various aspects of plant development, including shoot branching, root architecture, and leaf senescence, and modulates plant responses to both abiotic and biotic stress – . Additionally, SLs play a pivotal role in rhizospheric communications, manifesting both negative and positive effects , . On one hand, they induce the germination of seeds of root parasitic plants, which is followed by infestation that causes substantial yield losses in numerous crops , . On the other hand, they act as chemical signals attracting arbuscular mycorrhizal (AM) fungi and facilitating the establishment of beneficial AM symbiosis . Recent studies have indicated that SLs also influence the composition of the rhizosphere microbial community , and regulate plant-pathogen interactions , . Zaxinone is emerging as a crucial regulator of rice growth, metabolism, hormone homeostasis, and AM symbiosis , . Its plant growth-promoting effect is mediated by an enhancement of root sugar uptake and metabolism, and a modulation of SL and cytokinin content , . The limited availability of zaxinone, due to a complex labor-intensive organic synthesis, has been overcome by the development of easily synthesizable and highly efficient zaxinone mimics (MiZax) . MiZax3 and MiZax5 exhibit zaxinone-like activities, such as rescuing root growth in zaxinone-deficient rice mutants, promoting overall growth, and reducing SL content in wild-type plants. Exogenous applications of zaxinone, MiZax3, and MiZax5 demonstrated their utility and growth-promoting effects on rice and various horticultural crops under both normal and desert conditions , . Additionally, these compounds have the ability to alleviate the infestation by the root parasitic plant Striga by decreasing SL biosynthesis , , . Interestingly, MiZax compounds were at least as efficient as zaxinone in reducing Striga infestation and had no negative impact on mycorrhization . These data highlight that zaxinone, and in particular, the highly efficient MiZax are excellent biostimulants and helpful tools for establishing sustainable agriculture and alleviating the infestation by parasitic plants. However, their impact on soil microbial community composition is still unknown. With the aim to promote the use of these novel growth-promoting compounds as biostimulants, we investigated whether the exogenous application of zaxinone or MiZax(s) on the soil could influence soil microbiota communities, the recruitment of rice root-associated microbes, and shoot and grain metabolism. Our results show that treatment with zaxinone and MiZax mostly impacted the prokaryotic component of the root endosphere. However, network analysis highlighted a partial perturbation of taxa-taxa interactions at the vegetative stage (tillering), followed by a full recovery of a complex network, structured by relevant beneficial microbial hubs, during the fruit set (milky stage). Using microbial ecology tools, we provide here new insights into the role of zaxinone and MiZax in the interplay between plants and rice root-associated microbiota. Plant growth conditions, hormonal treatments, and sampling The impact of zaxinone, MiZax3, and MiZax5 on native paddy soil and rice rhizomicrobiota was studied in a greenhouse mesocosm experiment. Four soil treatments were considered, namely zaxinone, MiZax3, MiZax5, and the solvent acetone as control (ACE), for both rice plants and unplanted soil. Plants were sampled at three phenological stages, the tillering stage (60 days after transplanting, T1) the milky-stage maturation (120 days after transplanting, T2), and the over-ripe stage (180 days after transplanting, T3). For each treatment a total of 33 plants were grown: 15 for phenotyping sampled at T3 and 18 for microbiota profiling (sampled at T1 and T2, 9 biological replicates each). For the unplanted soil treatment, 18 replicates were collected (9 for each sampling point). Mesocosm systems were set up using native paddy soil harvested from the experimental fields of the CREA-CI research center (Vercelli, Italy) which was used in previous studies , . Soil physico-chemical parameters measured on a representative batch used for this study are reported in Table . Rice seeds ( Oryza sativa cv. ‘Nipponbare’) were sown in alveolar trays filled with soil. After 1 month of growth under controlled conditions, plants were transferred to the final plastic pots (10 × 9 × 17 cm) filled with the same soil. For unplanted soil treatment, one alveolar tray was left unsown and the resulting soil cores were transferred into pots with fresh soils (as for rice seedlings) for the unplanted soil conditions. Plants and unplanted soils were grown in the greenhouse at the Department of Life Sciences and Systems Biology of the University of Torino from June 2021 to October 2021 (rice growing season) without monitoring light, temperature and humidity. Plants were watered once a week with tap water and once with distilled water containing zaxinone, MiZax3, and MiZax5 molecules dissolved individually to reach the final concentration of 5µM (10 −6 ), which has already been shown to promote growth activity in rice plant and to alleviate infestation by the root parasitic plant Striga , . Fifty mL of the solution was poured into the soil of each pot once every two weeks for about 5 months (10 treatments) to cover the vegetative and rice reproductive growth stages. Compartment isolation and microbiota profiling Plants were sampled for microbiota profiling at the tillering stage (T1) and at the milky-stage maturation (T2). At sampling, plants were removed from pots, vigorously shaken, and 10–15 g of roots collected within 3–4 cm from the base of the stem into a 50-mL Falcon tube. Unplanted soil samples were collected using a sterile spoon from the middle core of the pot, discarding edges that were in contact with plastic or any other portions where plant roots were visible. Samples were stored at + 4 °C and processed within 24 h to separate plant root compartments under sterile conditions. Samples were then processed to isolate the rhizosphere from the root endosphere according to the protocol by Bulgarelli et al. with minor modifications. Roots fragments were first washed into 10 mM sterile phosphate-buffer saline with 0.02% Tween-80 added (PBS-T), under continuous stirring on a horizontal shaker (15 min, 70 rpm). To obtain the rhizosphere soil slurry, roots were removed and tubes were centrifuged (4000 g, 10 min). The rhizosphere was then resuspended in 2 mL PBS-T and snap-freezed in liquid nitrogen. Roots were enriched in the endospheric compartment by two washes of 10 sonication cycles (30 s pulses, 30 s rest each) in PBS-T, discarding and replacing the buffer after each wash. Samples were finally rinsed in 50 mL of sterile dH 2 0, blotted on sterile filter paper, and stored at −80 °C until DNA isolation. For each time point, at least two aliquots of the same PBS-T buffer used to wash samples were collected and further processed along samples as blanks. DNA was extracted under sterile conditions from 0.5 g of unplanted soil or 500 µL of rhizospheric soil slurry with the NucleoSpin Soil kit (Macherey-Nagel, Düren, Germany) and from 20 mg freeze-dried root material using the NucleoSpin Plant II Mini kit (Macherey-Nagel) following manufacturer’s recommendations. DNA quantity and purity were assessed using a Nanodrop-1000 instrument (Thermo Scientific, Wilmington, Germany). DNA materials were sent for gene marker amplification and sequencing to IGA Technology Services (Udine, Italy; http://igatechnology.com/ ). For Prokaryotic communities profiling the V4 16S region was selected using primers pairs 515F (5’-GTGYCAGCMGCCGCGGTAA-3’) and 806R (5’-GACTACNVGGGTWTCTAAT-3’) , linked with the Illumina adapters overhang. Amplification on organellar rDNA was prevented by peptide nucleic acids (PNAs) clamping using universal pPNA and mPNA clamps for plastidial and mitochondrial sequences following the manufacturer’s protocol (PNA Bio Inc, Newbury Park, CA). For fungi, the ITS2 region of the rRNA gene was adopted as marker using primers pair fTIS7 (5’-GTGARTCATCGAATCTTTG-3’) and ITS4 (5’-TCCTCCGCTTATTGATATGC-3’) . Libraries from both target regions (16 S and ITS2) were then constructed and sequenced on an Illumina NovaSeq6000 platform (Illumina, San Diego, CA, USA) with a 2 × 250 bp sequencing layout. Bioinformatics Amplicon libraries were inspected for quality using FastQC v0.11.9 and multiQC v1.11 software and raw reads imported into QIIME 2 (Quantitative Insights Into Microbial Ecology) v2022.02 for denoising, Amplicon Sequence Variants (ASVs) detection and taxonomy mapping. First, primers were fully removed from reads using the cutadapt ‘trim-paired’ plugin discarding untrimmed sequences. For ITS2 libraries the full-length ITS2 region was selected using ITSxpress plugin with the built-in fungal database to increase taxonomic resolution . Clean reads were then denoised and merged into ASVs using DADA2 plugin in ‘pooled’ chimera method detection and applying a reads truncation of 180 and 160 bp based on quality profiles for R1 and R2 sequences, respectively. No reads truncation was applied for ITS2 libraries (--p-trunc-len 0). Variants were then taxonomically annotated using a Naïve-Bayes classifier via the ‘feature-classifier classify-sklearn’ plugin . The SILVA v138 database (99% clustering) pre-formatted for QIIME , and the UNITE + INSDC v8.3 database in developer and dynamic mode were used as reference databases for 16S and ITS2 libraries, respectively. Tables were further taxonomy-filtered to obtain the final feature table analyzed. For the 16S dataset, ASVs matching organellar (mitochondria and chloroplast) rDNA, or without any match (unassigned at the domain level) were removed while for ITS2 libraries non-fungal sequences were discarded. The obtained feature tables was imported into R v4.2.1 environment (R Core Team, 2023) and contaminants were removed using the extraction-blank samples with the microDecon package . Alpha-and beta-diversity analyses were performed using ‘phyloseq’ v1.40.0 , ‘vegan’ v2.6-2 , and ‘QsRutils’ v0.1.5 . The ASVs count table was first filtered by removing low-abundance ASVs using ‘HTSFilter’ v1.36.0 and then normalized with a rarefaction-free approach using DEseq2 v1.36.0 , . Analyses of β- diversity were performed on the resulting normalized table. PERMANOVA and pairwise PERMANOVA analyses were performed using the adonis function of the R package ‘vegan’ and the package ‘pairwiseAdonis’ v0.4.1 , respectively. Principal coordinate analysis (PCoA) was performed by multidimensional scaling (MDS) of Bray–Curtis distance matrices using cmdscale R function. Constrained ordination (cPCoA) were computed using the ‘vegan’ capscale function (which implements CAP (Canonical analysis of principal coordinates) by constraining the factor of interest. Significance of constraints was assessed using the ANOVA-like permutation test implemented in the anova.cca function from the vegan package (999 permutations, P < 0.05). Compartment enrichment and differential abundance analyses were performined using DESeq2 package applying a zero-tolerant geometric mean formula, as detailed in phyloseq package vignettes, and adopting an FDR threshold of 0.05 to define enriched/depleted taxa. Phylogenetic heatmaps were obtained using the ggtreeExtra R package keeping only highly-abundant taxa (relative abundance > 5%) annotated at least at family level. Briefly, ASV sequences of differentially abundant taxa were aligned using MUSCLE (default parameters), and approximately-ML phylogenetic trees were obtained using FastTree v2.1.11 using the GTR model and enabling ‘-no2nd’ option and setting SPRs number as 4. Graphical elaborations were performed using ‘ggtern’ v3.3.5 and ggplot2 v3.3.6 packages. Network analysis Co-abundance networks of the root endosphere prokaryotic community for each treatment and time points were inferred using network analysis. The most abundant taxa (occurring in > 50% of the samples with at least 350 reads across all the considered samples) were selected for each of the subsampled tables and co-abundance networks inferred using the SPIEC-EASI (Sparse Inverse Covariance Estimation for Ecological Association and Statistical Inference) algorithm in SpiecEasi R package v1.1.2 using the Meinshausen and Bühlmann neighborhood selection model, a lambda path of 100 and other parameters at default values. The final model was selected by random subsampling and interaction re-estimation using stability Approach to Regularization Selection (StARS) and pulsar packages using 100 random subsamples at 0.05% variability threshold. The obtained adjacency matrices were imported into igraph objects and networks plotted using ggraph R package v2.1.0 . Network statistics and node’s topological parameters were calculated using igraph R package v1.5.1 and differences across conditions were assessed using the Student’s t -test ( P < 0.05). For each of the obtained networks, 1000 re-wired random networks (same number of nodes and edges as the real ones) were obtained using the Erdős–Rényi model with the ‘sample_gnm’ function in igraph and metrics calculated as detailed above. Differences of the topological metrics between random and real networks were assessed using the Z-test ( p < 0.05) within the BSDA v1.2.2 R package . Keystone species (hubs-taxa) were identified as the top 5% ASVs showing maximum closeness centrality and betweenness centrality metrics according to a log-normal distribution , . Gas chromatography-mass spectrometry (GC-MS) analysis Powdered shoot and root material (50 mg ± 10%) was extracted in 700 µL of methanol by adding 10 µg/mL of methyl α-D-glucopyranoside as internal standard. The samples were homogenized by vortexing and by using a Ball Mill (Retsch, MM 300) with 5 mm zirconia balls (3 min, 20 Hz) and then centrifuged (10 min, 21000 g) recovering 600 µL of the resulting supernatant. The sample was mixed by vortexing with 300 µL of chloroform and 750 µL of water and centrifuged (10 min, 21000 g ). Aliquots of 100 µL and 300 µL from the polar phase were dried in a SpeedVac™ concentrator for GC-MS analysis, respectively. Samples for GC-MS were derivatized according to Lisec et al. . The samples were re-suspended in 40 µL of methoxyaminhydrochloride (20 mg/mL in pyridine) and shaked for 2 h (37 C°, 900 rpm). After that, 70 µL MSTFA were added, and the samples were mixed for additional 30 min (37 C°) and transferred to a glass vial for GC-MS analysis. GC-MS analysis was performed on an Agilent 7890A GC system coupled to a Pegasus HT high throughput TOF/MS (LECO). 1 µL of the sample was injected at 230 °C in splitless mode with He as a carrier gas (2 mL/min). The flow rate was kept constant with electronic pressure control enabled. Chromatography was performed in a 30 m MDN-35 capillary column, with the following temperature program: isothermal for 2 min at 80 °C, followed by a 15 °C per min ramp to 330 °C, and isothermal for 6 min at 330 °C. Transfer line and ion source temperatures were set to 250 °C. The recorded mass range was set from m/z 70 to m/z 600 at 20 scans per second. The remaining monitored chromatography time was preceded by a 170 s solvent delay with filaments turned off. The manual mass defect was set to filament bias current to − 70 V, and detector voltage to ~ 1700–1850 V. The obtained chromatograms were converted to .raw file format and analyzed using the Xcalibur 2.2 software (Thermo Fisher). GC-MS peaks were annotated by comparing retention indexes relative to a mixture of fatty acid methyl esters (FAMES) and spectra similarity against metabolites from the Golm metabolome database (GMD) . Statistical analysis was conducted using MetaboAnalyst software. Five different biological replicates were analyzed for each condition. Biochemical analyses Sample extraction The Pereira-Caro et al. method was adapted to simultaneously extract the target lipophilic components from rice seeds, with some modifications. Dehulled rice seeds collected from panicles were grounded using a TissueLyser (Retsch) machine (25 Hz, 2 min) to obtain 2 g for each genotype of rice flour from whole grain (brown rice) and extracted for 1 h in an ultrasonicator with 10 mL of ethanol/hexane (4:3, v/v) mixture containing 0.1% ascorbic acid (w/v). After homogenization, samples were centrifuged for 15 min (9000 rpm at 20 °C). The obtained pellets were re-extracted twice using 5 mL hexane, mixed by vortexing, sonicated for 1 h and centrifuged as described above. The resuspended pellet was pooled and washed first with 10 mL distilled water and 5 mL of 10% NaCl solution. The organic phase was retained and reduced to almost dryness using a rotary vacuum evaporator at 35 °C. A small amount of 90% ethanol was added to remove adhering residues on the wall. The concentrated extract was frozen at −20 °C, freeze-dried for 24 h and stored in the dark at 4 °C for further analyses. Three different biological replicates were analyzed for each condition. Antioxidant analysis The antioxidant assays such as DPPH (2,2-Diphenyl-1-picrylhydrazyl), FRAP (Ferric Reducing Antioxidant Power), and ABTS (2,2’-Azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)) were performed using the previous method . Briefly, Trolox in ethanol (serial dilutions) was used as a positive control, and a blank control was prepared. DPPH, FRAP, and ABTS are measured in 515 nm, 620 nm, and 734 nm, respectively. All antioxidant values were expressed as Trolox equivalents/100 g rice (µmol TE/100 g). Three different biological replicates were analyzed for each condition. Total starch quantification The rice seeds were dehulled and milled as described before and total starch spectrophotometrically-quantified using the Total Starch (AA/AMG) Assay Kit (Megazyme Ltd., Ireland), following the manufacturer’s instructions. Three different biological replicates were analyzed for each condition. Mineral determination The minerals were quantified in the rice seed samples using the Inductively Coupled Plasma Optical Emission spectroscopy (ICP-OES) following a previously published method . From the digested samples of rice, twelve minerals such as Al, Ca, Cu, Fe, K, Mg, Mn, Mo, Na, P, S, and Zn were quantified. Statistical analysis All the statistical analyses were performed in the R statistical environment (R Core Team, 2023). Data normality and homoscedasticity were tested using Shapiro–Wilk and Levene’s test using the ‘stats’ and ‘car’ v3.1–2packages , respectively (P < 0.05). According to data distributions, ANOVA for normal homoscedastic data or Kruskal–Wallis test for non-normal homoscedastic data were applied (P < 0.05) using custom base R function or the ‘agricolae’ package . Multiple comparisons between treatments were performed using Tukey’s HSD or Dunn’s post-hoc tests after ANOVA or Kruskal-Wallis respectively (p < 0.05), using the package ‘agricolae’ v1.3 or ‘rstatix’ 0.7.2 . When comparing two experimental groups the Welch t-test was applied as implemented in package ‘ggpubr’ v0.6.0 . All data visualizations were performed in R using ‘ggplot2’ v3.3.6 . The impact of zaxinone, MiZax3, and MiZax5 on native paddy soil and rice rhizomicrobiota was studied in a greenhouse mesocosm experiment. Four soil treatments were considered, namely zaxinone, MiZax3, MiZax5, and the solvent acetone as control (ACE), for both rice plants and unplanted soil. Plants were sampled at three phenological stages, the tillering stage (60 days after transplanting, T1) the milky-stage maturation (120 days after transplanting, T2), and the over-ripe stage (180 days after transplanting, T3). For each treatment a total of 33 plants were grown: 15 for phenotyping sampled at T3 and 18 for microbiota profiling (sampled at T1 and T2, 9 biological replicates each). For the unplanted soil treatment, 18 replicates were collected (9 for each sampling point). Mesocosm systems were set up using native paddy soil harvested from the experimental fields of the CREA-CI research center (Vercelli, Italy) which was used in previous studies , . Soil physico-chemical parameters measured on a representative batch used for this study are reported in Table . Rice seeds ( Oryza sativa cv. ‘Nipponbare’) were sown in alveolar trays filled with soil. After 1 month of growth under controlled conditions, plants were transferred to the final plastic pots (10 × 9 × 17 cm) filled with the same soil. For unplanted soil treatment, one alveolar tray was left unsown and the resulting soil cores were transferred into pots with fresh soils (as for rice seedlings) for the unplanted soil conditions. Plants and unplanted soils were grown in the greenhouse at the Department of Life Sciences and Systems Biology of the University of Torino from June 2021 to October 2021 (rice growing season) without monitoring light, temperature and humidity. Plants were watered once a week with tap water and once with distilled water containing zaxinone, MiZax3, and MiZax5 molecules dissolved individually to reach the final concentration of 5µM (10 −6 ), which has already been shown to promote growth activity in rice plant and to alleviate infestation by the root parasitic plant Striga , . Fifty mL of the solution was poured into the soil of each pot once every two weeks for about 5 months (10 treatments) to cover the vegetative and rice reproductive growth stages. Plants were sampled for microbiota profiling at the tillering stage (T1) and at the milky-stage maturation (T2). At sampling, plants were removed from pots, vigorously shaken, and 10–15 g of roots collected within 3–4 cm from the base of the stem into a 50-mL Falcon tube. Unplanted soil samples were collected using a sterile spoon from the middle core of the pot, discarding edges that were in contact with plastic or any other portions where plant roots were visible. Samples were stored at + 4 °C and processed within 24 h to separate plant root compartments under sterile conditions. Samples were then processed to isolate the rhizosphere from the root endosphere according to the protocol by Bulgarelli et al. with minor modifications. Roots fragments were first washed into 10 mM sterile phosphate-buffer saline with 0.02% Tween-80 added (PBS-T), under continuous stirring on a horizontal shaker (15 min, 70 rpm). To obtain the rhizosphere soil slurry, roots were removed and tubes were centrifuged (4000 g, 10 min). The rhizosphere was then resuspended in 2 mL PBS-T and snap-freezed in liquid nitrogen. Roots were enriched in the endospheric compartment by two washes of 10 sonication cycles (30 s pulses, 30 s rest each) in PBS-T, discarding and replacing the buffer after each wash. Samples were finally rinsed in 50 mL of sterile dH 2 0, blotted on sterile filter paper, and stored at −80 °C until DNA isolation. For each time point, at least two aliquots of the same PBS-T buffer used to wash samples were collected and further processed along samples as blanks. DNA was extracted under sterile conditions from 0.5 g of unplanted soil or 500 µL of rhizospheric soil slurry with the NucleoSpin Soil kit (Macherey-Nagel, Düren, Germany) and from 20 mg freeze-dried root material using the NucleoSpin Plant II Mini kit (Macherey-Nagel) following manufacturer’s recommendations. DNA quantity and purity were assessed using a Nanodrop-1000 instrument (Thermo Scientific, Wilmington, Germany). DNA materials were sent for gene marker amplification and sequencing to IGA Technology Services (Udine, Italy; http://igatechnology.com/ ). For Prokaryotic communities profiling the V4 16S region was selected using primers pairs 515F (5’-GTGYCAGCMGCCGCGGTAA-3’) and 806R (5’-GACTACNVGGGTWTCTAAT-3’) , linked with the Illumina adapters overhang. Amplification on organellar rDNA was prevented by peptide nucleic acids (PNAs) clamping using universal pPNA and mPNA clamps for plastidial and mitochondrial sequences following the manufacturer’s protocol (PNA Bio Inc, Newbury Park, CA). For fungi, the ITS2 region of the rRNA gene was adopted as marker using primers pair fTIS7 (5’-GTGARTCATCGAATCTTTG-3’) and ITS4 (5’-TCCTCCGCTTATTGATATGC-3’) . Libraries from both target regions (16 S and ITS2) were then constructed and sequenced on an Illumina NovaSeq6000 platform (Illumina, San Diego, CA, USA) with a 2 × 250 bp sequencing layout. Amplicon libraries were inspected for quality using FastQC v0.11.9 and multiQC v1.11 software and raw reads imported into QIIME 2 (Quantitative Insights Into Microbial Ecology) v2022.02 for denoising, Amplicon Sequence Variants (ASVs) detection and taxonomy mapping. First, primers were fully removed from reads using the cutadapt ‘trim-paired’ plugin discarding untrimmed sequences. For ITS2 libraries the full-length ITS2 region was selected using ITSxpress plugin with the built-in fungal database to increase taxonomic resolution . Clean reads were then denoised and merged into ASVs using DADA2 plugin in ‘pooled’ chimera method detection and applying a reads truncation of 180 and 160 bp based on quality profiles for R1 and R2 sequences, respectively. No reads truncation was applied for ITS2 libraries (--p-trunc-len 0). Variants were then taxonomically annotated using a Naïve-Bayes classifier via the ‘feature-classifier classify-sklearn’ plugin . The SILVA v138 database (99% clustering) pre-formatted for QIIME , and the UNITE + INSDC v8.3 database in developer and dynamic mode were used as reference databases for 16S and ITS2 libraries, respectively. Tables were further taxonomy-filtered to obtain the final feature table analyzed. For the 16S dataset, ASVs matching organellar (mitochondria and chloroplast) rDNA, or without any match (unassigned at the domain level) were removed while for ITS2 libraries non-fungal sequences were discarded. The obtained feature tables was imported into R v4.2.1 environment (R Core Team, 2023) and contaminants were removed using the extraction-blank samples with the microDecon package . Alpha-and beta-diversity analyses were performed using ‘phyloseq’ v1.40.0 , ‘vegan’ v2.6-2 , and ‘QsRutils’ v0.1.5 . The ASVs count table was first filtered by removing low-abundance ASVs using ‘HTSFilter’ v1.36.0 and then normalized with a rarefaction-free approach using DEseq2 v1.36.0 , . Analyses of β- diversity were performed on the resulting normalized table. PERMANOVA and pairwise PERMANOVA analyses were performed using the adonis function of the R package ‘vegan’ and the package ‘pairwiseAdonis’ v0.4.1 , respectively. Principal coordinate analysis (PCoA) was performed by multidimensional scaling (MDS) of Bray–Curtis distance matrices using cmdscale R function. Constrained ordination (cPCoA) were computed using the ‘vegan’ capscale function (which implements CAP (Canonical analysis of principal coordinates) by constraining the factor of interest. Significance of constraints was assessed using the ANOVA-like permutation test implemented in the anova.cca function from the vegan package (999 permutations, P < 0.05). Compartment enrichment and differential abundance analyses were performined using DESeq2 package applying a zero-tolerant geometric mean formula, as detailed in phyloseq package vignettes, and adopting an FDR threshold of 0.05 to define enriched/depleted taxa. Phylogenetic heatmaps were obtained using the ggtreeExtra R package keeping only highly-abundant taxa (relative abundance > 5%) annotated at least at family level. Briefly, ASV sequences of differentially abundant taxa were aligned using MUSCLE (default parameters), and approximately-ML phylogenetic trees were obtained using FastTree v2.1.11 using the GTR model and enabling ‘-no2nd’ option and setting SPRs number as 4. Graphical elaborations were performed using ‘ggtern’ v3.3.5 and ggplot2 v3.3.6 packages. Co-abundance networks of the root endosphere prokaryotic community for each treatment and time points were inferred using network analysis. The most abundant taxa (occurring in > 50% of the samples with at least 350 reads across all the considered samples) were selected for each of the subsampled tables and co-abundance networks inferred using the SPIEC-EASI (Sparse Inverse Covariance Estimation for Ecological Association and Statistical Inference) algorithm in SpiecEasi R package v1.1.2 using the Meinshausen and Bühlmann neighborhood selection model, a lambda path of 100 and other parameters at default values. The final model was selected by random subsampling and interaction re-estimation using stability Approach to Regularization Selection (StARS) and pulsar packages using 100 random subsamples at 0.05% variability threshold. The obtained adjacency matrices were imported into igraph objects and networks plotted using ggraph R package v2.1.0 . Network statistics and node’s topological parameters were calculated using igraph R package v1.5.1 and differences across conditions were assessed using the Student’s t -test ( P < 0.05). For each of the obtained networks, 1000 re-wired random networks (same number of nodes and edges as the real ones) were obtained using the Erdős–Rényi model with the ‘sample_gnm’ function in igraph and metrics calculated as detailed above. Differences of the topological metrics between random and real networks were assessed using the Z-test ( p < 0.05) within the BSDA v1.2.2 R package . Keystone species (hubs-taxa) were identified as the top 5% ASVs showing maximum closeness centrality and betweenness centrality metrics according to a log-normal distribution , . Powdered shoot and root material (50 mg ± 10%) was extracted in 700 µL of methanol by adding 10 µg/mL of methyl α-D-glucopyranoside as internal standard. The samples were homogenized by vortexing and by using a Ball Mill (Retsch, MM 300) with 5 mm zirconia balls (3 min, 20 Hz) and then centrifuged (10 min, 21000 g) recovering 600 µL of the resulting supernatant. The sample was mixed by vortexing with 300 µL of chloroform and 750 µL of water and centrifuged (10 min, 21000 g ). Aliquots of 100 µL and 300 µL from the polar phase were dried in a SpeedVac™ concentrator for GC-MS analysis, respectively. Samples for GC-MS were derivatized according to Lisec et al. . The samples were re-suspended in 40 µL of methoxyaminhydrochloride (20 mg/mL in pyridine) and shaked for 2 h (37 C°, 900 rpm). After that, 70 µL MSTFA were added, and the samples were mixed for additional 30 min (37 C°) and transferred to a glass vial for GC-MS analysis. GC-MS analysis was performed on an Agilent 7890A GC system coupled to a Pegasus HT high throughput TOF/MS (LECO). 1 µL of the sample was injected at 230 °C in splitless mode with He as a carrier gas (2 mL/min). The flow rate was kept constant with electronic pressure control enabled. Chromatography was performed in a 30 m MDN-35 capillary column, with the following temperature program: isothermal for 2 min at 80 °C, followed by a 15 °C per min ramp to 330 °C, and isothermal for 6 min at 330 °C. Transfer line and ion source temperatures were set to 250 °C. The recorded mass range was set from m/z 70 to m/z 600 at 20 scans per second. The remaining monitored chromatography time was preceded by a 170 s solvent delay with filaments turned off. The manual mass defect was set to filament bias current to − 70 V, and detector voltage to ~ 1700–1850 V. The obtained chromatograms were converted to .raw file format and analyzed using the Xcalibur 2.2 software (Thermo Fisher). GC-MS peaks were annotated by comparing retention indexes relative to a mixture of fatty acid methyl esters (FAMES) and spectra similarity against metabolites from the Golm metabolome database (GMD) . Statistical analysis was conducted using MetaboAnalyst software. Five different biological replicates were analyzed for each condition. Sample extraction The Pereira-Caro et al. method was adapted to simultaneously extract the target lipophilic components from rice seeds, with some modifications. Dehulled rice seeds collected from panicles were grounded using a TissueLyser (Retsch) machine (25 Hz, 2 min) to obtain 2 g for each genotype of rice flour from whole grain (brown rice) and extracted for 1 h in an ultrasonicator with 10 mL of ethanol/hexane (4:3, v/v) mixture containing 0.1% ascorbic acid (w/v). After homogenization, samples were centrifuged for 15 min (9000 rpm at 20 °C). The obtained pellets were re-extracted twice using 5 mL hexane, mixed by vortexing, sonicated for 1 h and centrifuged as described above. The resuspended pellet was pooled and washed first with 10 mL distilled water and 5 mL of 10% NaCl solution. The organic phase was retained and reduced to almost dryness using a rotary vacuum evaporator at 35 °C. A small amount of 90% ethanol was added to remove adhering residues on the wall. The concentrated extract was frozen at −20 °C, freeze-dried for 24 h and stored in the dark at 4 °C for further analyses. Three different biological replicates were analyzed for each condition. Antioxidant analysis The antioxidant assays such as DPPH (2,2-Diphenyl-1-picrylhydrazyl), FRAP (Ferric Reducing Antioxidant Power), and ABTS (2,2’-Azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)) were performed using the previous method . Briefly, Trolox in ethanol (serial dilutions) was used as a positive control, and a blank control was prepared. DPPH, FRAP, and ABTS are measured in 515 nm, 620 nm, and 734 nm, respectively. All antioxidant values were expressed as Trolox equivalents/100 g rice (µmol TE/100 g). Three different biological replicates were analyzed for each condition. Total starch quantification The rice seeds were dehulled and milled as described before and total starch spectrophotometrically-quantified using the Total Starch (AA/AMG) Assay Kit (Megazyme Ltd., Ireland), following the manufacturer’s instructions. Three different biological replicates were analyzed for each condition. Mineral determination The minerals were quantified in the rice seed samples using the Inductively Coupled Plasma Optical Emission spectroscopy (ICP-OES) following a previously published method . From the digested samples of rice, twelve minerals such as Al, Ca, Cu, Fe, K, Mg, Mn, Mo, Na, P, S, and Zn were quantified. All the statistical analyses were performed in the R statistical environment (R Core Team, 2023). Data normality and homoscedasticity were tested using Shapiro–Wilk and Levene’s test using the ‘stats’ and ‘car’ v3.1–2packages , respectively (P < 0.05). According to data distributions, ANOVA for normal homoscedastic data or Kruskal–Wallis test for non-normal homoscedastic data were applied (P < 0.05) using custom base R function or the ‘agricolae’ package . Multiple comparisons between treatments were performed using Tukey’s HSD or Dunn’s post-hoc tests after ANOVA or Kruskal-Wallis respectively (p < 0.05), using the package ‘agricolae’ v1.3 or ‘rstatix’ 0.7.2 . When comparing two experimental groups the Welch t-test was applied as implemented in package ‘ggpubr’ v0.6.0 . All data visualizations were performed in R using ‘ggplot2’ v3.3.6 . Zaxinone and its mimics shape soil and root prokaryotic communities but not mycobiota assembly To investigate the impact of exogenous treatment with zaxinone and its mimic molecules (MiZax3 and MiZax5) on soil and plant-associated microbial communities, deep 16S and ITS2 rRNA gene amplicon sequencing on unplanted soil and rice rhizosphere and endosphere collected at two timepoints (tillering and milky stage) were performed. A total of ~ 40 and ~ 50 M high-quality reads for the 16S and ITS2 markers, were obtained respectively. After primers removal, sequence denoising, ASV calling, and removal of non-target sequences (plant organellar DNA and non-fungal sequences), a total of 20,875,118 and 20,600,037 fragments were retained and used for subsequent analyses for 16S and ITS2 markers, respectively (Dataset S1). A total of 31,797 16S ASVs (bASVs) and 2712 ITS2 ASVs (fASVs) were obtained, optimally covering diversity for both markers (Figure - ). The root endospheric prokaryotic community was dominated by Proteobacteria, Myxococcota, Chloroflexi, Actinobacteriota, and Bacteroidota, while in rhizosphere and soil, a higher abundance of Acidobacteria and Planctomycetes was detected (Fig. A); this trend is in line with literature data describing the rice microbiota , . The root endophytic fungal community was dominated by Sordariomycetes and Dothideomycetes class (Ascomycota), while in rhizosphere and soil, there was a prevalence of Agaricomycetes (Basidiomycota) and Mortierellomycetes (Mortierellomycotina, Fig. B). Principal Coordinate Analysis (PCoA) on both datasets showed that the influence of zaxinone and MiZax treatments was less marked compared to the effect of compartment factors, validating the robustness of the protocol used to collect rhizocompartments (Figure , Table ). However, at least for prokaryotic communities, PERMANOVA analysis indicated a significant effect of treatments ( P < 0.05, 9999 permutations; Table ). Indeed, constraining PCoA ordinations by treatment factor using canonical analysis of principal coordinates in all compartments and timepoints, a clear separation of bacterial communities across treatments emerged ( P < 0.05; Fig. C) with a closer clustering of Prokaryotic communities in plants/soils treated with MiZax3 and MiZax5 and a neat separation of zaxinone-treated samples at both timepoints, with the two mimics exerting a less marked influence compared with the acetone control. By contrast, the treatments had a lower impact on fungal communities with a less clear separation between conditions and a non-significant effect on β-diversity (PERMANOVA, P > 0.05) (Fig. D). Since all the factors as well as their interactions showed a high impact on prokaryotic community assembly, PERMANOVA was performed testing of the impact of treatments at individual time points/compartments combinations (Table ). Treatments had a significant impact on bacterial communities at T1 in both root endosphere and unplanted soil but not on rhizosphere. At T2, treatments significantly influenced the assembly of prokaryotic communities in all rhizo-compartments with the highest effect on the root endosphere (29.36% of explained variance). Still, no treatment effect was detected in fungal communities. In pairwise PERMANOVA analysis (Table ), it was evident that the individual contribution of the different molecules influences the prokaryotic community abundance: the influence of zaxinone and MiZax5 treatments was significant in the unplanted soil, compared to the control, and in endosphere, especially at T2, all the tested molecules gave a significant impact. Such variations in community assembly were evident when comparing relative abundances of the most abundant bacterial orders. For example, zaxinone treatment decreased the relative abundance of Pedospherales in the endosphere, while MiZax3 promoted their abundance in the rhizosphere at T1. Both MiZax molecules significantly decreased the amount of Polyangiales in the soil at T1 while at T2 MiZax5 decreased Rhizobiales levels. Besides, at T2, zaxinone significantly lowered the relative abundance of Anaerolinales in soil, while MiZax3 increased Chitinophagales in the endosphere (Figure A-B). As reflected in previous PERMANOVA analysis, minor variations occurred in mycobiota at T1: at this time point, MiZax5 led to significantly higher Pleosporales levels in the root endosphere (Figure S5 A-B). No effect of the treatments was observed on bacterial and fungal Shannon index ( α - diversity) (Fig. E and Figure S6) at both time points for both soil and rhizosphere samples. A discernible trend toward an increase in diversity was evident in the root endosphere during the reproductive stage (T2) of the 16S rDNA amplicon dataset irrespective of the treatments (Figure S6C). Furthermore, looking at the Shannon diversity index, differences between T1 and T2 within each treatment emerged indicating that zaxinone and MiZax can influence microbiota dynamics across plant phenological stages (Fig. E-F and S6). In more detail, in the endosphere the prokaryotic -diversity increased across time points for all the treatments considered, including the acetone control (Fig. E and S6). Regarding the fungal communities, zaxinone and MiZax5 elicited a substantial elevation in -diversity from T1 to T2 (Fig. F) while no effect emerged in the control. This trend is also detectable in the rhizosphere, where zaxinone increased fungal -diversity across time points (Figure S6B). Notably, no discernible alterations in α -diversity were noted in the unplanted soil samples among T1 and T2, when compared with the trend of the acetone control, except for an increase in MiZax5 treatment for the prokaryotic community (Figure S6A). Overall, these results indicate that zaxinone and MiZax treatments exerted a mild effect on the prokaryotic community assembly with variations mainly related to compartments and time points while having a minor impact on the fungal community. Nevertheless, when comparing time points within the same treatment, it was found that zaxinone increased fungal α -diversity in root endosphere and rhizosphere in T2 vs. T1 compared to the control acetone condition, while MiZax5 had the same effect in unplanted soil considering the prokaryotic communities. This highlights the possible role of Zaxinone and its mimics in shaping root and soil communities across different plant life stages. Zaxinone and MiZax modulate microbial recruitment dynamics along the soil-root interface To investigate the recruitment of microbiota by rice plants along the soil-rhizosphere-endosphere continuum and to assess the potential interference of zaxinone and MiZax treatments in this process, compartment-specific ASVs, defined as those occurring in higher abundance in a particular compartment compared to others across treatments and time points were analyzed. The results revealed distinct patterns of bacterial taxa enrichment across compartments in the various treatments, exhibiting a pronounced timepoint-dependent trend. Specifically, at T1, there was an increase in the number of rhizosphere-enriched taxa under zaxinone, MiZax3, and MiZax5 treatments (Fig. ). Concurrently, there was an increase in root endosphere-enriched taxa compared to the acetone control in all treatments except for the zaxinone treatment. At T2, a decrease in rhizosphere- and root-enriched ASVs upon the treatment with zaxinone and MiZax3 was observed, whereas the application of MiZax5 caused an increase in root- and rhizosphere-enriched taxa (Fig. ). Overall, MiZax5 treatment proved most effective in increasing the number of highly specific rhizosphere taxa at both time points. Notably, treatments had also an effect on soil-enriched ASVs. With the exception of MiZax5, all the treatments reduced and increased soil-enriched ASVs at T1 and T2, respectively (Fig. ). At each time point, all treatments showed a shared core of root-enriched taxa which included Comamonadaceae, Rhizobiaceae, Chloroflexaceae, and Microscillaceae at T1 with the addition of a Novosphingobium sp. at T2. Among this root-specific set of taxa, MiZax treatments consistently recruited specific Acidibacter sp., Devosia sp., and Comamonadaceae ASVs at T1 (Figure S7). Altogether, these data suggest that zaxinone and MiZax treatments exert different effects on microbiota recruitment according to the timepoint considered. Remarkably, during the 15 days following the application of treatments (T1 sampling), plants exhibited the recruitment of distinct endosphere and rhizosphere communities. Nevertheless, by T2, the bacterial community assemblies in both compartments displayed increased homogeneity, marked by a reduction in compartment-specific taxa and an augmentation of taxa shared among different compartments. The sole exception to this trend was observed in MiZax5, which consistently maintained highly compartment-exclusive communities at both time points and across compartments (Fig. and S8). Zaxinone and MiZax modulate bacterial and fungal taxa in the root endosphere The analysis of differential abundance (Fig. ) at the ASV level revealed that zaxinone and MiZax molecules exerted the most substantial impacts on root endosphere communities, evidenced by a higher number of differentially abundant taxa at both T1 and T2, for both prokaryotic and fungal communities. In contrast, the rhizosphere and soil exhibited a lower number of taxa with altered abundance following treatments. This pattern was notably pronounced in prokaryotic communities (Fig. A), and a comparable trend was observed in Fungi, albeit with a limited number of differentially abundant taxa across conditions in the latter case (Fig. C). Furthermore, our observations indicate that at T1, the majority of taxa exhibited a negative response to almost all treatments in the root, rhizosphere, and soil. In the root endosphere, the number of enriched/depleted prokaryotic taxa at T2 diverged upon zaxinone and MiZax3 treatments. In particular, upon MiZax3 treatment an increase of depleted taxa was identified while the MiZax5 treatment resulted in a higher number of enriched taxa compared to the other treatments (Fig. A). Considering the taxonomic diversity of differentially abundant taxa in the root endosphere, it was observed that depleted taxa spanned across the bacterial phylogeny, with the exception of Actinobacteriota which mostly increased in their abundance in all treatments at T1 (Fig. B). The phyla Acidobacteriota and Proteobacteria are generally depleted. In particular, within Proteobacteria the family Nitrosomonadaceae (genus Ellin6067 ) is decreased in our dataset after the treatments at the first timepoint, while the relative abundance of Comamonadaceae (Proteobacteria) increased with MiZax5 treatment. The application of the molecules was consistently linked to a reduction in Chloroflexi abundance both in the vegetative and in reproductive plant stages. Among depleted taxa, Fimbriimonadaceae (Armatimonadota phylum) emerged especially upon zaxinone and MiZax5 treatments. This taxon is usually detected in ANAMMOX (ANaerobic AMMonium OXidation) consortia, implying that the Fimbriimonadaceae family either contains ammonia-oxidizing taxa or has positive interactions with ammonia-oxidizing bacteria, favoring the ammonia-oxidizing processes . Additionally, a diminished abundance of sulfate-reducing bacteria, such as Geobacteraceae, was observed across all treatments at T2. Notably, treatments significantly influenced Archaeal taxa, with methanogens (primarily Methanosarcina , Methanobacterium , Methanosaeta , and Methanocella genera) generally exhibiting increased abundance in treated samples compared to the control, particularly at T1. Notably, an increase in the abundance of Actinobacteriota, particularly at T1 was detected. This group encompasses taxa well-known to establish beneficial interactions with plants, acting both in the rhizosphere and as endophytes, stimulating plant growth and enhancing disease resistance . Conversely, other groups well-acknowledged to include plant-beneficial species including Sphingomonadaceae (Proteobacteria) and Bacillaceae (Firmicutes), decreased in abundance suggesting a treatment-induced shift of these components. Considering Fungi, our findings revealed no distinct phylogeny-related differences, as the taxa depleted in the root endosphere were distributed across all major phyla, with only a few exceptions. In particular, almost all the ASVs belonging to Helotiales (namely Talaromyces , Dimorphospora , Meliniomyces , and Hyaloscypha ASV) were enriched across treatments (Fig. D). This group includes soil fungi with marked organic matter degradation abilities that are known to associate with plant roots as endophytes and symbionts . Within Mortierellomycota, Glomeromycotina formerly Glomeromycota, genera such as Funneliformis and Claroideoglomus -related ASVs in Fig. D ) and Mortierellomycotina (formerly Mortierellomycota), seem to be mostly depleted in this compartment, with the exception of zaxinone- and MiZax5-treated plants at T2, which show an increment of different taxa, especially the genera Funneliformis and Mortierella . However, these changes do not reflect any significant changes in the whole AMF community at higher taxonomic level such as families, which is mainly composed of Glomeraceae, Paraglomeraceae, and Claroideoglomeraceae at variable abundances according to the phenological stage and additionally being slightly affected by treatment (Figure S8). Lastly, two ASVs pointing to the Trichoderma genus showed an increase in its abundance at T2 upon both Zaxinone and Mizax5 treatment. This fungus is known to exert plant-beneficial abilities, being particularly active as a biocontrol agent and increasing yield in rice . Taken together, these data suggest an overall impact of zaxinone and its mimicking molecules on the abundances of numerous microbial groups, which turned out to be mostly decreased at the first time point, with some exceptions. At T2, the influence of the treatments seems to be globally less pronounced. Zaxinone and MiZax treatment impact microbiota network dynamics more at the vegetative than at the reproductive stage To track for changes in co-occurrence dynamics determined by treatments on the bacterial endosphere community and to identify hubs-taxa (i.e. ’keystone microbes which drive the structure of the community), co-occurrence networks were constructed using Sparse Inverse Covariance estimation (SPIEC-EASI) for each timepoint and treatment considered (Fig. ). Globally, structures of the community in each treatment at each timepoint showed comparable network-level metrics, with a higher ratio of positive correlations between taxa (mean 61.1% positive edges) and an overall similar modularity, centrality metrics, and cohesion (Table S6). Notably, at the milky-stage (T2), communities showed increased connectivity, nodes degree, and number of hubs-taxa identified, indicating a higher community complexity. However, a treatment-dependent modulation was observed in most of the node-level metrics analyzed (Figure S9). At T1 all the treatments significantly decreased the degree distribution (number of connections, i.e . edges, established by each node). Zaxinone and MiZax5 treatments significantly increased betweenness i.e . the extent to which a node lies in the shortest path connecting other nodes and decreased closeness centrality i.e . the average distance to all other nodes. Further, the eigenvector-centrality and the HUB score, which both indicate the amount of connections towards highly influential taxa, were positively impacted by the treatments with the exception of MiZax5. In addition, in both zaxinone and MiZax5 the betweenness centrality significantly increased while closeness centrality decreased (Figure S9). Altogether, these metrics indicate that treatments increased distance between taxa and increased the occurrences of more isolated sub-communities, in terms of connections while decreasing the overall connections between taxa (lower degree). At T2, the degree distribution became more uniform across treatments, with the exception of MiZax3, where significantly fewer connections emerged. Concurrently, at the same timepoint, all treatments led to an increase in node closeness, and, except for MiZax5, also in betweenness centrality. Interestingly, by comparing the inferred network metrics with those of randomly generated graphs (see methods; Table S6), it was found that at both time points network structures were not casual, highlighting that reconstructed community dynamics were meaningful. To offer a more detailed insight into the impact of treatments on the community network structure, hub-taxa, i.e. the nodes characterized by higher closeness and betweenness centrality (top 5% of the distribution, Fig. ) were identified. At T1 most of the hubs-taxa detected across treatments belong to Proteobacteria, Myxococcota, and Acidobacteria including Sphingomonas , Haliangium , Vicinamibacter , and Tahibacter genera. Most of them were already known as keystone species in root- or plant-associated communities and hold plant-growth-promoting capacities. The analysis indicated that treatments resulted in a reduction of hub-taxa at T1 (from 6 to 1–2 hub-taxa in the control and treatments, respectively), while minor to no differences were detected at T2. In the control (acetone), hub-taxa primarily consisted of Chloroflexi (A4b family), Acidobacteriota, Firmicutes, Actinobacteriota, and Bacteroidota members, with the inclusion of Proteobacteria (Comamonadaceae) at T2. Moreover, in the examination of the node’s closeness distribution, a notable shift towards lower values was observed in zaxinone and MiZax5 at T1, reflecting the reduced number of established edges (degree metrics, Fig. A). At T2, keystone taxa increased compared to T1 in all treatments. Several Comamonadaceae ASVs (proteobacteria), acknowledged for their prevalence in rice-associated root communities , were identified as keystone taxa in all treatments as well as in the control. Under MiZax5 treatment, Novosphingobium and Streptomyces , two well-acknowledged PGP species, emerged as hub-taxa. Interestingly, in both MiZax3 and MiZax5 treatments, Haliangium occurred as a keystone species. Overall, analysis of co-occurrences indicated that all the treatments decreased the overall network complexity promoting the isolation of sub-communities and decreasing the number of hubs at the vegetative stage (T1), while at the reproductive stage (T2), in treated plants and in particular upon MiZax5 application, significant interactions between taxa were re-established in a similar manner to the acetone control, though with an array of hub-taxa that seems to be specific for each condition considered. Zaxinone and MiZax treatment induce alterations in plant’s primary metabolism and grain nutrient content At the metabolomic level, the effects of zaxinone-related compounds and the associated root microbiota composition on treated and non-treated rice plants were determined by collecting shoots from the same rice plants used for the metabarcoding analysis at T1 and T2. Through targeted GC-MS analysis, approximately 40 primary metabolites, including amino acids, organic acids, and sugars, were identified as differentially accumulated metabolites (Fig. ). Noteworthy alterations in metabolite levels were observed across various metabolite classes in green tissues. During the vegetative stage (T1), MiZax3 and MiZax5 treatments induce a trend of increase in the sugar content. In particular, threitol and myoinositol are significantly more abundant in the shoots upon MiZax3 treatment. Conversely, zaxinone exhibited a contrasting trend in sugar content at T1 compared to its synthetic mimics. As the plants progressed to the reproductive stage (T2), a higher trend in sugar content (glucose, glycerol, myoinositol, and fructose) has been reported under zaxinone treatment, whereas the same compounds appeared rather reduced following MiZax5 treatments, though none of the detected changes turned out to be statistically significant. A general accumulation of amino acids was observed upon treatments: in particular, an increase in the levels of different free amino acids (alanine, threonine, and GABA) was observed at T1 upon MiZax treatment. At the same timepoint, zaxinone treatment induced an accumulation of isoleucine, proline, and threonine while at T2 an increase in alanine was observed upon MiZax5. While the effect of the treatments on sugar and amino acid accumulation seems to be limited, the organic acid pattern turned out to be more influenced, especially at T1. At that time point, MiZax3 induced TCA cycle intermediates, including glycerate, pyruvate, and (2)-oxoglutarate. Similarly, at both time points MiZax5 increased glycerate and (2)-oxoglutarate content. While at T1 malate and citrate showed a statistically-supported lower abundance in shoot treated with MiZax3 and zaxinone, a slight decrease in shoot treated with MiZax5 was observed. Ribonic acid was also negatively affected mainly by both MiZax. A substantial reduction in phosphoric acid levels in the shoots after all zaxinone-related compounds treatments was also observed at T1. By contrast, an increase in salicylate levels in the shoots was observed at T2 following zaxinone treatment, whereas MiZax3 and MiZax5 exhibited an opposite trend. To a certain extent, the metabolomic profiles are in agreement with the results obtained by Wang et al. , which reported an increased concentration of free amino acids and succinate in the shoot of rice plants treated with zaxinone. However, there are some discrepancies in other metabolic pathways, namely sugar metabolism and organic acids. This may be ascribed to differences in the experimental set-up ( i.e. plant’s phenological stage and compound applications in hydroponic or soil system). Furthermore, our findings indicate that MiZax3 is the most effective compound in modulating the plant metabolome, particularly at T1. In order to determine the impact of zaxinone and MiZax compounds and the relative root-associated microbiota community on seed nutrient content, the starch content, the antioxidant capacity, and the mineral nutrient profile were evaluated. While the antioxidant activity and the starch content did not change across treatments, some differences in grain mineral nutrition were observed. Indeed, a higher content of zinc (33.87 ppm) and copper (5.73 ppm) in seeds of plants treated with zaxinone compared to control acetone plants was detected. Conversely, a decrease in manganese content was observed in seeds of plants treated with both MiZax3 and MiZax5 (Figure S10). To investigate the impact of exogenous treatment with zaxinone and its mimic molecules (MiZax3 and MiZax5) on soil and plant-associated microbial communities, deep 16S and ITS2 rRNA gene amplicon sequencing on unplanted soil and rice rhizosphere and endosphere collected at two timepoints (tillering and milky stage) were performed. A total of ~ 40 and ~ 50 M high-quality reads for the 16S and ITS2 markers, were obtained respectively. After primers removal, sequence denoising, ASV calling, and removal of non-target sequences (plant organellar DNA and non-fungal sequences), a total of 20,875,118 and 20,600,037 fragments were retained and used for subsequent analyses for 16S and ITS2 markers, respectively (Dataset S1). A total of 31,797 16S ASVs (bASVs) and 2712 ITS2 ASVs (fASVs) were obtained, optimally covering diversity for both markers (Figure - ). The root endospheric prokaryotic community was dominated by Proteobacteria, Myxococcota, Chloroflexi, Actinobacteriota, and Bacteroidota, while in rhizosphere and soil, a higher abundance of Acidobacteria and Planctomycetes was detected (Fig. A); this trend is in line with literature data describing the rice microbiota , . The root endophytic fungal community was dominated by Sordariomycetes and Dothideomycetes class (Ascomycota), while in rhizosphere and soil, there was a prevalence of Agaricomycetes (Basidiomycota) and Mortierellomycetes (Mortierellomycotina, Fig. B). Principal Coordinate Analysis (PCoA) on both datasets showed that the influence of zaxinone and MiZax treatments was less marked compared to the effect of compartment factors, validating the robustness of the protocol used to collect rhizocompartments (Figure , Table ). However, at least for prokaryotic communities, PERMANOVA analysis indicated a significant effect of treatments ( P < 0.05, 9999 permutations; Table ). Indeed, constraining PCoA ordinations by treatment factor using canonical analysis of principal coordinates in all compartments and timepoints, a clear separation of bacterial communities across treatments emerged ( P < 0.05; Fig. C) with a closer clustering of Prokaryotic communities in plants/soils treated with MiZax3 and MiZax5 and a neat separation of zaxinone-treated samples at both timepoints, with the two mimics exerting a less marked influence compared with the acetone control. By contrast, the treatments had a lower impact on fungal communities with a less clear separation between conditions and a non-significant effect on β-diversity (PERMANOVA, P > 0.05) (Fig. D). Since all the factors as well as their interactions showed a high impact on prokaryotic community assembly, PERMANOVA was performed testing of the impact of treatments at individual time points/compartments combinations (Table ). Treatments had a significant impact on bacterial communities at T1 in both root endosphere and unplanted soil but not on rhizosphere. At T2, treatments significantly influenced the assembly of prokaryotic communities in all rhizo-compartments with the highest effect on the root endosphere (29.36% of explained variance). Still, no treatment effect was detected in fungal communities. In pairwise PERMANOVA analysis (Table ), it was evident that the individual contribution of the different molecules influences the prokaryotic community abundance: the influence of zaxinone and MiZax5 treatments was significant in the unplanted soil, compared to the control, and in endosphere, especially at T2, all the tested molecules gave a significant impact. Such variations in community assembly were evident when comparing relative abundances of the most abundant bacterial orders. For example, zaxinone treatment decreased the relative abundance of Pedospherales in the endosphere, while MiZax3 promoted their abundance in the rhizosphere at T1. Both MiZax molecules significantly decreased the amount of Polyangiales in the soil at T1 while at T2 MiZax5 decreased Rhizobiales levels. Besides, at T2, zaxinone significantly lowered the relative abundance of Anaerolinales in soil, while MiZax3 increased Chitinophagales in the endosphere (Figure A-B). As reflected in previous PERMANOVA analysis, minor variations occurred in mycobiota at T1: at this time point, MiZax5 led to significantly higher Pleosporales levels in the root endosphere (Figure S5 A-B). No effect of the treatments was observed on bacterial and fungal Shannon index ( α - diversity) (Fig. E and Figure S6) at both time points for both soil and rhizosphere samples. A discernible trend toward an increase in diversity was evident in the root endosphere during the reproductive stage (T2) of the 16S rDNA amplicon dataset irrespective of the treatments (Figure S6C). Furthermore, looking at the Shannon diversity index, differences between T1 and T2 within each treatment emerged indicating that zaxinone and MiZax can influence microbiota dynamics across plant phenological stages (Fig. E-F and S6). In more detail, in the endosphere the prokaryotic -diversity increased across time points for all the treatments considered, including the acetone control (Fig. E and S6). Regarding the fungal communities, zaxinone and MiZax5 elicited a substantial elevation in -diversity from T1 to T2 (Fig. F) while no effect emerged in the control. This trend is also detectable in the rhizosphere, where zaxinone increased fungal -diversity across time points (Figure S6B). Notably, no discernible alterations in α -diversity were noted in the unplanted soil samples among T1 and T2, when compared with the trend of the acetone control, except for an increase in MiZax5 treatment for the prokaryotic community (Figure S6A). Overall, these results indicate that zaxinone and MiZax treatments exerted a mild effect on the prokaryotic community assembly with variations mainly related to compartments and time points while having a minor impact on the fungal community. Nevertheless, when comparing time points within the same treatment, it was found that zaxinone increased fungal α -diversity in root endosphere and rhizosphere in T2 vs. T1 compared to the control acetone condition, while MiZax5 had the same effect in unplanted soil considering the prokaryotic communities. This highlights the possible role of Zaxinone and its mimics in shaping root and soil communities across different plant life stages. To investigate the recruitment of microbiota by rice plants along the soil-rhizosphere-endosphere continuum and to assess the potential interference of zaxinone and MiZax treatments in this process, compartment-specific ASVs, defined as those occurring in higher abundance in a particular compartment compared to others across treatments and time points were analyzed. The results revealed distinct patterns of bacterial taxa enrichment across compartments in the various treatments, exhibiting a pronounced timepoint-dependent trend. Specifically, at T1, there was an increase in the number of rhizosphere-enriched taxa under zaxinone, MiZax3, and MiZax5 treatments (Fig. ). Concurrently, there was an increase in root endosphere-enriched taxa compared to the acetone control in all treatments except for the zaxinone treatment. At T2, a decrease in rhizosphere- and root-enriched ASVs upon the treatment with zaxinone and MiZax3 was observed, whereas the application of MiZax5 caused an increase in root- and rhizosphere-enriched taxa (Fig. ). Overall, MiZax5 treatment proved most effective in increasing the number of highly specific rhizosphere taxa at both time points. Notably, treatments had also an effect on soil-enriched ASVs. With the exception of MiZax5, all the treatments reduced and increased soil-enriched ASVs at T1 and T2, respectively (Fig. ). At each time point, all treatments showed a shared core of root-enriched taxa which included Comamonadaceae, Rhizobiaceae, Chloroflexaceae, and Microscillaceae at T1 with the addition of a Novosphingobium sp. at T2. Among this root-specific set of taxa, MiZax treatments consistently recruited specific Acidibacter sp., Devosia sp., and Comamonadaceae ASVs at T1 (Figure S7). Altogether, these data suggest that zaxinone and MiZax treatments exert different effects on microbiota recruitment according to the timepoint considered. Remarkably, during the 15 days following the application of treatments (T1 sampling), plants exhibited the recruitment of distinct endosphere and rhizosphere communities. Nevertheless, by T2, the bacterial community assemblies in both compartments displayed increased homogeneity, marked by a reduction in compartment-specific taxa and an augmentation of taxa shared among different compartments. The sole exception to this trend was observed in MiZax5, which consistently maintained highly compartment-exclusive communities at both time points and across compartments (Fig. and S8). The analysis of differential abundance (Fig. ) at the ASV level revealed that zaxinone and MiZax molecules exerted the most substantial impacts on root endosphere communities, evidenced by a higher number of differentially abundant taxa at both T1 and T2, for both prokaryotic and fungal communities. In contrast, the rhizosphere and soil exhibited a lower number of taxa with altered abundance following treatments. This pattern was notably pronounced in prokaryotic communities (Fig. A), and a comparable trend was observed in Fungi, albeit with a limited number of differentially abundant taxa across conditions in the latter case (Fig. C). Furthermore, our observations indicate that at T1, the majority of taxa exhibited a negative response to almost all treatments in the root, rhizosphere, and soil. In the root endosphere, the number of enriched/depleted prokaryotic taxa at T2 diverged upon zaxinone and MiZax3 treatments. In particular, upon MiZax3 treatment an increase of depleted taxa was identified while the MiZax5 treatment resulted in a higher number of enriched taxa compared to the other treatments (Fig. A). Considering the taxonomic diversity of differentially abundant taxa in the root endosphere, it was observed that depleted taxa spanned across the bacterial phylogeny, with the exception of Actinobacteriota which mostly increased in their abundance in all treatments at T1 (Fig. B). The phyla Acidobacteriota and Proteobacteria are generally depleted. In particular, within Proteobacteria the family Nitrosomonadaceae (genus Ellin6067 ) is decreased in our dataset after the treatments at the first timepoint, while the relative abundance of Comamonadaceae (Proteobacteria) increased with MiZax5 treatment. The application of the molecules was consistently linked to a reduction in Chloroflexi abundance both in the vegetative and in reproductive plant stages. Among depleted taxa, Fimbriimonadaceae (Armatimonadota phylum) emerged especially upon zaxinone and MiZax5 treatments. This taxon is usually detected in ANAMMOX (ANaerobic AMMonium OXidation) consortia, implying that the Fimbriimonadaceae family either contains ammonia-oxidizing taxa or has positive interactions with ammonia-oxidizing bacteria, favoring the ammonia-oxidizing processes . Additionally, a diminished abundance of sulfate-reducing bacteria, such as Geobacteraceae, was observed across all treatments at T2. Notably, treatments significantly influenced Archaeal taxa, with methanogens (primarily Methanosarcina , Methanobacterium , Methanosaeta , and Methanocella genera) generally exhibiting increased abundance in treated samples compared to the control, particularly at T1. Notably, an increase in the abundance of Actinobacteriota, particularly at T1 was detected. This group encompasses taxa well-known to establish beneficial interactions with plants, acting both in the rhizosphere and as endophytes, stimulating plant growth and enhancing disease resistance . Conversely, other groups well-acknowledged to include plant-beneficial species including Sphingomonadaceae (Proteobacteria) and Bacillaceae (Firmicutes), decreased in abundance suggesting a treatment-induced shift of these components. Considering Fungi, our findings revealed no distinct phylogeny-related differences, as the taxa depleted in the root endosphere were distributed across all major phyla, with only a few exceptions. In particular, almost all the ASVs belonging to Helotiales (namely Talaromyces , Dimorphospora , Meliniomyces , and Hyaloscypha ASV) were enriched across treatments (Fig. D). This group includes soil fungi with marked organic matter degradation abilities that are known to associate with plant roots as endophytes and symbionts . Within Mortierellomycota, Glomeromycotina formerly Glomeromycota, genera such as Funneliformis and Claroideoglomus -related ASVs in Fig. D ) and Mortierellomycotina (formerly Mortierellomycota), seem to be mostly depleted in this compartment, with the exception of zaxinone- and MiZax5-treated plants at T2, which show an increment of different taxa, especially the genera Funneliformis and Mortierella . However, these changes do not reflect any significant changes in the whole AMF community at higher taxonomic level such as families, which is mainly composed of Glomeraceae, Paraglomeraceae, and Claroideoglomeraceae at variable abundances according to the phenological stage and additionally being slightly affected by treatment (Figure S8). Lastly, two ASVs pointing to the Trichoderma genus showed an increase in its abundance at T2 upon both Zaxinone and Mizax5 treatment. This fungus is known to exert plant-beneficial abilities, being particularly active as a biocontrol agent and increasing yield in rice . Taken together, these data suggest an overall impact of zaxinone and its mimicking molecules on the abundances of numerous microbial groups, which turned out to be mostly decreased at the first time point, with some exceptions. At T2, the influence of the treatments seems to be globally less pronounced. To track for changes in co-occurrence dynamics determined by treatments on the bacterial endosphere community and to identify hubs-taxa (i.e. ’keystone microbes which drive the structure of the community), co-occurrence networks were constructed using Sparse Inverse Covariance estimation (SPIEC-EASI) for each timepoint and treatment considered (Fig. ). Globally, structures of the community in each treatment at each timepoint showed comparable network-level metrics, with a higher ratio of positive correlations between taxa (mean 61.1% positive edges) and an overall similar modularity, centrality metrics, and cohesion (Table S6). Notably, at the milky-stage (T2), communities showed increased connectivity, nodes degree, and number of hubs-taxa identified, indicating a higher community complexity. However, a treatment-dependent modulation was observed in most of the node-level metrics analyzed (Figure S9). At T1 all the treatments significantly decreased the degree distribution (number of connections, i.e . edges, established by each node). Zaxinone and MiZax5 treatments significantly increased betweenness i.e . the extent to which a node lies in the shortest path connecting other nodes and decreased closeness centrality i.e . the average distance to all other nodes. Further, the eigenvector-centrality and the HUB score, which both indicate the amount of connections towards highly influential taxa, were positively impacted by the treatments with the exception of MiZax5. In addition, in both zaxinone and MiZax5 the betweenness centrality significantly increased while closeness centrality decreased (Figure S9). Altogether, these metrics indicate that treatments increased distance between taxa and increased the occurrences of more isolated sub-communities, in terms of connections while decreasing the overall connections between taxa (lower degree). At T2, the degree distribution became more uniform across treatments, with the exception of MiZax3, where significantly fewer connections emerged. Concurrently, at the same timepoint, all treatments led to an increase in node closeness, and, except for MiZax5, also in betweenness centrality. Interestingly, by comparing the inferred network metrics with those of randomly generated graphs (see methods; Table S6), it was found that at both time points network structures were not casual, highlighting that reconstructed community dynamics were meaningful. To offer a more detailed insight into the impact of treatments on the community network structure, hub-taxa, i.e. the nodes characterized by higher closeness and betweenness centrality (top 5% of the distribution, Fig. ) were identified. At T1 most of the hubs-taxa detected across treatments belong to Proteobacteria, Myxococcota, and Acidobacteria including Sphingomonas , Haliangium , Vicinamibacter , and Tahibacter genera. Most of them were already known as keystone species in root- or plant-associated communities and hold plant-growth-promoting capacities. The analysis indicated that treatments resulted in a reduction of hub-taxa at T1 (from 6 to 1–2 hub-taxa in the control and treatments, respectively), while minor to no differences were detected at T2. In the control (acetone), hub-taxa primarily consisted of Chloroflexi (A4b family), Acidobacteriota, Firmicutes, Actinobacteriota, and Bacteroidota members, with the inclusion of Proteobacteria (Comamonadaceae) at T2. Moreover, in the examination of the node’s closeness distribution, a notable shift towards lower values was observed in zaxinone and MiZax5 at T1, reflecting the reduced number of established edges (degree metrics, Fig. A). At T2, keystone taxa increased compared to T1 in all treatments. Several Comamonadaceae ASVs (proteobacteria), acknowledged for their prevalence in rice-associated root communities , were identified as keystone taxa in all treatments as well as in the control. Under MiZax5 treatment, Novosphingobium and Streptomyces , two well-acknowledged PGP species, emerged as hub-taxa. Interestingly, in both MiZax3 and MiZax5 treatments, Haliangium occurred as a keystone species. Overall, analysis of co-occurrences indicated that all the treatments decreased the overall network complexity promoting the isolation of sub-communities and decreasing the number of hubs at the vegetative stage (T1), while at the reproductive stage (T2), in treated plants and in particular upon MiZax5 application, significant interactions between taxa were re-established in a similar manner to the acetone control, though with an array of hub-taxa that seems to be specific for each condition considered. At the metabolomic level, the effects of zaxinone-related compounds and the associated root microbiota composition on treated and non-treated rice plants were determined by collecting shoots from the same rice plants used for the metabarcoding analysis at T1 and T2. Through targeted GC-MS analysis, approximately 40 primary metabolites, including amino acids, organic acids, and sugars, were identified as differentially accumulated metabolites (Fig. ). Noteworthy alterations in metabolite levels were observed across various metabolite classes in green tissues. During the vegetative stage (T1), MiZax3 and MiZax5 treatments induce a trend of increase in the sugar content. In particular, threitol and myoinositol are significantly more abundant in the shoots upon MiZax3 treatment. Conversely, zaxinone exhibited a contrasting trend in sugar content at T1 compared to its synthetic mimics. As the plants progressed to the reproductive stage (T2), a higher trend in sugar content (glucose, glycerol, myoinositol, and fructose) has been reported under zaxinone treatment, whereas the same compounds appeared rather reduced following MiZax5 treatments, though none of the detected changes turned out to be statistically significant. A general accumulation of amino acids was observed upon treatments: in particular, an increase in the levels of different free amino acids (alanine, threonine, and GABA) was observed at T1 upon MiZax treatment. At the same timepoint, zaxinone treatment induced an accumulation of isoleucine, proline, and threonine while at T2 an increase in alanine was observed upon MiZax5. While the effect of the treatments on sugar and amino acid accumulation seems to be limited, the organic acid pattern turned out to be more influenced, especially at T1. At that time point, MiZax3 induced TCA cycle intermediates, including glycerate, pyruvate, and (2)-oxoglutarate. Similarly, at both time points MiZax5 increased glycerate and (2)-oxoglutarate content. While at T1 malate and citrate showed a statistically-supported lower abundance in shoot treated with MiZax3 and zaxinone, a slight decrease in shoot treated with MiZax5 was observed. Ribonic acid was also negatively affected mainly by both MiZax. A substantial reduction in phosphoric acid levels in the shoots after all zaxinone-related compounds treatments was also observed at T1. By contrast, an increase in salicylate levels in the shoots was observed at T2 following zaxinone treatment, whereas MiZax3 and MiZax5 exhibited an opposite trend. To a certain extent, the metabolomic profiles are in agreement with the results obtained by Wang et al. , which reported an increased concentration of free amino acids and succinate in the shoot of rice plants treated with zaxinone. However, there are some discrepancies in other metabolic pathways, namely sugar metabolism and organic acids. This may be ascribed to differences in the experimental set-up ( i.e. plant’s phenological stage and compound applications in hydroponic or soil system). Furthermore, our findings indicate that MiZax3 is the most effective compound in modulating the plant metabolome, particularly at T1. In order to determine the impact of zaxinone and MiZax compounds and the relative root-associated microbiota community on seed nutrient content, the starch content, the antioxidant capacity, and the mineral nutrient profile were evaluated. While the antioxidant activity and the starch content did not change across treatments, some differences in grain mineral nutrition were observed. Indeed, a higher content of zinc (33.87 ppm) and copper (5.73 ppm) in seeds of plants treated with zaxinone compared to control acetone plants was detected. Conversely, a decrease in manganese content was observed in seeds of plants treated with both MiZax3 and MiZax5 (Figure S10). The microbiota analysis reveals a community typical of paddy soils Rice fields represent a peculiar environment for the microbial soil communities. Due to the periodic flooding, this habitat is characterized by oxygen-limited conditions that shape the microbiota assembly. Bacterial communities typically include both aerobic and anaerobic taxa . The overall microbiota assembly from unplanted soil revealed a prokaryotic composition typical of paddy environments, including Gemmatimonadetes, Chloroflexi, Acidobacteria and Actinobacteria, and the archaeal phylum Crenarchaeota – . On the fungal side, our analysis revealed a relevant proportion of Leotiomycetes in all the conditions considered. This group of fungi includes good organic matter decomposers that can tolerate high levels of heavy metal contaminants usually found in paddy soils , and can associate with plant roots living as endophytes. Impact of zaxinone and its mimics on rhizomicrobiome diversity and composition So far, the effects of exogenous treatment with zaxinone and its mimics (MiZax3 and MiZax5) were investigated considering the growth promotion activity, the regulation of SLs biosynthesis, and the AM symbiosis , . Here, we provide comprehensively new information about the impact of these compounds on paddy soil and rice root-associated microbes and how these effects systemically influence rice metabolomic and grain nutrient profiles considering different developmental stages. The metabarcoding analysis highlighted that rice microbiome assembly is mainly influenced by compartment niche and developmental stage as already reported for rice and other plants – , regardless of zaxinone and MiZax treatments. In contrast to the findings reported by Zangh and colleagues , which observed a reduction in root microbial community richness in the later stage of the rice life cycle, our data reveals a general increase in α -diversity at T2 compared to T1 across the three considered compartments (unplanted soil, rhizosphere, and endosphere), regardless of the treatments. This increment was evident in the endosphere particularly for the prokaryotic community while the α -diversity of the fungal community increased over time only under zaxinone and MiZax5 treatments. Notwithstanding the compartment and developmental stage, zaxinone and MiZax treatments significantly influenced the β -diversity of prokaryotic microbial communities. Notably, we highlighted a shared pattern between MiZax3 and MiZax5 more similar to the acetone control, whereas a clear separation was evident for samples treated with zaxinone. Studies focused on plant traits demonstrated that MiZax3 and MiZax5 exhibit zaxinone-like activity by rescuing the root growth of a zaxinone-deficient rice mutant. They also promote overall growth and decrease SLs content in both roots and root exudates of wild-type plants . However, contrasting results were observed when mycorrhization was considered. Specifically, the application of 5 µM MiZax did not negatively impact AM fungal root colonization, whereas zaxinone treatment markedly reduced AM mycorrhization , . These findings suggested that zaxinone and MiZax compounds can be interchanged when plant traits are concerned but more caution is needed considering plant-microbe interactions. Our metabarcoding data are consistent with these observations. As shown by the analysis of compartment-specific taxa, the different pattern between MiZax and zaxinone treatment seems to be more evident at T1. Herein, a stronger polarization between root and rhizosphere prokaryotic communities occurs in MiZax3 and MiZax5 compared to the acetone control. Conversely, in the zaxinone treatment, a higher overlap between these two compartments is detected at both developmental stages. In particular, in the endosphere at T1, we identified an increased number of ASVs which are key soil-beneficial bacterial taxa, such as Comamonadaceae, Acidibacter , and Devosia , enriched under MiZax3 and MiZax5 treatments. Acidibacter and Comamonadaceae have normally high phosphorus solubilizing activity , and in addition, Comamonadaceae has also the ability to solubilize potassium, zinc, and nitrogen and its abundance has been related to disease-suppressive soil , . Devosia has been reported as nitrogen-fixing bacteria which also alleviate abiotic stress and recently has been reported to be enriched on AM extraradical hyphae and mycorrhizal root . Furthermore, two novel Devosia species recently isolated from the rice rhizosphere have been demonstrated to produce IAA and siderophores . Further, in the rhizosphere MiZax3 and MiZax5 shared the accumulation of keystone rhizobacteria taxa such as Pedosphaeraceae , and Nitrosarchaeum which is predicted to be an anaerobic hydrocarbon-degrading bacteria in the subsurface soil . By contrast, at T2 in both zaxinone and MiZax3, rhizosphere and endosphere prokaryotic communities were more similar, while MiZax5 still maintained the pattern shown at T1. By contrast, the impact of the treatments on soil communities was detectable but negligible in term of the amount of regulated taxa. It is plausible to hypothesize that the soil environment more effectively buffers exogenous environmental changes, potentially due to its higher microbial abundance compared to root tissues. As an alternative, we can hypothesize that the applied molecules only exert a limited direct effect on the microbial communities, such an outcome being magnified by the plant. Zaxinone and its mimics are in fact perceived by the plant, which might undertake modification of its hormonal balance and/or rhizodeposition that could result in a differential recruitment of endosphere/rhizosphere microbiota. Coupled with the evidence that the application of MiZax3 and MiZax5 differentially modulated the abundances of bacterial ASVs both in the endosphere and in the rhizosphere, this data suggests that each molecule has a specific impact on shaping the bacterial community. Compared with the acetone control, at T1, all the molecules exerted the most significant impact on the root endosphere bacterial and fungal communities with a widespread depletion of taxa, while an opposite pattern was observed in the rhizosphere where the compounds supply incremented the number of enriched taxa, at least for prokaryotic communities. At T2 a higher ratio of depleted prokaryotic taxa was observed in the endosphere following treatment with zaxinone and MiZax3 while MiZax5 has a milder effect, with enriched taxa prevailing on the depleted ones. The endosphere mycobiota was slightly less impacted by treatments compared to T1. In the rhizosphere and in the unplanted soil, all the treatments displayed a limited impact in terms of enriched/depleted taxa, irrespective of the timepoint. These findings suggest that zaxinone and MiZax treatments exert the highest impact on microbial communities in the root endosphere, suggesting once again that these dynamics are likely influenced by plant-mediated processes. In the analysis of the taxonomic diversity of differentially abundant taxa in the root endosphere, it was observed that depleted taxa were widely distributed across the bacterial phylogeny. Notably, Actinobacteriota was an exception, mostly showing an increase in their abundance in all treatments at T1. This phylum is recognized as the producer of many bioactive metabolites in agriculture, such as insecticides, herbicides, fungicides, and growth-promoting substances for plants . By contrast, the abundance of Acidobacteria decreased drastically, overall. Interestingly, a lower abundance of Acidobacteria was also found in rice d17 mutant lines defective for SLs biosynthesis , suggesting that the decreased Acidobacteria abundance in the endosphere of zaxinone and MiZax treated plants could be dependent on the negative impact of these compounds on SLs biosynthesis. At the same time, we observed an enrichment of Chitinophagaceae upon MiZax3 treatment. Since Chitinophagaceae was negatively associated with orobanchol , their higher abundance could be related to the negative impact of MiZax on SLs biosynthesis. Regarding the fungal community, we observed a depletion of Mucoromycota phylum. It is worth noting that different Funneliformis ASVs belonging to Glomeromycotina subphylum were depleted by zaxinone treatment at T1 while enriched at T2. By contrast, MiZax3 and MiZax5 have a mild impact at both time points. These findings are in line with our previous reports which displayed at early time points a strong reduction of AM symbiosis colonization in rice root treated with zaxinone while no negative effects were reported upon MiZax treatment , . The decrease of these taxa at a very early sampling time may be attributed to the lower SL content in root exudates induced by zaxinone treatment . The SLs reduction potentially delays the mycorrhization process, which however was fully recovered over time. The analysis of prokaryotic community interactions in the endosphere, revealed in zaxinone and MiZax-treated rice roots at T1 a decrease in network complexity and in a number of keystone taxa compared to the control, with the most pronounced effects for zaxinone and MiZax5 treatments, although few but significant plant-beneficial microbes emerged as hubs-taxa. However, at T1 all the network metrics indicated for most of the treatments a decentralization of microbe-microbe interactions compared to the control, with few but wider connections spanning through the network and possibly resulting in a community less dependent on a few hub species. These features suggest that at T1 the community may be more resilient to external perturbations, but further experiments are needed to confirm these speculations. At T2, in all treated plants and in particular, upon MiZax5, significant interactions between taxa were re-established in a similar way to the acetone control, with a shift in microbes potentially holding plant-beneficial traits among the identified keystone taxa. Overall, our results indicate that all the treatments, to different extents, induce significant changes in root-associated rice microbial communities particularly at T1, with a stronger effect on prokaryotes. Notably, our data indicate that these changes modify the community network structure involving keystone taxa. However, at T2 the microbial communities dynamics were re-established suggesting a temporary delay in the microbial interaction established by treated plants compared to the control ones. Despite these changes, under most of the tested conditions, the hub taxa still include genera with well-acknowledged plant-beneficial traits such as Sphingomonas , Streptomyces , and Haliangium . This last species has been already recognized as a hub taxon in rice paddies, being very sensitive to external solicitations such as abiotic stresses (negative correlation) or biofertilization (positive correlation) , and to be enriched in the rice root endosphere under the disturbance of barnyard grass ( Echinochloa crus-galli ). Shoot metabolome and its correlation with the rhizomicrobiome In the current study, we found that the soil treatment with zaxinone and its synthetic mimics systematically promote different metabolic pathways in the shoot of rice plants. Notwithstanding the treatment with zaxinone and MiZax exert comparable activity in the plant , the impact on shoot metabolites and grain nutrient content display some differences that may be also attributed to the different rhizomicrobiota composition. It was determined in different host plants that different rhizomicrobiota communities play a significant role in metabolic profiles and mineral nutrient uptake at local and systemic level – . The main metabolic differences in sugar, organic acid, and amino acid content have been detected at T1 upon MiZax3 treatment, while at T2 few differences have been observed between all treatments. Wang et al. demonstrated that the growth-promoting effect of zaxinone is strongly linked with an increase in sugar metabolism in rice shoot at 24 h after treatment. In our conditions, differences in sugar content were not detected in the shoot of rice plants treated with zaxinone. In contrast, MiZax3 at T1 exhibited an increase in myoinositol and threitol content. Concerning the organic acids content, we observed at T1 an accumulation of oxoglutarate, pyruvate, dehydroascorbate, and glycerate in the shoot of MiZax3 treated plants, the content of some of them (oxoglutarate and glycerate) increased also upon MiZax5 treatment and as a general trend we observed an accumulation of these metabolites in shoots of plants treated with zaxinone. Interestingly, Wang et al. previously demonstrated the early accumulation of 2-oxoglutaric acid and glyceric acid in shoot as a response to zaxinone treatment, suggesting that also MiZax compounds play a role in organic acid metabolism. However, we measured a lower content of citrate and malate. Since both organic acids contribute to the acquisition of phosphorus in soils , the reduced accumulation of these acids may be associated with the lower content of phosphoric acid observed in the shoots of all treated plants. Concerning amino acid content, we identified an increment of GABA, alanine, and threonine content in both MiZax at T1. It is worth noting that upon these treatments at T1, we observed an increase of nitrogen-fixing bacteria such as Comamonadaceae and Devosia . Given their known ability to enhance plant nitrogen uptake and to produce amino acids, these bacteria may be accountable for the enhanced amino acid content in these plants . In line with this hypothesis, Rahmoune and colleagues (2019) demonstrated that the inoculation of plant growth-promoting rhizobacteria (PGPR) in Datura stramonium affected significantly the amino acid content in both organs (root and shoot), highlighting a consistent increment of alanine in the shoot of inoculated plants. Impact of zaxinone and its mimics on grain nutrient content Since it has been reported that zaxinone and MiZax also promoted the growth and yield of horticultural crops under open-field conditions – , we, therefore, investigated their impact on grain nutritional content. Regarding the mineral nutrients profile, we observed that zaxinone increased Zinc (Zn) and Cu content, while Mn was reduced under both MiZax treatments. Notably, elevated Zn concentration has been reported in rice grains of plants inoculated with β and ɣ-proteobacteria (i.e. Sphingomonas sp., Burkholderia cepacia , Pantoea rodasii , and Enterobacter sp.) indicating a key role of rice-associated microbes in mineral grain content , . Zinc deficiency is a major constraint to rice production and Zn is also often deficient in humans with rice-based diets . The increase in Zn in the seeds of plants treated with zaxinone validates the use of this metabolite not only for improving rice growth but also for enhancing the nutritional aspect of the grain. Concerning Mn and Cu, both are considered essential elements for the growth and development of plants participating in many metabolic processes, including oxidation-reduction (redox), and photosynthesis . The lower amount of Mn in grains of MiZax treated plants could be related to the increase in the level of chlorophyll and in the enhancement of photosynthetic activities reported in rice that could affect Mn homeostasis and translocation from shoot to the grains. Rice fields represent a peculiar environment for the microbial soil communities. Due to the periodic flooding, this habitat is characterized by oxygen-limited conditions that shape the microbiota assembly. Bacterial communities typically include both aerobic and anaerobic taxa . The overall microbiota assembly from unplanted soil revealed a prokaryotic composition typical of paddy environments, including Gemmatimonadetes, Chloroflexi, Acidobacteria and Actinobacteria, and the archaeal phylum Crenarchaeota – . On the fungal side, our analysis revealed a relevant proportion of Leotiomycetes in all the conditions considered. This group of fungi includes good organic matter decomposers that can tolerate high levels of heavy metal contaminants usually found in paddy soils , and can associate with plant roots living as endophytes. So far, the effects of exogenous treatment with zaxinone and its mimics (MiZax3 and MiZax5) were investigated considering the growth promotion activity, the regulation of SLs biosynthesis, and the AM symbiosis , . Here, we provide comprehensively new information about the impact of these compounds on paddy soil and rice root-associated microbes and how these effects systemically influence rice metabolomic and grain nutrient profiles considering different developmental stages. The metabarcoding analysis highlighted that rice microbiome assembly is mainly influenced by compartment niche and developmental stage as already reported for rice and other plants – , regardless of zaxinone and MiZax treatments. In contrast to the findings reported by Zangh and colleagues , which observed a reduction in root microbial community richness in the later stage of the rice life cycle, our data reveals a general increase in α -diversity at T2 compared to T1 across the three considered compartments (unplanted soil, rhizosphere, and endosphere), regardless of the treatments. This increment was evident in the endosphere particularly for the prokaryotic community while the α -diversity of the fungal community increased over time only under zaxinone and MiZax5 treatments. Notwithstanding the compartment and developmental stage, zaxinone and MiZax treatments significantly influenced the β -diversity of prokaryotic microbial communities. Notably, we highlighted a shared pattern between MiZax3 and MiZax5 more similar to the acetone control, whereas a clear separation was evident for samples treated with zaxinone. Studies focused on plant traits demonstrated that MiZax3 and MiZax5 exhibit zaxinone-like activity by rescuing the root growth of a zaxinone-deficient rice mutant. They also promote overall growth and decrease SLs content in both roots and root exudates of wild-type plants . However, contrasting results were observed when mycorrhization was considered. Specifically, the application of 5 µM MiZax did not negatively impact AM fungal root colonization, whereas zaxinone treatment markedly reduced AM mycorrhization , . These findings suggested that zaxinone and MiZax compounds can be interchanged when plant traits are concerned but more caution is needed considering plant-microbe interactions. Our metabarcoding data are consistent with these observations. As shown by the analysis of compartment-specific taxa, the different pattern between MiZax and zaxinone treatment seems to be more evident at T1. Herein, a stronger polarization between root and rhizosphere prokaryotic communities occurs in MiZax3 and MiZax5 compared to the acetone control. Conversely, in the zaxinone treatment, a higher overlap between these two compartments is detected at both developmental stages. In particular, in the endosphere at T1, we identified an increased number of ASVs which are key soil-beneficial bacterial taxa, such as Comamonadaceae, Acidibacter , and Devosia , enriched under MiZax3 and MiZax5 treatments. Acidibacter and Comamonadaceae have normally high phosphorus solubilizing activity , and in addition, Comamonadaceae has also the ability to solubilize potassium, zinc, and nitrogen and its abundance has been related to disease-suppressive soil , . Devosia has been reported as nitrogen-fixing bacteria which also alleviate abiotic stress and recently has been reported to be enriched on AM extraradical hyphae and mycorrhizal root . Furthermore, two novel Devosia species recently isolated from the rice rhizosphere have been demonstrated to produce IAA and siderophores . Further, in the rhizosphere MiZax3 and MiZax5 shared the accumulation of keystone rhizobacteria taxa such as Pedosphaeraceae , and Nitrosarchaeum which is predicted to be an anaerobic hydrocarbon-degrading bacteria in the subsurface soil . By contrast, at T2 in both zaxinone and MiZax3, rhizosphere and endosphere prokaryotic communities were more similar, while MiZax5 still maintained the pattern shown at T1. By contrast, the impact of the treatments on soil communities was detectable but negligible in term of the amount of regulated taxa. It is plausible to hypothesize that the soil environment more effectively buffers exogenous environmental changes, potentially due to its higher microbial abundance compared to root tissues. As an alternative, we can hypothesize that the applied molecules only exert a limited direct effect on the microbial communities, such an outcome being magnified by the plant. Zaxinone and its mimics are in fact perceived by the plant, which might undertake modification of its hormonal balance and/or rhizodeposition that could result in a differential recruitment of endosphere/rhizosphere microbiota. Coupled with the evidence that the application of MiZax3 and MiZax5 differentially modulated the abundances of bacterial ASVs both in the endosphere and in the rhizosphere, this data suggests that each molecule has a specific impact on shaping the bacterial community. Compared with the acetone control, at T1, all the molecules exerted the most significant impact on the root endosphere bacterial and fungal communities with a widespread depletion of taxa, while an opposite pattern was observed in the rhizosphere where the compounds supply incremented the number of enriched taxa, at least for prokaryotic communities. At T2 a higher ratio of depleted prokaryotic taxa was observed in the endosphere following treatment with zaxinone and MiZax3 while MiZax5 has a milder effect, with enriched taxa prevailing on the depleted ones. The endosphere mycobiota was slightly less impacted by treatments compared to T1. In the rhizosphere and in the unplanted soil, all the treatments displayed a limited impact in terms of enriched/depleted taxa, irrespective of the timepoint. These findings suggest that zaxinone and MiZax treatments exert the highest impact on microbial communities in the root endosphere, suggesting once again that these dynamics are likely influenced by plant-mediated processes. In the analysis of the taxonomic diversity of differentially abundant taxa in the root endosphere, it was observed that depleted taxa were widely distributed across the bacterial phylogeny. Notably, Actinobacteriota was an exception, mostly showing an increase in their abundance in all treatments at T1. This phylum is recognized as the producer of many bioactive metabolites in agriculture, such as insecticides, herbicides, fungicides, and growth-promoting substances for plants . By contrast, the abundance of Acidobacteria decreased drastically, overall. Interestingly, a lower abundance of Acidobacteria was also found in rice d17 mutant lines defective for SLs biosynthesis , suggesting that the decreased Acidobacteria abundance in the endosphere of zaxinone and MiZax treated plants could be dependent on the negative impact of these compounds on SLs biosynthesis. At the same time, we observed an enrichment of Chitinophagaceae upon MiZax3 treatment. Since Chitinophagaceae was negatively associated with orobanchol , their higher abundance could be related to the negative impact of MiZax on SLs biosynthesis. Regarding the fungal community, we observed a depletion of Mucoromycota phylum. It is worth noting that different Funneliformis ASVs belonging to Glomeromycotina subphylum were depleted by zaxinone treatment at T1 while enriched at T2. By contrast, MiZax3 and MiZax5 have a mild impact at both time points. These findings are in line with our previous reports which displayed at early time points a strong reduction of AM symbiosis colonization in rice root treated with zaxinone while no negative effects were reported upon MiZax treatment , . The decrease of these taxa at a very early sampling time may be attributed to the lower SL content in root exudates induced by zaxinone treatment . The SLs reduction potentially delays the mycorrhization process, which however was fully recovered over time. The analysis of prokaryotic community interactions in the endosphere, revealed in zaxinone and MiZax-treated rice roots at T1 a decrease in network complexity and in a number of keystone taxa compared to the control, with the most pronounced effects for zaxinone and MiZax5 treatments, although few but significant plant-beneficial microbes emerged as hubs-taxa. However, at T1 all the network metrics indicated for most of the treatments a decentralization of microbe-microbe interactions compared to the control, with few but wider connections spanning through the network and possibly resulting in a community less dependent on a few hub species. These features suggest that at T1 the community may be more resilient to external perturbations, but further experiments are needed to confirm these speculations. At T2, in all treated plants and in particular, upon MiZax5, significant interactions between taxa were re-established in a similar way to the acetone control, with a shift in microbes potentially holding plant-beneficial traits among the identified keystone taxa. Overall, our results indicate that all the treatments, to different extents, induce significant changes in root-associated rice microbial communities particularly at T1, with a stronger effect on prokaryotes. Notably, our data indicate that these changes modify the community network structure involving keystone taxa. However, at T2 the microbial communities dynamics were re-established suggesting a temporary delay in the microbial interaction established by treated plants compared to the control ones. Despite these changes, under most of the tested conditions, the hub taxa still include genera with well-acknowledged plant-beneficial traits such as Sphingomonas , Streptomyces , and Haliangium . This last species has been already recognized as a hub taxon in rice paddies, being very sensitive to external solicitations such as abiotic stresses (negative correlation) or biofertilization (positive correlation) , and to be enriched in the rice root endosphere under the disturbance of barnyard grass ( Echinochloa crus-galli ). In the current study, we found that the soil treatment with zaxinone and its synthetic mimics systematically promote different metabolic pathways in the shoot of rice plants. Notwithstanding the treatment with zaxinone and MiZax exert comparable activity in the plant , the impact on shoot metabolites and grain nutrient content display some differences that may be also attributed to the different rhizomicrobiota composition. It was determined in different host plants that different rhizomicrobiota communities play a significant role in metabolic profiles and mineral nutrient uptake at local and systemic level – . The main metabolic differences in sugar, organic acid, and amino acid content have been detected at T1 upon MiZax3 treatment, while at T2 few differences have been observed between all treatments. Wang et al. demonstrated that the growth-promoting effect of zaxinone is strongly linked with an increase in sugar metabolism in rice shoot at 24 h after treatment. In our conditions, differences in sugar content were not detected in the shoot of rice plants treated with zaxinone. In contrast, MiZax3 at T1 exhibited an increase in myoinositol and threitol content. Concerning the organic acids content, we observed at T1 an accumulation of oxoglutarate, pyruvate, dehydroascorbate, and glycerate in the shoot of MiZax3 treated plants, the content of some of them (oxoglutarate and glycerate) increased also upon MiZax5 treatment and as a general trend we observed an accumulation of these metabolites in shoots of plants treated with zaxinone. Interestingly, Wang et al. previously demonstrated the early accumulation of 2-oxoglutaric acid and glyceric acid in shoot as a response to zaxinone treatment, suggesting that also MiZax compounds play a role in organic acid metabolism. However, we measured a lower content of citrate and malate. Since both organic acids contribute to the acquisition of phosphorus in soils , the reduced accumulation of these acids may be associated with the lower content of phosphoric acid observed in the shoots of all treated plants. Concerning amino acid content, we identified an increment of GABA, alanine, and threonine content in both MiZax at T1. It is worth noting that upon these treatments at T1, we observed an increase of nitrogen-fixing bacteria such as Comamonadaceae and Devosia . Given their known ability to enhance plant nitrogen uptake and to produce amino acids, these bacteria may be accountable for the enhanced amino acid content in these plants . In line with this hypothesis, Rahmoune and colleagues (2019) demonstrated that the inoculation of plant growth-promoting rhizobacteria (PGPR) in Datura stramonium affected significantly the amino acid content in both organs (root and shoot), highlighting a consistent increment of alanine in the shoot of inoculated plants. Since it has been reported that zaxinone and MiZax also promoted the growth and yield of horticultural crops under open-field conditions – , we, therefore, investigated their impact on grain nutritional content. Regarding the mineral nutrients profile, we observed that zaxinone increased Zinc (Zn) and Cu content, while Mn was reduced under both MiZax treatments. Notably, elevated Zn concentration has been reported in rice grains of plants inoculated with β and ɣ-proteobacteria (i.e. Sphingomonas sp., Burkholderia cepacia , Pantoea rodasii , and Enterobacter sp.) indicating a key role of rice-associated microbes in mineral grain content , . Zinc deficiency is a major constraint to rice production and Zn is also often deficient in humans with rice-based diets . The increase in Zn in the seeds of plants treated with zaxinone validates the use of this metabolite not only for improving rice growth but also for enhancing the nutritional aspect of the grain. Concerning Mn and Cu, both are considered essential elements for the growth and development of plants participating in many metabolic processes, including oxidation-reduction (redox), and photosynthesis . The lower amount of Mn in grains of MiZax treated plants could be related to the increase in the level of chlorophyll and in the enhancement of photosynthetic activities reported in rice that could affect Mn homeostasis and translocation from shoot to the grains. Taken together, soil application of zaxinone and MiZax exerted a temporary strong effect on the endosphere microbial communities, particularly prokaryotes, while a minor impact on fungal communities was observed. Moreover, if the T1 showed a general depletion of the prokaryotes communities and a reduction of hub-taxa at T2 all treated plants, especially those treated with MiZax5, re-established significant interactions between taxa at a level comparable to the acetone control. This recovery can be due to the difference in the plant phenological stage or to enhanced resilience of the rhizomicrobiota community to tolerate exogenous treatments over time. Among the communities impacted by treatments, a number of taxa potentially holding plant-beneficial traits were observed. However, since plant-growth promoting capacities cannot be confirmed by our metabarcoding approach, this evidence needs further validation by isolating strains or by in vitro tests. Interestingly, the differences in microbial communities assembly highlighted at T1 among treatments and between the control are also partially mirrored in the metabolites profiles which displayed changes in sugar, organic acid, and amino acid content depending on the condition considered, while few differences between shoot of treated and control plants are observed at T2. Furthermore, the grain biochemical characterization revealed that these compounds are not only beneficial for increasing plant biomass and yield – , but also hold promise for enhancing the accumulation of zinc content in rice seeds. With the adopted experimental set-up it was not possible to disentangle the effects of the treatments themselves on plant-associated communities from the changes arising from the different SLs exudation pattern exerted by MiZax application – . However, even to a much lower extent, treatments also impacted the soil prokaryotic community in absence of the plant suggesting a role for these molecules in shaping plant-associated microbial assemblages in a direct manner. This evidence opens new research questions, such as the understanding of contribution of the plant-mediated signaling on the changes observed here and whether these occur in a similar way across different crop plant models. Overall, our results reinforce the practical use of zaxinone and MiZax application in the field as ecologically friendly biostimulants to enhance crop productivity without causing permanent disruption to the native rice root-associated microbiota, thereby paving the way for new strategies towards sustainable agriculture worldwide. Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 |
Interventions Addressing Health Literacy in Cancer Care: A Systematic Review of Reviews | 6ed597e4-0b60-4671-b46a-58d38edaf8e0 | 11855911 | Health Literacy[mh] | Over the last decades, health literacy (HL) has gained critical importance in public health and healthcare . Although several definitions of the concept exist , HL is generally agreed to refer to people’s knowledge, motivation, and competencies to access, understand, appraise, and apply health information to decision-making in healthcare, disease prevention, and health promotion . This specific competence has become increasingly relevant in the context of patient-centered care, where the patients and their informal carers are actively involved in the decision-making regarding their health in different domains, such as disease prevention and treatment. Informed decision-making not only requires that information is communicated by health professionals or by the health system, but also that it is adapted to each patient’s level of comprehension of medical jargon, and their aptitude for navigating health services . Patients with limited HL have more difficulty navigating through the system and making decisions, especially in high-burden diseases such as cancer . In addition, low HL has been associated with less participation in preventive cancer screening and with more risky and problematic behavior, resulting in poor treatment adherence and poor illness self-management . As such, adequate levels of HL are essential when dealing with and managing chronic illnesses as complex as cancer . Cancer care offers a range of treatment options, varying in duration and complexity, with some being short-term, others long-term, and often requiring a combination of approaches to address a single issue effectively. These treatments can be surgeries, preventative screenings, chemotherapy, radiotherapy, and others, which is why cancer care necessitates a certain type of HL level and care coordination . While emphasizing its role in informed health decision-making in everyday life, the above-mentioned definition of HL by Sorensen et al. recognizes its multidimensional character and its applicability within the healthcare, disease prevention and health promotion setting. This has also been highlighted by other authors. For instance, Nutbeam distinguishes between functional, communicative and critical HL, while Stocks et al. shift the focus from understanding health information in healthcare to motivating health-related actions and Wu et al. (2010) consider the empowerment of health literate individuals in controlling their health behaviors and living conditions as key. Likewise, Dodson et al. consider HL to include a broad range of competencies enabling sound health decisions and proactive engagement with factors that impact health. Freedman et al. stress the communal dimension, defining HL as the ability of not only individuals, but also of groups, to use information for public health decisions. HL is indeed often considered in relationship with health inequities, in the sense that acts as a social determinant of health , or as a mediator between social and economic determinants and specific health outcomes, health-related behaviors, and access to health services . As a set of competencies linked to general literacy, HL can be conceived of as the product of health education . In that regard, Berkman and colleagues focus on the educational purposes of HL rather than on specific skills such as analysis, filtration of information, application, etc. Not surprisingly, the growing awareness of the role of HL for healthcare, health behaviors, health actions, and health inequalities has been accompanied by efforts to improve HL levels in populations. This has resulted in a large number of interventional studies and strategies targeting limited HL through patient education, health education, and health promotion. Specifically, operationalizing the HL concept in interventions is meant to target a set of functional skills and more complex competencies related to health behaviors, such as self-management, problem-solving skills, decision-making, application, and others . A review of intervention studies by Berkman and colleagues revealed that there is a large variety of intervention types, ranging from single-features, such as one-time information sessions, flyers, booklets, and/or other tools meant to be used by patients for educational purposes, to more encompassing intervention programs targeting self-management, self-efficacy, adherence, and skill building, aimed at behavior change. In the past years, several reviews have been performed to synthesize the findings from the intervention studies on the quality, outcomes, feasibility, and efficacy of interventions addressing HL . However, given the variety of intervention types and outcome indicators that were involved in these reviews, the results could not easily be compared, nor could the conclusions be generalized. Therefore, this review of reviews is set to explore the conclusions drawn by the included reviews that have addressed interventions with a HL factor by going over the included studies, the analyses, and the role HL played in the theory or application of the interventions. Exploring the above could allow for a better understanding of the position of HL in what is considered a “health literacy intervention”. The current review addresses the existing reviews in the form of a meta-review, focusing specifically on the dimensions of HL when considering the results of the interventions concerned. This focus on HL allows for a better understanding of the usage of the term in itself, its role in interventions, and its influence on the results. Specifically, the review will synthesize the existing systematic reviews that study cancer-specific interventional studies including HL, bringing the HL aspect forward through compiling the information from different sources, looking into (1) the features of the different interventions included in existing review studies, (2) the outcomes of existing systematic reviews, (3) the conclusions drawn from the interventional studies in terms of the importance, efficacy, and impact of/on HL, as presented in the existing reviews, and (4) the different aspects of HL brought out in the interventions through the analysis of the reviewers. This study is a review of systematic reviews looking into interventions targeting HL as a mediating variable or as an outcome among patients with cancer. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) declaration and the methodological guidelines for utilizing existing systematic reviews were followed in the conduct of the study . The review protocol was submitted to Prospero in February 2022. 2.1. Search Strategy A literature search was performed using Embase, Pubmed, PsycINFO, and Science Direct. The search string used in performing the literature search focused on 5 themes: health literacy, oncology, interventions, population (adult patients and health professionals), and the nature of the article (review) (can be accessed in ). Due to the recent developments and peaked interest in HL, articles published before 2010 were not considered. Specifically, HL started being officially recognized around 2009, after a report by WHO “Health Promotion and Health Literacy” , in addition to the US department of Health and Human Services launching the “National Action Plan to Improve Health Literacy” . The literature search included articles in French, English, and Spanish. The first literature search wave was conducted in February 2022, with updates in March 2023 and January 2024 in order to include any newly released studies during the review process. The articles that were identified were imported into CADIMA. 2.2. Study Selection Reviews that studied interventions addressing health literacy within cancer care (targeting organizations, professionals, and/or patients) were included. The inclusion criteria were (1) the article should include a cancer-related population, (2) the review should include interventions targeting adult patients and/or health professionals, (3) health literacy or health competence should be explicitly mentioned. Studies that included health literacy as a targeted outcome, a determinant of intervention effectiveness, a moderator, a mediating variable, used HL-specific tools, or factored in the influence on health-related outcomes were included. Studies could have addressed health literacy as a whole or focused on specific aspects (e.g., comprehension, information use, decision-making, or healthcare-system navigation). First, the titles and abstracts of reviews found through the literature search were screened by two evaluators (CJ and CL), with a third evaluator (SV) solicited in the case of disagreements. The interrater validity between the two evaluators was assured through the repeated testing of 3 random studies and discussions on clarifying the criteria until the evaluators reached full agreement. In the second step, full-text screening was conducted for the reviews that passed the inclusion criteria based on their titles and abstracts. The full texts were also screened by the two evaluators, with a third evaluator involved in the case of opposing results. Comments were also added on the excluded studies, in order to compare notes in cases of potential disagreements. While adhering to the inclusion criteria and screening for intervention studies addressing oncology populations and health literacy, three of the included reviews examined multiple diseases, with cancer being one among them. Consistent with our criteria, these reviews were included, and the data, along with conclusions specific to cancer-related studies, were extracted for the analysis. As a result of the screening process, 10 review studies (out of 148 screened) met the criteria and were considered eligible for data extraction. shows the PRISMA flow diagram representing the flow of information through the different phases of a systematic review. 2.3. Data Extraction Data extraction was performed by two individual reviewers (CJ and CL), using a data extraction form created to fit the characteristics evaluated in this review. Four separate sheets were created, each concerning a different aspect of each systematic review: (1) general information about the studies (type of studies, populations, criteria, countries, etc.); (2) quality data; (3) the content of the interventions included in each review (type of intervention, frequency, tools, persons involved, etc.); and (4) outcomes. 2.4. Quality Assessment and Risk of Bias A quality assessment of the 10 review articles included in the paper was performed using a systematic review and a meta-analysis assessment tool from the National Institute of Health (NIH), specifically the National Heart, Lung, and Blood Institute’s (NHLBI). The criteria of this tool assess the review’s objectives, including a well-formulated question and pre-defined inclusion and exclusion criteria. It scrutinizes the process of the literature search, the screening process, and the thoroughness of the evaluation of each study’s quality. Additionally, the criteria consider transparent reporting of included studies, potential publication bias, and heterogeneity in meta-analyses, when applicable. Each of the 8 criteria was assessed by two reviewers (CJ and CL). Similar to the screening process, a third reviewer (SV) was involved in case of inconsistencies. For each criterium, a score was given corresponding with the categories Good, Fair, Poor, Not Applicable, Cannot Determine, and Not Reported. The majority of the studies scored ‘good’ on the first three criteria and on eligibility criteria: clear questions, screening process, and study presentation. The rest varied. A table with details regarding the quality scores of each article is given in . The studies’ overall quality can be characterized as ranging from fair to good. For a deeper analysis, the risk of bias was assessed through the ROBIS tool . The reviews scored between moderate- and low-risk. No study scored a high risk of bias. All study objectives, included articles, and methodologies were aligned. The ROBIS table can be found in the . Most studies used databases in the search strategy. An overlap assessment was performed . Reviews Hill et al. (2) and Mustermann et al. (9) had the most overlap, with 12 articles overlapping between each other. This can be explained through the specificity of the population of both articles (deaf patients with cancer). There was a difference in the analysis of results, as Hill et al. focused more on barriers and health disparities in the interventions whilst Mustermann et al. focused more on the interventions and their outcomes. Two meta-analyses are included in the reviews (4,10) . A literature search was performed using Embase, Pubmed, PsycINFO, and Science Direct. The search string used in performing the literature search focused on 5 themes: health literacy, oncology, interventions, population (adult patients and health professionals), and the nature of the article (review) (can be accessed in ). Due to the recent developments and peaked interest in HL, articles published before 2010 were not considered. Specifically, HL started being officially recognized around 2009, after a report by WHO “Health Promotion and Health Literacy” , in addition to the US department of Health and Human Services launching the “National Action Plan to Improve Health Literacy” . The literature search included articles in French, English, and Spanish. The first literature search wave was conducted in February 2022, with updates in March 2023 and January 2024 in order to include any newly released studies during the review process. The articles that were identified were imported into CADIMA. Reviews that studied interventions addressing health literacy within cancer care (targeting organizations, professionals, and/or patients) were included. The inclusion criteria were (1) the article should include a cancer-related population, (2) the review should include interventions targeting adult patients and/or health professionals, (3) health literacy or health competence should be explicitly mentioned. Studies that included health literacy as a targeted outcome, a determinant of intervention effectiveness, a moderator, a mediating variable, used HL-specific tools, or factored in the influence on health-related outcomes were included. Studies could have addressed health literacy as a whole or focused on specific aspects (e.g., comprehension, information use, decision-making, or healthcare-system navigation). First, the titles and abstracts of reviews found through the literature search were screened by two evaluators (CJ and CL), with a third evaluator (SV) solicited in the case of disagreements. The interrater validity between the two evaluators was assured through the repeated testing of 3 random studies and discussions on clarifying the criteria until the evaluators reached full agreement. In the second step, full-text screening was conducted for the reviews that passed the inclusion criteria based on their titles and abstracts. The full texts were also screened by the two evaluators, with a third evaluator involved in the case of opposing results. Comments were also added on the excluded studies, in order to compare notes in cases of potential disagreements. While adhering to the inclusion criteria and screening for intervention studies addressing oncology populations and health literacy, three of the included reviews examined multiple diseases, with cancer being one among them. Consistent with our criteria, these reviews were included, and the data, along with conclusions specific to cancer-related studies, were extracted for the analysis. As a result of the screening process, 10 review studies (out of 148 screened) met the criteria and were considered eligible for data extraction. shows the PRISMA flow diagram representing the flow of information through the different phases of a systematic review. Data extraction was performed by two individual reviewers (CJ and CL), using a data extraction form created to fit the characteristics evaluated in this review. Four separate sheets were created, each concerning a different aspect of each systematic review: (1) general information about the studies (type of studies, populations, criteria, countries, etc.); (2) quality data; (3) the content of the interventions included in each review (type of intervention, frequency, tools, persons involved, etc.); and (4) outcomes. A quality assessment of the 10 review articles included in the paper was performed using a systematic review and a meta-analysis assessment tool from the National Institute of Health (NIH), specifically the National Heart, Lung, and Blood Institute’s (NHLBI). The criteria of this tool assess the review’s objectives, including a well-formulated question and pre-defined inclusion and exclusion criteria. It scrutinizes the process of the literature search, the screening process, and the thoroughness of the evaluation of each study’s quality. Additionally, the criteria consider transparent reporting of included studies, potential publication bias, and heterogeneity in meta-analyses, when applicable. Each of the 8 criteria was assessed by two reviewers (CJ and CL). Similar to the screening process, a third reviewer (SV) was involved in case of inconsistencies. For each criterium, a score was given corresponding with the categories Good, Fair, Poor, Not Applicable, Cannot Determine, and Not Reported. The majority of the studies scored ‘good’ on the first three criteria and on eligibility criteria: clear questions, screening process, and study presentation. The rest varied. A table with details regarding the quality scores of each article is given in . The studies’ overall quality can be characterized as ranging from fair to good. For a deeper analysis, the risk of bias was assessed through the ROBIS tool . The reviews scored between moderate- and low-risk. No study scored a high risk of bias. All study objectives, included articles, and methodologies were aligned. The ROBIS table can be found in the . Most studies used databases in the search strategy. An overlap assessment was performed . Reviews Hill et al. (2) and Mustermann et al. (9) had the most overlap, with 12 articles overlapping between each other. This can be explained through the specificity of the population of both articles (deaf patients with cancer). There was a difference in the analysis of results, as Hill et al. focused more on barriers and health disparities in the interventions whilst Mustermann et al. focused more on the interventions and their outcomes. Two meta-analyses are included in the reviews (4,10) . shows a summary of the basic characteristics of each of the 10 review articles that were included. 3.1. Study Types The review papers included in this review included between ten (7) and fifty-three (4) primary studies. The total number of participants was not always clear. They included interventional studies, surveys, and simple comparative studies focusing on HL in the context of oncology. However, they all included at least one interventional study that targeted at least one of the aspects of HL (e.g., informed decision-making). The intervention studies varied from testing the efficacy of online interventions to analyzing the importance of promoting certain aspects such as decision-making, competence, and patient education. 3.2. Types of Cancer Although none of the reviews in this study targeted breast cancer specifically, breast cancer was the most often studied type of cancer, and was represented in all ten the reviews. The reviews by DeRosa et al. (3) and by Fernandez-Gonzalez and Bravo-Valenzuela (7) were concerned with HL in breast and prostate cancer, while the ones by McAlpine et al. (1) , Housten et al. (5) , and Cabanes et al. (8) had 50%, 40%, and one-third of their study population suffering from breast cancer, respectively. As mentioned above, three of the review studies (4,6,10) included cancer as one of several non-communicable diseases. The review by Heine et al. (4) included only one interventional study that targeted patients with cancer; in the interventional studies included in the review by van der Kruk et al. (6) , half were concerned with cancer; and the review by Verweel et al. (10) contained seventeen studies on chronic illnesses, four of which focused specifically on cancer, and two of which included cancer along with other diseases. In the review by Verweel et al. (10) , six of the studies were concerned with cancer, three of which were specifically focused on breast cancer. Other types of cancer, such as prostate and colorectal cancer, were highly present. 3.3. Participants Most of the reviews focused on HL among individual adult patients, with an average age ranging from 42 to 70 years. However, the review by Housten et al. (5) also included the HL of peers and caregivers. DeRosa et al.’s review (3) also mentioned the importance of including family and peers (social support) in education and decision-aid of patients with cancer, as they play a crucial role in addressing HL. The review by Hill et al. (2) also included interventions involving health professionals, since they fit into their target of building patient-specific culturally competent care. Four of the reviews (2,3,7,9) targeted minority populations, notably deaf patients (2,9) and African, Hispanic, Latin, or Asian populations (3,7) . 3.4. General Description of Interventions Almost all the reviews included studies involving primarily online interventions or online assistance for face-to-face interventions. Five reviews (1,6,7,9,10) were concerned with studies that were completely online, while the others (2,3,4,5,7) considered mixed methods, such as face-to-face interventions (group or individual), interviews, etc. The interventions that did not occur online, such as interviews (2), educational programs (1,2,3,4,6,7,8) , information sessions (2,3,4,7,8) , workshops (3,5,6,8) , and handouts (4,5) , were mostly delivered by the research team, health educators, or medical staff such as nurses, pharmacists, and/or social workers. The review by Verweel et al. (10) focused specifically on digital health literacy, and therefore included solely digital tools, whether internet-based or using other digital means (such as computer software, smart devices, websites, learning management systems, and electronic personal health records (ePHR)), while study 6 had a virtual reality approach. Educational videos were the most commonly used approach to address health literacy, featured in at least one intervention across all ten reviews. The subjects of the educational interventions varied from specific topics, such as fatigue, insomnia, fertility, diet, and smoking cessation, to more general topics, involving cancer-related knowledge, screening information, and general symptoms. Other interventions included coping skills training, communication skills training (for professionals), symptom monitoring and self-management, self-care, treatment adherence, and decision-making. While these programs are clearly linked to HL, the reviews did not provide enough detail about the content to clearly define this connection. Nevertheless, it is clear that education, digital or not, is the main tool to address HL among patients with cancer. 3.5. The Role of Health Literacy 3.5.1. Operationalization of HL The outcomes of the reviews comprised in this review are shown in . All reviews included in this study involved one or more interventions, and, in accordance with the inclusion criteria, had to focus on HL as a theme within the interventions to be included in the study. While all of them analyzed the impact of certain factors related to HL, not all necessarily considered HL itself as a measured outcome. Other outcomes that were measured were decision-making, adherence, knowledge, understanding, communication, self-efficacy and self-management, which are related to HL, or a part of its definition. In some studies, however, the outcomes were less directly related to HL, such as quality of life (QoL), mood, physical symptoms, support, coping, PTSD, pain, hope, sleep, or motivation. Moreover, the measurement of HL and related variables varied greatly across the studies, with only a few conducted using thorough evaluations and specific scales to measure HL, such as the Health Literacy Questionnaire (HLQ), the Health Literacy Survey questionnaire (HLS-Q), the Short Test of Functional HL (STFHL), the Rapid Estimate of Adult Literacy in medicine, or the eHealth Literacy scale. Given also that some interventions did not concretely assess HL, determining the intervention’s effect on HL levels in some reviews remains unclear and heterogeneous, making it difficult to draw definitive conclusions. 3.5.2. Health Literacy as an Outcome Of the ten reviews included in this study, only four contained studies that explicitly considered HL as an intervention outcome. Of these, two remain rather general concerning the operationalization of HL: Heine et al.’s (4) review clearly mentions HL as an outcome but does not specify how it was operationalized in the studies included in their review. Housten et al. (5) mention that the interventions, which aimed to improve specific aspects of HL, led to a significant improvement in two interventions , but that the outcomes varied depending on whether the baseline level of HL was limited or adequate. The review by Fernandez-Gonzalez and Bravo-Valenzuela (7) , via various questionnaires, measured HL as an outcome, along with other factors, such as self-efficacy, motivation, etc. The fourth review by Verweel et al. (10) is rather specific, in the sense that it looked at HL interventions among patients with cancer through digital media. Only one study in this review measured HL as an outcome, while the other ones measured specific skills such as ‘cancer competence’. Interestingly, the one that considered digital health literacy as the outcome showed a significant improvement in comparison to the control group, whereas the other four studies yielded contradictory evidence, with only some of them showing significant change in competence levels, compared to the control groups. The review by McAlpine et al. (1) also describes health competence as an outcome of education interventions, but does not provide any further information on the measurement, impact, or implications. The review included only one study that explicitly targeted health competence, measured via an 8-item questionnaire, but found no significant improvement. The review by Hill et al. (2) mentions the impact of the different interventions on the levels of HL without detailing the operationalizations of HL and finding mixed evidence of efficacy for online interventions, while DeRosa et al. (3) concluded that in most of the studies they reviewed, HL increased as a result of the introduction of navigators that help and support decision making. While ‘better HL awareness’ was identified as a positive decision-making outcome, there was no mention of HL measurement. Van der Kruk and colleagues (6) , who reviewed the impact of using virtual reality (VR) in patient education, reported that most participants had a low baseline HL level, which improved as a result of the interventions, but no measurements of HL were mentioned. Finally, the review by Munstermann et al. (9) used the term ‘HL interventions’ to refer to educational interventions and evaluated the impact of these interventions on the participants’ HL levels, but again, does not specify which specific HL measurement methods were used. 3.5.3. Health Literacy as a Moderator Contrary to the reviews that considered HL as an intervention outcome (4,5,9) , some reviews looked at the moderating effects of HL (8) . Some of these reviews also considered HL as an intervention outcome, and although the specificities regarding to the measurement of the potential moderating impact of HL were mostly unclear, the analyses and conclusion that are drawn by the reviewers suggest that HL is mainly seen to have a moderating effect. Specifically, low HL is seen to act as a barrier preventing the access to population-appropriate healthcare, while higher HL facilitates access to care and affects patient–physician relationships positively. Munstermann and colleagues (9) reported that lower HL was related to inequalities and the inaccessibility of appropriate care in deaf and hard-of-hearing patients, influencing the effect of educational interventions on cancer-knowledge and quality of life. On a similar note, the review by Cabanes et al. (8) considers HL as a means of ‘supportive care’, allowing for a more positive impact on the quality of life and on the reduction in the ‘burden’ of cancer. While these two reviews consider the moderating effect of HL indirectly, the reviews by Heine et al. (4) and Housten, et al. (5) are more explicit about the moderator role of HL on the effects of interventions. The first identifies HL as a moderator of the effects of educational interventions with patients with cancer regarding lifestyle and dietary changes, whereas the second explicitly uses the term ‘modification’ to describe how HL influenced the effects of the interventions with patients with cancer regarding screening and tasks such as recall and recognition. 3.5.4. Effects of HL Interventions on Other Outcome Variables In addition to being an outcome of an educational intervention, or a moderator of its effects on other outcome measures, HL can also be considered the main theme of an intervention, the effects of which are then assessed via other variables. The review by DeRosa et al. (3) concluded that interventions aiming to enhance ‘HL awareness’ resulted in more satisfaction and self-efficacy, which are in turn linked to decision adherence. Heine et al.’s (4) review, which considered various non-communicable diseases, showed an increase in knowledge, attitude (self-efficacy, motivation, etc.), and self-management behavior, albeit more so among diabetes patients than among patients with cancer. The review by Verweel and colleagues (10) mentions an effect of digital HL interventions among patients with cancer on other outcome variables. Van der Kruk and colleagues (6) , who reviewed the impact of using virtual reality (VR) in HL-based patient education, found a significant improvement of knowledge, comprehension, and understanding in most of the studies. The review conducted by Fernandez–Gonzalez and Bravo–Valenzuela (7) showed correlations between HL and other variables such as self-care, knowledge, self-efficacy, and adherence, although the role of HL within those links was not made very clear. Finally, in 6 of the 35 articles reviewed by Cabanes et al. (8) , HL-based interventions had an overall positive effect on the QoL, which showed a significant improvement. However, while it is possible for the above-mentioned factors to have a link to HL, the diversity in the types of interventions and outcome measures and the lack of clarity regarding their measurements make it difficult to draw conclusions regarding the actual effect of HL on these variables. 3.5.5. Interventions The included reviews included various interventions, which tackled similar objectives with comparable strategies. Online interventions showed positive effects on patients with cancer; however, their significance was questioned, as the findings could’ve been affected by the outcome measures (1) . ‘Online’ interventions included platforms linking patients with clinics (and even with other patients) (1) , videos with illness-specific information, surveys, media, and talk sessions . Technology was said to have limited effects if administered alone, and should rather take various, accessible approaches (10) . Accessibility was thoroughly discussed and highlighted in most of the reviews included, especially through tailored interventions. Tailored interventions showed better results, whether the tailoring relates to the type of patient, to the type of cancer, or even to each individual; however, one review (10) that included results of cancer-specific interventions showed less significant results than, for example, HIV-, diabetes-, or COPD-specific interventions. The advantage of modifying said interventions to the person’s needs, and targeting each patient individually, was the conclusion of not only each intervention included in the reviews, but also by the reviewers of each review included (1,2,3,4,5,7,8,9,10) . Contextual appropriateness was also considered in the interventions; whether regarding age, culture, specific needs, etc. (2,3,4) . Finally, multilevel interventions showed better results, in the few studies where they were attempted (5,8) . 3.6. Conclusions The authors of the ten reviews drew different, yet concurrent, conclusions from their analysis. The mixed results that were reported after the interventions included in the reviews were partly attributed to the weak theoretical basis and weak operational definitions of HL used in many of the studies. Most studies considered HL as an idea, a concept or, at best, as one of the outcomes, but rarely as the primary one. Another point that was raised in several of the reviews (4,5,7,8) is the fact that HL, as a complex concept, is often approached holistically, rather than being broken down into its various dimensions for a deeper understanding and a more precise targeting. Moreover, the reported outcomes varied largely depending on the type of interventions, the target population, and the frequency, duration, and timing of the interventions (1) , which makes generalizable interventions more difficult. Tailored interventions were believed to be more effective, as shown in the studies involving deaf patients (2,9) , at-risk groups (3,7) , and specific type of cancers (7) . It was also pointed out that the effectiveness of HL interventions highly depends on the availability and accessibility of the resources that are required to implement them at different levels of care (prevention, screening, interventions) (4) , and that individual needs, disease-specific information, and preferences must be accounted for when designing HL interventions (10) . The integration of technology and education, delivered by trusted sources, was believed to be effective, even though no significant effects were reported in cancer-specific interventions. Stand-alone technology interventions had limited effects, while the education-based interventions combined with the technology-based approaches showed more promise (10) . The review papers included in this review included between ten (7) and fifty-three (4) primary studies. The total number of participants was not always clear. They included interventional studies, surveys, and simple comparative studies focusing on HL in the context of oncology. However, they all included at least one interventional study that targeted at least one of the aspects of HL (e.g., informed decision-making). The intervention studies varied from testing the efficacy of online interventions to analyzing the importance of promoting certain aspects such as decision-making, competence, and patient education. Although none of the reviews in this study targeted breast cancer specifically, breast cancer was the most often studied type of cancer, and was represented in all ten the reviews. The reviews by DeRosa et al. (3) and by Fernandez-Gonzalez and Bravo-Valenzuela (7) were concerned with HL in breast and prostate cancer, while the ones by McAlpine et al. (1) , Housten et al. (5) , and Cabanes et al. (8) had 50%, 40%, and one-third of their study population suffering from breast cancer, respectively. As mentioned above, three of the review studies (4,6,10) included cancer as one of several non-communicable diseases. The review by Heine et al. (4) included only one interventional study that targeted patients with cancer; in the interventional studies included in the review by van der Kruk et al. (6) , half were concerned with cancer; and the review by Verweel et al. (10) contained seventeen studies on chronic illnesses, four of which focused specifically on cancer, and two of which included cancer along with other diseases. In the review by Verweel et al. (10) , six of the studies were concerned with cancer, three of which were specifically focused on breast cancer. Other types of cancer, such as prostate and colorectal cancer, were highly present. Most of the reviews focused on HL among individual adult patients, with an average age ranging from 42 to 70 years. However, the review by Housten et al. (5) also included the HL of peers and caregivers. DeRosa et al.’s review (3) also mentioned the importance of including family and peers (social support) in education and decision-aid of patients with cancer, as they play a crucial role in addressing HL. The review by Hill et al. (2) also included interventions involving health professionals, since they fit into their target of building patient-specific culturally competent care. Four of the reviews (2,3,7,9) targeted minority populations, notably deaf patients (2,9) and African, Hispanic, Latin, or Asian populations (3,7) . Almost all the reviews included studies involving primarily online interventions or online assistance for face-to-face interventions. Five reviews (1,6,7,9,10) were concerned with studies that were completely online, while the others (2,3,4,5,7) considered mixed methods, such as face-to-face interventions (group or individual), interviews, etc. The interventions that did not occur online, such as interviews (2), educational programs (1,2,3,4,6,7,8) , information sessions (2,3,4,7,8) , workshops (3,5,6,8) , and handouts (4,5) , were mostly delivered by the research team, health educators, or medical staff such as nurses, pharmacists, and/or social workers. The review by Verweel et al. (10) focused specifically on digital health literacy, and therefore included solely digital tools, whether internet-based or using other digital means (such as computer software, smart devices, websites, learning management systems, and electronic personal health records (ePHR)), while study 6 had a virtual reality approach. Educational videos were the most commonly used approach to address health literacy, featured in at least one intervention across all ten reviews. The subjects of the educational interventions varied from specific topics, such as fatigue, insomnia, fertility, diet, and smoking cessation, to more general topics, involving cancer-related knowledge, screening information, and general symptoms. Other interventions included coping skills training, communication skills training (for professionals), symptom monitoring and self-management, self-care, treatment adherence, and decision-making. While these programs are clearly linked to HL, the reviews did not provide enough detail about the content to clearly define this connection. Nevertheless, it is clear that education, digital or not, is the main tool to address HL among patients with cancer. 3.5.1. Operationalization of HL The outcomes of the reviews comprised in this review are shown in . All reviews included in this study involved one or more interventions, and, in accordance with the inclusion criteria, had to focus on HL as a theme within the interventions to be included in the study. While all of them analyzed the impact of certain factors related to HL, not all necessarily considered HL itself as a measured outcome. Other outcomes that were measured were decision-making, adherence, knowledge, understanding, communication, self-efficacy and self-management, which are related to HL, or a part of its definition. In some studies, however, the outcomes were less directly related to HL, such as quality of life (QoL), mood, physical symptoms, support, coping, PTSD, pain, hope, sleep, or motivation. Moreover, the measurement of HL and related variables varied greatly across the studies, with only a few conducted using thorough evaluations and specific scales to measure HL, such as the Health Literacy Questionnaire (HLQ), the Health Literacy Survey questionnaire (HLS-Q), the Short Test of Functional HL (STFHL), the Rapid Estimate of Adult Literacy in medicine, or the eHealth Literacy scale. Given also that some interventions did not concretely assess HL, determining the intervention’s effect on HL levels in some reviews remains unclear and heterogeneous, making it difficult to draw definitive conclusions. 3.5.2. Health Literacy as an Outcome Of the ten reviews included in this study, only four contained studies that explicitly considered HL as an intervention outcome. Of these, two remain rather general concerning the operationalization of HL: Heine et al.’s (4) review clearly mentions HL as an outcome but does not specify how it was operationalized in the studies included in their review. Housten et al. (5) mention that the interventions, which aimed to improve specific aspects of HL, led to a significant improvement in two interventions , but that the outcomes varied depending on whether the baseline level of HL was limited or adequate. The review by Fernandez-Gonzalez and Bravo-Valenzuela (7) , via various questionnaires, measured HL as an outcome, along with other factors, such as self-efficacy, motivation, etc. The fourth review by Verweel et al. (10) is rather specific, in the sense that it looked at HL interventions among patients with cancer through digital media. Only one study in this review measured HL as an outcome, while the other ones measured specific skills such as ‘cancer competence’. Interestingly, the one that considered digital health literacy as the outcome showed a significant improvement in comparison to the control group, whereas the other four studies yielded contradictory evidence, with only some of them showing significant change in competence levels, compared to the control groups. The review by McAlpine et al. (1) also describes health competence as an outcome of education interventions, but does not provide any further information on the measurement, impact, or implications. The review included only one study that explicitly targeted health competence, measured via an 8-item questionnaire, but found no significant improvement. The review by Hill et al. (2) mentions the impact of the different interventions on the levels of HL without detailing the operationalizations of HL and finding mixed evidence of efficacy for online interventions, while DeRosa et al. (3) concluded that in most of the studies they reviewed, HL increased as a result of the introduction of navigators that help and support decision making. While ‘better HL awareness’ was identified as a positive decision-making outcome, there was no mention of HL measurement. Van der Kruk and colleagues (6) , who reviewed the impact of using virtual reality (VR) in patient education, reported that most participants had a low baseline HL level, which improved as a result of the interventions, but no measurements of HL were mentioned. Finally, the review by Munstermann et al. (9) used the term ‘HL interventions’ to refer to educational interventions and evaluated the impact of these interventions on the participants’ HL levels, but again, does not specify which specific HL measurement methods were used. 3.5.3. Health Literacy as a Moderator Contrary to the reviews that considered HL as an intervention outcome (4,5,9) , some reviews looked at the moderating effects of HL (8) . Some of these reviews also considered HL as an intervention outcome, and although the specificities regarding to the measurement of the potential moderating impact of HL were mostly unclear, the analyses and conclusion that are drawn by the reviewers suggest that HL is mainly seen to have a moderating effect. Specifically, low HL is seen to act as a barrier preventing the access to population-appropriate healthcare, while higher HL facilitates access to care and affects patient–physician relationships positively. Munstermann and colleagues (9) reported that lower HL was related to inequalities and the inaccessibility of appropriate care in deaf and hard-of-hearing patients, influencing the effect of educational interventions on cancer-knowledge and quality of life. On a similar note, the review by Cabanes et al. (8) considers HL as a means of ‘supportive care’, allowing for a more positive impact on the quality of life and on the reduction in the ‘burden’ of cancer. While these two reviews consider the moderating effect of HL indirectly, the reviews by Heine et al. (4) and Housten, et al. (5) are more explicit about the moderator role of HL on the effects of interventions. The first identifies HL as a moderator of the effects of educational interventions with patients with cancer regarding lifestyle and dietary changes, whereas the second explicitly uses the term ‘modification’ to describe how HL influenced the effects of the interventions with patients with cancer regarding screening and tasks such as recall and recognition. 3.5.4. Effects of HL Interventions on Other Outcome Variables In addition to being an outcome of an educational intervention, or a moderator of its effects on other outcome measures, HL can also be considered the main theme of an intervention, the effects of which are then assessed via other variables. The review by DeRosa et al. (3) concluded that interventions aiming to enhance ‘HL awareness’ resulted in more satisfaction and self-efficacy, which are in turn linked to decision adherence. Heine et al.’s (4) review, which considered various non-communicable diseases, showed an increase in knowledge, attitude (self-efficacy, motivation, etc.), and self-management behavior, albeit more so among diabetes patients than among patients with cancer. The review by Verweel and colleagues (10) mentions an effect of digital HL interventions among patients with cancer on other outcome variables. Van der Kruk and colleagues (6) , who reviewed the impact of using virtual reality (VR) in HL-based patient education, found a significant improvement of knowledge, comprehension, and understanding in most of the studies. The review conducted by Fernandez–Gonzalez and Bravo–Valenzuela (7) showed correlations between HL and other variables such as self-care, knowledge, self-efficacy, and adherence, although the role of HL within those links was not made very clear. Finally, in 6 of the 35 articles reviewed by Cabanes et al. (8) , HL-based interventions had an overall positive effect on the QoL, which showed a significant improvement. However, while it is possible for the above-mentioned factors to have a link to HL, the diversity in the types of interventions and outcome measures and the lack of clarity regarding their measurements make it difficult to draw conclusions regarding the actual effect of HL on these variables. 3.5.5. Interventions The included reviews included various interventions, which tackled similar objectives with comparable strategies. Online interventions showed positive effects on patients with cancer; however, their significance was questioned, as the findings could’ve been affected by the outcome measures (1) . ‘Online’ interventions included platforms linking patients with clinics (and even with other patients) (1) , videos with illness-specific information, surveys, media, and talk sessions . Technology was said to have limited effects if administered alone, and should rather take various, accessible approaches (10) . Accessibility was thoroughly discussed and highlighted in most of the reviews included, especially through tailored interventions. Tailored interventions showed better results, whether the tailoring relates to the type of patient, to the type of cancer, or even to each individual; however, one review (10) that included results of cancer-specific interventions showed less significant results than, for example, HIV-, diabetes-, or COPD-specific interventions. The advantage of modifying said interventions to the person’s needs, and targeting each patient individually, was the conclusion of not only each intervention included in the reviews, but also by the reviewers of each review included (1,2,3,4,5,7,8,9,10) . Contextual appropriateness was also considered in the interventions; whether regarding age, culture, specific needs, etc. (2,3,4) . Finally, multilevel interventions showed better results, in the few studies where they were attempted (5,8) . The outcomes of the reviews comprised in this review are shown in . All reviews included in this study involved one or more interventions, and, in accordance with the inclusion criteria, had to focus on HL as a theme within the interventions to be included in the study. While all of them analyzed the impact of certain factors related to HL, not all necessarily considered HL itself as a measured outcome. Other outcomes that were measured were decision-making, adherence, knowledge, understanding, communication, self-efficacy and self-management, which are related to HL, or a part of its definition. In some studies, however, the outcomes were less directly related to HL, such as quality of life (QoL), mood, physical symptoms, support, coping, PTSD, pain, hope, sleep, or motivation. Moreover, the measurement of HL and related variables varied greatly across the studies, with only a few conducted using thorough evaluations and specific scales to measure HL, such as the Health Literacy Questionnaire (HLQ), the Health Literacy Survey questionnaire (HLS-Q), the Short Test of Functional HL (STFHL), the Rapid Estimate of Adult Literacy in medicine, or the eHealth Literacy scale. Given also that some interventions did not concretely assess HL, determining the intervention’s effect on HL levels in some reviews remains unclear and heterogeneous, making it difficult to draw definitive conclusions. Of the ten reviews included in this study, only four contained studies that explicitly considered HL as an intervention outcome. Of these, two remain rather general concerning the operationalization of HL: Heine et al.’s (4) review clearly mentions HL as an outcome but does not specify how it was operationalized in the studies included in their review. Housten et al. (5) mention that the interventions, which aimed to improve specific aspects of HL, led to a significant improvement in two interventions , but that the outcomes varied depending on whether the baseline level of HL was limited or adequate. The review by Fernandez-Gonzalez and Bravo-Valenzuela (7) , via various questionnaires, measured HL as an outcome, along with other factors, such as self-efficacy, motivation, etc. The fourth review by Verweel et al. (10) is rather specific, in the sense that it looked at HL interventions among patients with cancer through digital media. Only one study in this review measured HL as an outcome, while the other ones measured specific skills such as ‘cancer competence’. Interestingly, the one that considered digital health literacy as the outcome showed a significant improvement in comparison to the control group, whereas the other four studies yielded contradictory evidence, with only some of them showing significant change in competence levels, compared to the control groups. The review by McAlpine et al. (1) also describes health competence as an outcome of education interventions, but does not provide any further information on the measurement, impact, or implications. The review included only one study that explicitly targeted health competence, measured via an 8-item questionnaire, but found no significant improvement. The review by Hill et al. (2) mentions the impact of the different interventions on the levels of HL without detailing the operationalizations of HL and finding mixed evidence of efficacy for online interventions, while DeRosa et al. (3) concluded that in most of the studies they reviewed, HL increased as a result of the introduction of navigators that help and support decision making. While ‘better HL awareness’ was identified as a positive decision-making outcome, there was no mention of HL measurement. Van der Kruk and colleagues (6) , who reviewed the impact of using virtual reality (VR) in patient education, reported that most participants had a low baseline HL level, which improved as a result of the interventions, but no measurements of HL were mentioned. Finally, the review by Munstermann et al. (9) used the term ‘HL interventions’ to refer to educational interventions and evaluated the impact of these interventions on the participants’ HL levels, but again, does not specify which specific HL measurement methods were used. Contrary to the reviews that considered HL as an intervention outcome (4,5,9) , some reviews looked at the moderating effects of HL (8) . Some of these reviews also considered HL as an intervention outcome, and although the specificities regarding to the measurement of the potential moderating impact of HL were mostly unclear, the analyses and conclusion that are drawn by the reviewers suggest that HL is mainly seen to have a moderating effect. Specifically, low HL is seen to act as a barrier preventing the access to population-appropriate healthcare, while higher HL facilitates access to care and affects patient–physician relationships positively. Munstermann and colleagues (9) reported that lower HL was related to inequalities and the inaccessibility of appropriate care in deaf and hard-of-hearing patients, influencing the effect of educational interventions on cancer-knowledge and quality of life. On a similar note, the review by Cabanes et al. (8) considers HL as a means of ‘supportive care’, allowing for a more positive impact on the quality of life and on the reduction in the ‘burden’ of cancer. While these two reviews consider the moderating effect of HL indirectly, the reviews by Heine et al. (4) and Housten, et al. (5) are more explicit about the moderator role of HL on the effects of interventions. The first identifies HL as a moderator of the effects of educational interventions with patients with cancer regarding lifestyle and dietary changes, whereas the second explicitly uses the term ‘modification’ to describe how HL influenced the effects of the interventions with patients with cancer regarding screening and tasks such as recall and recognition. In addition to being an outcome of an educational intervention, or a moderator of its effects on other outcome measures, HL can also be considered the main theme of an intervention, the effects of which are then assessed via other variables. The review by DeRosa et al. (3) concluded that interventions aiming to enhance ‘HL awareness’ resulted in more satisfaction and self-efficacy, which are in turn linked to decision adherence. Heine et al.’s (4) review, which considered various non-communicable diseases, showed an increase in knowledge, attitude (self-efficacy, motivation, etc.), and self-management behavior, albeit more so among diabetes patients than among patients with cancer. The review by Verweel and colleagues (10) mentions an effect of digital HL interventions among patients with cancer on other outcome variables. Van der Kruk and colleagues (6) , who reviewed the impact of using virtual reality (VR) in HL-based patient education, found a significant improvement of knowledge, comprehension, and understanding in most of the studies. The review conducted by Fernandez–Gonzalez and Bravo–Valenzuela (7) showed correlations between HL and other variables such as self-care, knowledge, self-efficacy, and adherence, although the role of HL within those links was not made very clear. Finally, in 6 of the 35 articles reviewed by Cabanes et al. (8) , HL-based interventions had an overall positive effect on the QoL, which showed a significant improvement. However, while it is possible for the above-mentioned factors to have a link to HL, the diversity in the types of interventions and outcome measures and the lack of clarity regarding their measurements make it difficult to draw conclusions regarding the actual effect of HL on these variables. The included reviews included various interventions, which tackled similar objectives with comparable strategies. Online interventions showed positive effects on patients with cancer; however, their significance was questioned, as the findings could’ve been affected by the outcome measures (1) . ‘Online’ interventions included platforms linking patients with clinics (and even with other patients) (1) , videos with illness-specific information, surveys, media, and talk sessions . Technology was said to have limited effects if administered alone, and should rather take various, accessible approaches (10) . Accessibility was thoroughly discussed and highlighted in most of the reviews included, especially through tailored interventions. Tailored interventions showed better results, whether the tailoring relates to the type of patient, to the type of cancer, or even to each individual; however, one review (10) that included results of cancer-specific interventions showed less significant results than, for example, HIV-, diabetes-, or COPD-specific interventions. The advantage of modifying said interventions to the person’s needs, and targeting each patient individually, was the conclusion of not only each intervention included in the reviews, but also by the reviewers of each review included (1,2,3,4,5,7,8,9,10) . Contextual appropriateness was also considered in the interventions; whether regarding age, culture, specific needs, etc. (2,3,4) . Finally, multilevel interventions showed better results, in the few studies where they were attempted (5,8) . The authors of the ten reviews drew different, yet concurrent, conclusions from their analysis. The mixed results that were reported after the interventions included in the reviews were partly attributed to the weak theoretical basis and weak operational definitions of HL used in many of the studies. Most studies considered HL as an idea, a concept or, at best, as one of the outcomes, but rarely as the primary one. Another point that was raised in several of the reviews (4,5,7,8) is the fact that HL, as a complex concept, is often approached holistically, rather than being broken down into its various dimensions for a deeper understanding and a more precise targeting. Moreover, the reported outcomes varied largely depending on the type of interventions, the target population, and the frequency, duration, and timing of the interventions (1) , which makes generalizable interventions more difficult. Tailored interventions were believed to be more effective, as shown in the studies involving deaf patients (2,9) , at-risk groups (3,7) , and specific type of cancers (7) . It was also pointed out that the effectiveness of HL interventions highly depends on the availability and accessibility of the resources that are required to implement them at different levels of care (prevention, screening, interventions) (4) , and that individual needs, disease-specific information, and preferences must be accounted for when designing HL interventions (10) . The integration of technology and education, delivered by trusted sources, was believed to be effective, even though no significant effects were reported in cancer-specific interventions. Stand-alone technology interventions had limited effects, while the education-based interventions combined with the technology-based approaches showed more promise (10) . This review of reviews summarized the existing systematic reviews and meta-analyses of published studies investigating HL as an outcome, a moderator of outcomes, or a component of interventions in patients with cancer. Although all the reviews included in the study mention HL as a part of the interventions, the role of HL varies significantly between the reviews. Many reviews saw HL as an outcome, even though the outcome measurement was not always appropriate, clearly defined, or detailed. Others considered HL as a moderator, but the measurement methods are unclear. Some reviews considered HL as both an outcome and a moderator, and some saw it mainly as a type or a component of an educational intervention with patients with cancer. The majority of interventions focused on adult patients, but some also included peers, family, or healthcare providers. The interventions in the studies varied widely in terms of format, but several involved an online component, and some included VR and eHealth. Other than HL or outcomes moderated by HL, the most commonly used outcomes measured were knowledge, adherence, attitudes (self-efficacy), self-care, and decision-making-related variables. The reviews pointed to the importance of tailored interventions. Contextual appropriateness as well as individual needs were themes that emerged frequently and could fall under the umbrella of ‘tailored’ interventions. Sudore and Schillinger stress the importance of tailoring communication to the patient’s perceived barriers and their needs for better quality interactions. Brooks and colleagues took a different approach by showing the importance of building trust to improve the health literacy of elderly patients. This was conducted through more thought-out, tailored interventions that met individual needs. According to Salter et al., HL levels vary depending on the healthcare system requirements and a person’s particular skills. However, in elderly patients, while the difficulty of dealing with online tools was recognized in discussions of the reviews, it was not specified how to adapt these digital tools to the population in question, or even the part of the population that does not have access to digital tools. Also, DeMarco and Nystrom emphasized the adaptation of patient education tools to patient-specific needs, which led to positive results. Therefore, the importance of intervening with tools that are adapted to the individual’s existing strengths and limitations is appears in the more positive results of tailored interventions. Nevertheless, tailored health literacy interventions require an exhaustive understanding of HL as a concept. In the articles included in this review, authors pointed out the weak understanding of the implementation and consideration of the different dimensions of HL, in addition to weak reported correlations between the factors measured as outcomes and their relation to HL, and the role that HL plays sometimes in moderating those effects. Definitions of HL have pointed out several aspects to be targeted when addressing a HL intervention. Sorensen et al. explain HL through a set of cognitive and behavioral skills needed to make decisions and apply health information. Nutbeam also underlined the social skills needed for HL and pointed out its different levels. Those different levels were to be taken into consideration within the interventions, therefore categorizing health literacy interventions into functional, interactive, and critical aspects, and focusing on the clinical settings, emphasizing the efforts to enhance health literacy among healthcare professionals and simplify healthcare organizations. In his review with Llyod , they explore interventions for community populations, underscoring the importance of transferable health literacy skills and ensuring accessibility to different populations. The review advocates for a shift in intervention focus towards improving communication quality, developing transferable skills, and prioritizing interventions for populations disproportionately affected by low health literacy. In a systematic review by Liu et al. the concepts that emerged from reviewing HL definitions were health decisions, functioning in a healthcare environment, promoting and maintaining health behavior, understanding, and gaining access to healthcare. Those concepts were further explored in Liu’s analysis and showed that every aspect of HL had many factors relating to the patient himself, the healthcare practitioners, and the healthcare system itself. For example, Liu and colleagues mentioned knowledge of health information as not only knowing the information but understanding the terms, being able to discern the relevant information, contextually adapting the information received, and using the information relevant to oneself which is linked to accessing the appropriate resources. In this review, the interventions included looked at the retention of information given in the intervention, but no testing of the usage, management, and processing of the information contextually was conducted clearly. The focus was mainly on giving out the information in an adapted manner, which is a part of implementing HL practices; however, the impact, implementation, management, and application of this information was either rarely measured or vaguely reported. On the other hand, decision-making was evaluated in some interventions following decision-aid procedures; however, no explicit link was made with the type of skills required for these decisions to be taken. Maintaining health through management and partnership with the health institutions was also pointed out in several HL definitions; communication with healthcare professionals appeared in some of the reviews included, but still, no measurement of the impact of that communication was measured in the long-term. The above results show a robust framework concerning HL in most cases, addressing the appropriate questions concerning education, decision making, increasing HL; however, the application and implantation of those practices, as well as the adapted clear measurement of its results and correlations, were unclear, insufficient or/and, in some cases, non-existent. The following lead to an incomprehension of how the definitions and frameworks, relating to understanding and increasing HL, are used, leading to inconclusive results for the most part. The authors pointed out that most of the interventions included were emerging and novel interventions that need further testing and analysis. An important point to cover is the long-term effect of the interventions. Interventions either did not measure long-term effects, or they did, which ended up showing a lack of effects being retained. This outcome can be attributed to the lack of interventions focusing on self-management and the acquisition of skills needed to maintain health. Another point this outcome can be attributed to is the complexity of the healthcare system that requires constant interventions in order to accompany the patient. Parnell et al. explained the importance of redefining HL as a complex concept, not only relating to the patients’ capacities, but also the healthcare system’s demands and resources. The organization plays a role in adapting to their patients’ HL levels, which can create better environments that promote self-efficacy, self-management, and a better understanding and application of health information . No intervention included in this review targeted organizational aspects that would allow for better HL outcomes, limiting the spectrum of HL interventions to mostly the knowledge and information aspects of HL. Kaper and colleagues studied the effectiveness level of OHL interventions at the different (patient, professional, and organization) levels. The review found promising results on the patient level, and intermediate outcomes on the professional and organizational levels. Recent studies are just starting to understand the importance of OHL interventions’ impact, but despite the advancements in the understanding, evidence and implementation remain weak. Kaper and colleagues called for a deeper assessment of the outcomes and development of reliable measurements for a comprehensive analysis. In addition, a longitudinal study conducted by Kaper and colleagues showed how the involvement of organizations and professionals helped to identify the implementation barriers, which led to more positive, long-term results in the assessment of OHL intervention implementation in Irish and Dutch hospitals. This information correlates with an observation in one of the reviews included (2) , which explains that inadequate HL poses as a barrier for patients with hearing difficulties, but that the responsibility lies with the organization. Access to appropriate care within institutions, in this case specifically the access to interpreters, adapted healthcare models, and linguistically and culturally competent providers, plays a huge role in breaking that barrier and allowing patients with all levels of HL to benefit from standard care, which will in turn increase the QoL. The review also called for the better training of professionals and for the creation of acceptable standards of health information delivery for different backgrounds. Liu et al. and Parnell et al. demonstrated the main aspect of HL interventions, primarily focusing on providing knowledge and information. In addition, as established in this review, most interventions did not measure HL itself as an outcome, but rather factors such as the QoL, self-efficacy, decision-making, etc. Even though factors such as self-efficacy appeared to be pertinent moderators/mediators of HL levels, and vice versa, and factors such as the QoL have proved to be impacted by HL levels , HL was not always clearly measured to check for impact and/or correlation. Only one study, by Verweel and colleagues (10) , correctly defined and included studies that, at least partially, mentioned and/or targeted HL within its role as an outcome or as a moderator. Even though it is unclear whether the articles included in the review that date from 2011 to 2022 correctly defined HL, digital HL, or competence, the review itself adopts an appropriate approach to the term. In one of the review’s studies, a health literacy assessment tool was employed, but it demonstrated no discernible impact. On the contrary, when various cancer-competence tools were utilized to gauge the intervention’s effectiveness—primarily employing illness-specific information tools—significant results were consistently observed. This might suggest that the use of a health literacy assessment might have led to insignificant findings, in contrast to the more targeted and specific cancer-related information competence questionnaires. Therefore, this calls not only for a better and more correct application of HL, based on its theoretical and practical framework, but also for the use of appropriate assessment tools based on what is being sought out during an intervention. On another note, although the patients’ digital education was a recurrent theme in that review, the recency of the review (2024) could attest to a better application and interpretation of HL, with recent focus on the subject. This unclear application and implementation of HL practices could be interpreted as an overuse of the term HL, particularly in contexts that are strictly focused on knowledge, education, and communication. This situation may also reflect a weak understanding of the term, due to its multidimensional and complex nature. A more suitable approach would be to specify the aspects that need to be implemented, especially if the goal is to focus on and enhance only one aspect of HL, and then to correctly assess the results, in order to understand its specific impact on the targeted variable, and to see if there a ripple effect would occur. Although the evaluated studies frequently included health literacy in patient education, they also frequently did not provide a thorough examination of its more complex aspects. Due to their focus on the particular needs of the groups and the provided targeted strategies, content, and tailored interventions that targeted particular populations—like migrant women or the deaf—showed greater effectiveness. However, there is little information available on the long-term results, making it difficult to pinpoint the exact effects and goals of the intervention components. Most interventions were short-term and direct, fulfilling the immediate goals of patient education, but perhaps being less successful in high-stress scenarios like cancer care. Furthermore, it was observed that one-shot therapies were time-consuming and frequently dependent on outside assistance, which made their incorporation into current system less feasible. Instead of only improving the patients’ health literacy skills, organizational health literacy interventions could help them manage their illnesses more effectively through improved self-management and an adjustment to the patients’ needs. The above-discussed remarks call for suitable application and assessment of health literacy when performing or applying interventions specific to HL or having a HL component. The use of self-efficacy measures or testing information received during an intervention do not reflect on HL in itself. The links between HL and the other components measured (quality of life, self-efficacy, satisfaction, decision-making) can be highlighted and explained to better map out the effects and outcomes of the interventions directly related to HL. In other words, a clearer portrayal of the role of HL (outcome, moderator, mediator) is crucial when presenting the findings. The results of the review also called for tailored interventions, as those show better results and are more accessible to minor populations. Furthermore, better results have also been found when using population-specific instruments, such as cancer-specific HL scales and tests. Finally, the outcomes of this review draw attention to the need for healthcare organizations to consider incorporating accessible interventions, tailored to individual needs, within oncology departments, as a way to render HL more accessible and to emphasize the responsibility of the organization to facilitate access to HL services, rather than focusing solely on educating the patient, which has shown to have little to no long-term effects. A call for a more multilevel, interventional approach, focusing on the three levels: organization, professionals, and patients, is indispensable, considering the different outcomes and the recent emerging theories. The inclusion criteria explicitly mentioned HL as an outcome or a moderator; however, many of the studies included, despite mentioning HL, did not measure it. This can be considered a limitation, since many other reviews, that included similar interventions which did not mention HL, but measured the same outcomes, were excluded. Therefore, being stricter with the term ‘HL’ and its use within the reviews would allow for a more specific review, solely focused on HL, rather than its simplified interpretations. Three languages were included, but only English search results were found, limiting the scope of the investigation. Standardized measures were not used across the studies, making it difficult to perform a meta-analysis. While we aimed to provide a comprehensive synthesis of the available evidence, the absence of a formal heterogeneity assessment and a meta-analysis prevented us from quantifying the degree of variability among the included studies. This limitation implies that the observed results should be interpreted with caution, and that the generalization of findings in diverse populations or settings may be influenced by a potential heterogeneity. Future research should strive to incorporate robust measures of heterogeneity to enhance the reliability and validity of the meta-analytic findings. This would contribute to a more nuanced understanding of the variability in study outcomes and strengthen the overall quality of evidence in the field. This review encourages adequate applications based on the definition of HL, extending beyond the sole focus on communication and education into targeting interventions on multiple levels of HL through mixed-method interventions. Furthermore, the positive outcomes of interventions involving healthcare professionals and peers call for changing the mindset that HL is solely the responsibility of patients, and that it instead addresses the different levels such as patients, environment, professionals, and organizations. Lastly, it addresses the need to use explicit measures of HL as a primary outcome or a moderator when the interventions’ target is HL, in addition to utilizing population-specific strategies and instruments. |
Enhancing ophthalmology education through a mobile flipped classroom: a new teaching method | 2afb8b8a-7e02-421b-a500-31c4dfb27de1 | 11753106 | Ophthalmology[mh] | Traditional didactic lectures in classrooms often lack effectiveness, as this passive method leads to superficial learning that is quickly forgotten . As Boyer noted, prolonged lectures without active student engagement are highly ineffective . In recent years, learners’ needs, goals, and performance have evolved. Students now expect rapid access to information and prefer collaborative environments with learner-centered activities . Advanced technologies, increased online content, and cognitive science have challenged traditional education approaches . One method to transform teaching and incorporate online resources is the flipped classroom model . The flipped classroom flips traditional learning - content acquisition is done individually outside of class through video lectures, while interactive group activities occur in-class . This emphasizes using pre-class educational videos for self-directed learning . Learner’s view content beforehand, so class time focuses on active application and workshops rather than passive listening . In medical education, flipped classrooms have improved student satisfaction, academic performance, and outcomes across disciplines including medicine , pharmacy , dentistry , and ophthalmology . However, no studies have examined effects in ophthalmology assistants specifically. Glaucoma is a major cause of irreversible blindness globally . Optical coherence tomography (OCT) aids early glaucoma diagnosis through high-resolution cross-sectional imaging . However, interpreting OCT scans requires familiarity with the technology and image analysis. This study investigates the impact of a mobile flipped classroom model for teaching OCT interpretation in glaucoma on satisfaction among ophthalmology assistants. Flipped classroom approaches may provide an effective method to improve glaucoma education. We aim to provide insights that could guide improvements in educational content and delivery methods. Study design and participants This quasi-experimental pre-test/post-test study was conducted among ophthalmology residents at Shahid Beheshti University of Medical Sciences. A total of 22 residents (from years 1–3) were recruited through census sampling to expand the sample size. This sample size was limited by the number of available residents in the program; thus, future studies with a larger sample size and multiple sites would be ideal for broader applicability. Inclusion and exclusion criteria’s Inclusion criteria were: interest in participating, and access to mobile devices. Exclusion criteria were: failure to complete questionnaires or tests, non-participation in the educational intervention, and stated inability to participate. Instruments The Educational Teaching Method Satisfaction Questionnaire (EMSQ) assessed learner satisfaction. This validated 10-item tool uses a 0–10 scale, with higher scores indicating greater satisfaction . Pre- and post-tests evaluated glaucoma OCT interpretation knowledge. Tests were developed by a committee of 3 glaucoma specialists and 1 medical education expert . The exam blueprint aligned questions with learning objectives (Table ). Validity was established through expert review. Reliability (Kuder-Richardson 20 = 0.80) was determined during piloting. Procedures Sixteen modules focused on the interpretation of OCT for glaucoma cover a range of essential topics, including an introduction to various OCT devices, the significance of each device, their distinguishing features, and methodologies for interpreting the resulting images. Each module is designed to be interactive and incorporates multimedia elements such as instructional videos, case studies, and quizzes to enhance learner engagement and retention. All educational content is hosted on the web application OphthalMobilE (Ophthalmology Mobile Education), which provides a user-friendly interface for accessing materials ( https://ophthalmobile.ir/ ). Access to this web application is free for all study participants, ensuring that they can easily engage with the content at their convenience. In addition to the core content, the modules include interactive features such as discussion forums and feedback mechanisms, allowing residents to ask questions and receive guidance from instructors in real-time. After ethical approval and written consent, participants completed a pre-test on the study’s mobile platform. The educational content was delivered through a series of structured modules that included both asynchronous learning materials accessed prior to class and synchronous in-class discussions. Sixteen glaucoma OCT interpretation modules were subsequently uploaded weekly over 4 sessions, each lasting 15 min. Learners accessed modules at least 6 days before each class session. In-class, case discussions reinforced module content. Groups investigated cases using module knowledge, presented findings, and received feedback. Post-tests occurred after each session, with the EMSQ administered at the end of the study period. Data analysis Quantitative analyses included means, standard deviations, frequencies, percentages, independent/paired t-tests, and ANOVA using SPSS v26. Significance was set at p < 0.05. Ethics statement The study was approved ethically. Informed consent was obtained. Participation was voluntary, data were anonymized, and confidentiality was maintained. This study was approved with ethical approved number IR.SBMU.SME.REC.1401.078 by the Ethic Committee of School of Medical Education and Learning Technologies, Shahid Beheshti University of Medical Sciences. This quasi-experimental pre-test/post-test study was conducted among ophthalmology residents at Shahid Beheshti University of Medical Sciences. A total of 22 residents (from years 1–3) were recruited through census sampling to expand the sample size. This sample size was limited by the number of available residents in the program; thus, future studies with a larger sample size and multiple sites would be ideal for broader applicability. Inclusion criteria were: interest in participating, and access to mobile devices. Exclusion criteria were: failure to complete questionnaires or tests, non-participation in the educational intervention, and stated inability to participate. The Educational Teaching Method Satisfaction Questionnaire (EMSQ) assessed learner satisfaction. This validated 10-item tool uses a 0–10 scale, with higher scores indicating greater satisfaction . Pre- and post-tests evaluated glaucoma OCT interpretation knowledge. Tests were developed by a committee of 3 glaucoma specialists and 1 medical education expert . The exam blueprint aligned questions with learning objectives (Table ). Validity was established through expert review. Reliability (Kuder-Richardson 20 = 0.80) was determined during piloting. Sixteen modules focused on the interpretation of OCT for glaucoma cover a range of essential topics, including an introduction to various OCT devices, the significance of each device, their distinguishing features, and methodologies for interpreting the resulting images. Each module is designed to be interactive and incorporates multimedia elements such as instructional videos, case studies, and quizzes to enhance learner engagement and retention. All educational content is hosted on the web application OphthalMobilE (Ophthalmology Mobile Education), which provides a user-friendly interface for accessing materials ( https://ophthalmobile.ir/ ). Access to this web application is free for all study participants, ensuring that they can easily engage with the content at their convenience. In addition to the core content, the modules include interactive features such as discussion forums and feedback mechanisms, allowing residents to ask questions and receive guidance from instructors in real-time. After ethical approval and written consent, participants completed a pre-test on the study’s mobile platform. The educational content was delivered through a series of structured modules that included both asynchronous learning materials accessed prior to class and synchronous in-class discussions. Sixteen glaucoma OCT interpretation modules were subsequently uploaded weekly over 4 sessions, each lasting 15 min. Learners accessed modules at least 6 days before each class session. In-class, case discussions reinforced module content. Groups investigated cases using module knowledge, presented findings, and received feedback. Post-tests occurred after each session, with the EMSQ administered at the end of the study period. Quantitative analyses included means, standard deviations, frequencies, percentages, independent/paired t-tests, and ANOVA using SPSS v26. Significance was set at p < 0.05. The study was approved ethically. Informed consent was obtained. Participation was voluntary, data were anonymized, and confidentiality was maintained. This study was approved with ethical approved number IR.SBMU.SME.REC.1401.078 by the Ethic Committee of School of Medical Education and Learning Technologies, Shahid Beheshti University of Medical Sciences. Of the 22 participants, 10 (45.5%) were male and 12 (54.5%) females. Mean age was 29.42 ± 2.09 years (range 25–33). Most ( n = 14, 63.6%) were single, with 8 (36.4%) married. By year, 9 (42.9%) were second-year residents, 5 (23.8%) first-year, and the remainder third-year (Table ). The mean overall satisfaction score on the EMSQ was 74.05 ± 16.9, indicating high satisfaction. Satisfaction was slightly higher among females (78.42 ± 17.53) versus males (68.8 ± 15.33), but not significantly different ( p = 0.191) (Table ). By year, first-years had the highest satisfaction (79.4 ± 10.24) and third-years the lowest (71.43 ± 21.67), though not statistically significant ( p = 0.728) (Table ). Satisfaction was higher among single (76.71 ± 17.6) versus married (69.38 ± 15.59) participants, but the difference was not significant ( p = 0.339). The mean pre-test score was 8.84 ± 2.67 (range 3–16) and post-test was 11.67 ± 3.07 (range 4–16), a significant improvement ( p < 0.001). On the Cirrus OCT test, mean pre-test scores were 8.52 ± 2.71 (range 3–14) and post-test were 12.39 ± 2.71 (range 4–16), showing significant improvement ( p < 0.001). On the Optovue OCT test, pre-test scores averaged 7.38 ± 1.6 (range 4–10) and post-test 8.86 ± 1.46 (range 5–10), also a significant increase ( p = 0.009). For the Topcon OCT test, pre-test scores were 8.22 ± 1.2 (range 6–10) and post-test 9.36 ± 1.22 (range 6–10), a significant gain ( p = 0.004). Finally, for the Spectralis OCT test, pre-test scores averaged 10.67 ± 2.9 (range 4–16) and post-test 13.65 ± 2.92 (range 4–16), again showing significant improvement ( p < 0.001). Pre-test scores were highest for Spectralis (10.67 ± 2.9) and lowest for Optovue (7.38 ± 1.6) ( p < 0.001). Post-test scores were also highest for Spectralis (13.65 ± 2.92) and lowest for Optovue (8.86 ± 1.46) ( p < 0.001) (Table ). This study investigated the effectiveness of a mobile flipped classroom model on learning outcomes and satisfaction in OCT interpretation education for ophthalmology residents. The findings revealed significant improvements in test scores following the intervention. In summary, the educational approach notably enhanced OCT interpretation skills across all assessments, indicating that the mobile flipped classroom model is an effective strategy for advancing glaucoma education among ophthalmology residents. Overall pre-test scores averaged 8.84 ± 2.67 and significantly improved to 11.67 ± 3.07 post-intervention ( p < 0.001). These findings align with other research showing mobile flipped classrooms can enhance medical student learning in ophthalmology and other specialties . The educational intervention also received high satisfaction ratings. The mean score on the learner satisfaction questionnaire was 74.05 ± 16.09, denoting participants were highly satisfied. Similar studies on flipped classrooms for physician continuing education likewise found increased satisfaction and engagement . While virtual platforms like mobile learning can improve access to materials, they may not fully replace traditional in-person teaching for clinical skills in ophthalmology . This study combined mobile self-directed learning with face-to-face interactive case discussions. Blended approaches integrating technology with traditional methods are often most effective for medical education . Regarding the use of multiple OCT devices, training on how to interpret OCT images for diagnosing glaucoma using the four devices is essential for ophthalmology residents for two main reasons: (1) All four devices are utilized across different hospitals in our country, and (2) While the functionality of these devices is similar, the methods of reading and interpreting their images differ significantly. We emphasize that the inclusion of various OCT devices is a one of strength of our study. The positive impact of the flipped classroom teaching method was observed across all devices, demonstrating its effectiveness in training residents to use multiple OCT devices, even when taught simultaneously. Notably, the Spectralis device is more commonly used in academic centers, leading to greater access for assistants, which likely contributes to their higher scores compared to devices like Optovue. Understanding these differences can help improve educational content and approaches. Gradually transitioning from teacher-centered to learner-centered approaches through blended learning models like the mobile flipped classroom warrants consideration in medical curricula. Future research should examine effects on long-term knowledge retention and clinical skill application. Study limitations include coordination challenges for scheduling participants. Nonetheless, results demonstrate the mobile flipped classroom’s potential to significantly improve glaucoma OCT interpretation competency and satisfaction among ophthalmology trainees. Further research with larger, multi-center samples is needed to increase generalizability. Randomized controlled trial designs comparing the flipped classroom to other teaching methods would provide higher quality evidence. Longitudinal follow up studies could evaluate long-term knowledge and skill retention. Qualitative studies could provide insights into participant experiences and perspectives. Comparative effectiveness studies on flipped classroom variants and integration with other educational technologies would help optimize learning. Finally, research on supporting learner access to mobile platforms and mitigating disparities warrants consideration given the increasing use of technology in medical education. In summary, while this study demonstrates promising results, further rigorous research on mobile flipped classrooms in medical education can continue to inform implementation and best practices. Limitations This study has several limitations to consider. The small sample size from a single training program restricts the generalizability of the findings to broader populations. Additionally, the absence of long-term follow-up means that knowledge retention over time remains uncertain. The study design did not include a control group for comparison with the intervention, which limits our ability to attribute improvements solely to the educational approach. Furthermore, participant attitudes and satisfaction ratings may be influenced by response bias, which could affect the reliability of the reported outcomes. Lastly, scheduling challenges in coordinating participants were noted, which could be improved through enhanced communication and greater instructor availability. This study has several limitations to consider. The small sample size from a single training program restricts the generalizability of the findings to broader populations. Additionally, the absence of long-term follow-up means that knowledge retention over time remains uncertain. The study design did not include a control group for comparison with the intervention, which limits our ability to attribute improvements solely to the educational approach. Furthermore, participant attitudes and satisfaction ratings may be influenced by response bias, which could affect the reliability of the reported outcomes. Lastly, scheduling challenges in coordinating participants were noted, which could be improved through enhanced communication and greater instructor availability. This study implemented a mobile flipped classroom model for teaching OCT interpretation skills to ophthalmology residents and analyzed the effects on learning outcomes and satisfaction. Results demonstrated the educational intervention significantly improved test scores in OCT image analysis. Participants also reported high levels of satisfaction with the teaching method as reflected in the questionnaire responses. These findings suggest that mobile flipped classrooms may be an effective approach for enhancing clinical knowledge and skills within ophthalmology training programs. The model integrates pre-class mobile self-directed learning with active, collaborative application of concepts through in-person case discussions. This blended approach is consistent with principles of active learning and learner-centered education. Given the characteristics of today’s learners, the implementation of flipped classroom designs warrants consideration in other medical education contexts beyond ophthalmology. However, further research is necessary to evaluate long-term knowledge retention, skill acquisition, and the comparative effectiveness of this approach relative to traditional curricula. As technology continues to evolve, exploring innovative integrations of emerging modalities with interactive in-person activities will be crucial for fostering engagement and clinical competency. In conclusion, this study provides preliminary evidence supporting the benefits of a mobile flipped classroom in ophthalmology education. Continued research into optimal implementation strategies and learner-centered curricula will further enhance medical education to meet the evolving needs of 21st-century students and ultimately improve patient care. |
The role of glycans in personalization of preventive health care | 263d2c30-d3e0-4263-887a-04be18320263 | 11157255 | Preventive Medicine[mh] | |
Bivariate multilevel modeling of antenatal care contacts and place of delivery among reproductive-aged women in Ethiopia | 9ff212c5-12c9-466d-a043-2328fd6fe8a9 | 11809890 | Surgical Procedures, Operative[mh] | Antenatal care (ANC) visits and improved delivery facilities are crucial for maternal and child healthcare, as they improve the health of mothers and their new-borns. Regular ANC allows healthcare providers to monitor the mother’s health, detect potential complications early, and administer necessary interventions. ANC also helps women maintain healthy pregnancies by identifying pre-existing conditions and preventing complications during delivery. Improved delivery facilities ensure safe care during childbirth, reducing maternal and infant mortality risk. Place of delivery (PD) refers to the location of childbirth, either at home or in a health facility . Pregnancy and childbirth in Sub-Saharan Africa are characterized by low rates of health facility births, with 99% of maternal deaths occurring in developing countries. Over 800 women die daily due to preventable complications worldwide . In Africa, direct obstetric problems during delivery, such as haemorrhage, hypertension, sepsis, and obstructed labour, account for 64% of all maternal deaths . Maternal healthcare service utilization remains a crucial measure for tracking advancements in maternal and child health outcomes. However, Sub-Saharan countries, which have the world’s highest maternal mortality rate (420 per 100,000 live births), are associated with low rates of recommended antenatal care (ANC) visits and skilled birth attendance . Maternal mortality in Ethiopia was high, at 412 per 100,000 live births. Providing access to antenatal care (ANC) and skilled birth attendance can help prevent maternal and infant deaths . A joint multilevel analysis of antenatal care (ANC) and place of delivery has not been previously well examined for women of reproductive age in Ethiopia at the zonal level. While the number of women who attended at least four ANC contacts and potential influencing factors have been identified, the variations at the zonal administration level have not been addressed and most of the studies have focused on a separate analysis using antenatal and delivery care utilization along with their determinant factors in Ethiopia . Prioritizing antenatal care and the quality of delivery services, and a fair distribution of these services in the zone plays a vital role in fostering healthier communities and improving the overall quality of life for women and their children . The joint analysis using both antenatal care visits and place of delivery and identifying the associated risk factor can improve the estimates and will help us to put deliverable recommendations at a zonal level which is the lowest administration region for better intervention. Therefore, this study intended to determine possible factors that affect both ANC visits and place of delivery jointly among reproductive-aged women at the zonal level of Ethiopia. Study setting and design This study was conducted in Ethiopia, a country located in the horn of Africa and part of Sub-Saharan Africa, situated between 33° and 14° east longitude and 3° and 15° north latitude. A cross-sectional study design was employed using data from the 2019 Ethiopian Mini Demographic and Health Survey (EDHS), which was carried out by the Central Statistical Agency in collaboration with the Federal Ministry of Health and the Ethiopian Public Health Institute. Administratively, Ethiopia is divided into nine regional states and two city administrations, further subdivided into 68 zones . Data sources and study population The dataset from the 2019 Ethiopian Mini Demographic and Health Survey (EMDHS) was used for this study. This survey was the second EMDHS and the fifth DHS conducted in Ethiopia, carried out from March 21, 2019, to June 28, 2019. We utilized the individual record (IR) dataset and extracted both the dependent and independent variables. The data set used in this study is freely available and possible to download by justifying the reason for requesting the data from the link: https://dhsprogram.com/data/available-datasets.cfm . All Ethiopian women (15–49 years old) in the reproductive age range are considered as the study’s population. Women in Ethiopia who had children within the previous five years of the survey for the most recent birth were involved in the study. Sample size and sampling procedure The 2019 EMDHS was carried out among 8,663 households to ensure national representation. The survey included interviews with 8,885 women aged 15–49 years, with a 98.6% response rate. A stratified two-stage cluster sampling procedure was used to collect data from 9 regional states and 2 city administrations, with probability proportional to enumeration area (EA) size based on the 2019 primary healthcare frame and independent selection in each sample stratum. In the first stage, 305 EAs (93 urban and 212 rural) were chosen. In the second step of selection, a set number of 30 households per cluster was chosen by an equal probability systematic sampling from the newly produced household list. Finally, a nationwide sample of 8885 eligible women was interviewed. Women who had not given birth within the five years before the survey were excluded from this study. As a result, the analytic sample for the current study consists of 3,926 women who had at least one live birth in the past five years. Study variables Outcome variables. In this study, two binary outcome variables were considered. These are antenatal care (ANC) and place of delivery (PD). ANC= y e s = 1 , if a women attended four or more ANC 4 contacts/visits no = 0, otherwise Place of delivery= 1, if the woman gave birth at a health facility 0, if the woman gave birth at home Independent variables. The relevant risk factors associated with ANC visits and the place of delivery were included as individual-level variables: mother’s current age, mother’s educational level, family wealth index, religion, marital status, birth order, birth interval, number of children ever born, exposure to mass media, and community-level variables such as place of residence and region . Data management and analysis The data were extracted using STATA 18 software before being analyzed using SAS version 9.4 with PROC LOGISTIC and PROC GLIMMIX, which used the LAPLACE approximation. Descriptive statistics, such as frequencies and percentages, were utilized to summarize the individuals’ background information. Bivariate binary logistic regression Bivariate binary logistic regression is an extension of univariate logistic regression when there are two correlated categorical response variables, such as ANC and place of delivery. This approach examines the relationship between the two correlated categorical dependent variables and their associated independent variables. In this study, we focused on these two correlated categorical dependent variables, each of which has two categories. Let Y 1 and Y 2 are the two dependent variables such that ANC and place of delivery respectively, and each can have one of the two values (0 or 1) as described in . The best method for measuring the relationship between categorical variables in the logistic regression model was the odds ratio . shows the joint probability of the response variable and based on and , the random variables. Y 11 , Y 10 Y 01 and Y 00 follows the multinomial distribution with a joint probability function defined by P Y 11 = y 11 , Y 10 = y 10 , Y 01 = y 01 , Y 00 = y 00 = ∏ 0 1 ∏ 0 1 p gh y gh y gh , 0 < p gh < 1 where: g , h = 0 , 1 ; y g h = 0 , 1 ; Y 00 = 1 − y 11 − Y 10 − Y 01 ; p 00 = 1 − p 11 − p 10 − p 01 . Denote that ψ x The odds ratio shows an association between Y 1 and Y 2 depend on x which shows that Y 1 and Y 2 are correlated. The variable Y 1 and Y 2 are independent if ψ x = 1 . The odds ratio is defined by ψ x = p 11 x p 00 x p 10 x p 01 x , ψ x ≥ 0. Multilevel bivariate logistic regression Separate multilevel analyses of ANC and place of delivery among reproductive-age women have been conducted in various research papers . However, implementing a separate analysis would ignore the dependency between the ANC and the place of delivery. To take under consideration the correlation between the ANC and place of delivery and the estimates of effects of one or more covariates, the multilevel bivariate logistic regression model is a more plausible alternative. Multilevel bivariate binary logit models were used to account for the hierarchical nature of the data that affect the outcome variable and how the interactions among covariates measured at different levels (zonal administrative level) affect the outcome variable . As the EMDHS 2019 data was collected from women living in different zones in Ethiopia, the likelihood of having a clustering effect is very high. In this study, we considered a two-level hierarchical analysis where women are nested within zonal administration. The clustering effect was checked using the intra-class correlation (ICC) coefficient . When the logistic model is used the residual at level one (women level) is assumed to follow the standard logistic distribution and the variance. π 2 3 = 3.29 was assumed. The ICC can be calculated as: I C C = σ μ 0 2 σ μ 0 2 + π 2 3 , where σ μ 0 2 is the estimated cluster variance (zonal level). The multilevel model includes both fixed and random effects terms. The results of the fixed effects of the model were presented as an adjusted odds ratio (AOR) while the random effects were assessed with an intra-class correlation coefficient (ICC). In this study, four models were fitted with the null model (Model 0), which shows the variations of place of delivery and ANC in the absence of any independent variables, Model I as an adjusted for the individual-level variables, Model II as adjusted for the community-level variables and model III as adjusted for both individual and community-level variables. Akaike’s information criterion (AIC) and the Bayesian information criterion (BIC) determine the multilevel model that fits the data well and select the final model. Finally, adjusted odds ratios (AORs) were estimated and statistically significant predictors of ANC and place of delivery were identified at a p-value less than 0.05. Ethical review statement This study is based on Demographic and health survey (DHS) data and the DHS Program maintains standards for protecting the privacy of respondents and household members in all DHS surveys. Procedures and questionnaires for standard DHS surveys have been reviewed and approved by the International Coaching Federation (ICF) Institutional Review Board (IRB). While the host country IRB ensures the survey conforms with local laws and customs, the ICF IRB ensures it adheres to the U.S. Department of Health and Human Services’ requirements for protecting human subjects (45 CFR 46). So, country-specific DHS survey protocols are reviewed by the ICF IRB and typically by an IRB in the host country and the procedures can be found in the link: https://www.dhsprogram.com/methodology/Protecting-the-Privacy-of-DHS-Survey-Respondents.cfm . This study was conducted in Ethiopia, a country located in the horn of Africa and part of Sub-Saharan Africa, situated between 33° and 14° east longitude and 3° and 15° north latitude. A cross-sectional study design was employed using data from the 2019 Ethiopian Mini Demographic and Health Survey (EDHS), which was carried out by the Central Statistical Agency in collaboration with the Federal Ministry of Health and the Ethiopian Public Health Institute. Administratively, Ethiopia is divided into nine regional states and two city administrations, further subdivided into 68 zones . The dataset from the 2019 Ethiopian Mini Demographic and Health Survey (EMDHS) was used for this study. This survey was the second EMDHS and the fifth DHS conducted in Ethiopia, carried out from March 21, 2019, to June 28, 2019. We utilized the individual record (IR) dataset and extracted both the dependent and independent variables. The data set used in this study is freely available and possible to download by justifying the reason for requesting the data from the link: https://dhsprogram.com/data/available-datasets.cfm . All Ethiopian women (15–49 years old) in the reproductive age range are considered as the study’s population. Women in Ethiopia who had children within the previous five years of the survey for the most recent birth were involved in the study. The 2019 EMDHS was carried out among 8,663 households to ensure national representation. The survey included interviews with 8,885 women aged 15–49 years, with a 98.6% response rate. A stratified two-stage cluster sampling procedure was used to collect data from 9 regional states and 2 city administrations, with probability proportional to enumeration area (EA) size based on the 2019 primary healthcare frame and independent selection in each sample stratum. In the first stage, 305 EAs (93 urban and 212 rural) were chosen. In the second step of selection, a set number of 30 households per cluster was chosen by an equal probability systematic sampling from the newly produced household list. Finally, a nationwide sample of 8885 eligible women was interviewed. Women who had not given birth within the five years before the survey were excluded from this study. As a result, the analytic sample for the current study consists of 3,926 women who had at least one live birth in the past five years. Outcome variables. In this study, two binary outcome variables were considered. These are antenatal care (ANC) and place of delivery (PD). ANC= y e s = 1 , if a women attended four or more ANC 4 contacts/visits no = 0, otherwise Place of delivery= 1, if the woman gave birth at a health facility 0, if the woman gave birth at home Independent variables. The relevant risk factors associated with ANC visits and the place of delivery were included as individual-level variables: mother’s current age, mother’s educational level, family wealth index, religion, marital status, birth order, birth interval, number of children ever born, exposure to mass media, and community-level variables such as place of residence and region . Data management and analysis The data were extracted using STATA 18 software before being analyzed using SAS version 9.4 with PROC LOGISTIC and PROC GLIMMIX, which used the LAPLACE approximation. Descriptive statistics, such as frequencies and percentages, were utilized to summarize the individuals’ background information. In this study, two binary outcome variables were considered. These are antenatal care (ANC) and place of delivery (PD). ANC= y e s = 1 , if a women attended four or more ANC 4 contacts/visits no = 0, otherwise Place of delivery= 1, if the woman gave birth at a health facility 0, if the woman gave birth at home The relevant risk factors associated with ANC visits and the place of delivery were included as individual-level variables: mother’s current age, mother’s educational level, family wealth index, religion, marital status, birth order, birth interval, number of children ever born, exposure to mass media, and community-level variables such as place of residence and region . The data were extracted using STATA 18 software before being analyzed using SAS version 9.4 with PROC LOGISTIC and PROC GLIMMIX, which used the LAPLACE approximation. Descriptive statistics, such as frequencies and percentages, were utilized to summarize the individuals’ background information. Bivariate binary logistic regression is an extension of univariate logistic regression when there are two correlated categorical response variables, such as ANC and place of delivery. This approach examines the relationship between the two correlated categorical dependent variables and their associated independent variables. In this study, we focused on these two correlated categorical dependent variables, each of which has two categories. Let Y 1 and Y 2 are the two dependent variables such that ANC and place of delivery respectively, and each can have one of the two values (0 or 1) as described in . The best method for measuring the relationship between categorical variables in the logistic regression model was the odds ratio . shows the joint probability of the response variable and based on and , the random variables. Y 11 , Y 10 Y 01 and Y 00 follows the multinomial distribution with a joint probability function defined by P Y 11 = y 11 , Y 10 = y 10 , Y 01 = y 01 , Y 00 = y 00 = ∏ 0 1 ∏ 0 1 p gh y gh y gh , 0 < p gh < 1 where: g , h = 0 , 1 ; y g h = 0 , 1 ; Y 00 = 1 − y 11 − Y 10 − Y 01 ; p 00 = 1 − p 11 − p 10 − p 01 . Denote that ψ x The odds ratio shows an association between Y 1 and Y 2 depend on x which shows that Y 1 and Y 2 are correlated. The variable Y 1 and Y 2 are independent if ψ x = 1 . The odds ratio is defined by ψ x = p 11 x p 00 x p 10 x p 01 x , ψ x ≥ 0. Separate multilevel analyses of ANC and place of delivery among reproductive-age women have been conducted in various research papers . However, implementing a separate analysis would ignore the dependency between the ANC and the place of delivery. To take under consideration the correlation between the ANC and place of delivery and the estimates of effects of one or more covariates, the multilevel bivariate logistic regression model is a more plausible alternative. Multilevel bivariate binary logit models were used to account for the hierarchical nature of the data that affect the outcome variable and how the interactions among covariates measured at different levels (zonal administrative level) affect the outcome variable . As the EMDHS 2019 data was collected from women living in different zones in Ethiopia, the likelihood of having a clustering effect is very high. In this study, we considered a two-level hierarchical analysis where women are nested within zonal administration. The clustering effect was checked using the intra-class correlation (ICC) coefficient . When the logistic model is used the residual at level one (women level) is assumed to follow the standard logistic distribution and the variance. π 2 3 = 3.29 was assumed. The ICC can be calculated as: I C C = σ μ 0 2 σ μ 0 2 + π 2 3 , where σ μ 0 2 is the estimated cluster variance (zonal level). The multilevel model includes both fixed and random effects terms. The results of the fixed effects of the model were presented as an adjusted odds ratio (AOR) while the random effects were assessed with an intra-class correlation coefficient (ICC). In this study, four models were fitted with the null model (Model 0), which shows the variations of place of delivery and ANC in the absence of any independent variables, Model I as an adjusted for the individual-level variables, Model II as adjusted for the community-level variables and model III as adjusted for both individual and community-level variables. Akaike’s information criterion (AIC) and the Bayesian information criterion (BIC) determine the multilevel model that fits the data well and select the final model. Finally, adjusted odds ratios (AORs) were estimated and statistically significant predictors of ANC and place of delivery were identified at a p-value less than 0.05. This study is based on Demographic and health survey (DHS) data and the DHS Program maintains standards for protecting the privacy of respondents and household members in all DHS surveys. Procedures and questionnaires for standard DHS surveys have been reviewed and approved by the International Coaching Federation (ICF) Institutional Review Board (IRB). While the host country IRB ensures the survey conforms with local laws and customs, the ICF IRB ensures it adheres to the U.S. Department of Health and Human Services’ requirements for protecting human subjects (45 CFR 46). So, country-specific DHS survey protocols are reviewed by the ICF IRB and typically by an IRB in the host country and the procedures can be found in the link: https://www.dhsprogram.com/methodology/Protecting-the-Privacy-of-DHS-Survey-Respondents.cfm . In this study, a total weighted sample of 3,926 women of reproductive age was considered. As depicted in , from the total women who participated in this study, 2238 (57%) women didn’t utilize minimum WHO-recommended ANC contacts, and 1864 (47.5%) made their delivery at home which shows a low proportion of several antenatal care visit and a high proportion of home delivery, respectively. In addition, as seen from , among women of reproductive age, the majority of respondents 1192 (30.37%) were between the ages of 25 and 29 years. The majority of participants, 2900 (73.86%) lived in rural areas. Among the reproductive-age women, 2501 (63.69%) had no media exposure; of them, approximately 1664 (74.35%) had no ANC contacts and 1447(77.62%) had home delivery. shows the joint and marginal probability of ANC contacts and place of delivery for pregnant women and the odds ratio between ANC contact and place of delivery was 5.435 (OR = 5.435, 95% CI: 4.72–6.25), indicating a statistically significant association between ANC contacts and place of delivery. Additionally, the results showed that 422 women (10.8%) never had ANC contacts during their entire pregnancy period and delivered their children at home. shows the joint frequency of factors and different combinations of ANC utilization and place of delivery. The descriptive results of this study indicate that the highest proportion of both without ANC contact and home delivery was observed among mothers in the age group of 25–29 years, with a total of 423 women (66.55%). Among different residential areas, the majority of women have no ANC contacts and made home deliveries were found among mothers residing in rural areas 1254 (69.13%). Similarly, in terms of women’s educational level, the descriptive statistics revealed that three-fourths of the respondents had not received any ANC contacts and home delivery took place 1032 (75.79%) . , shows the estimated intra-class correlation (ICC) for ANC and place of delivery. The estimated value of ICC for the null model for ANC contacts and place of delivery is 42%, which shows a high zonal clustering/zonal variability. This indicates that there is a clustering effect and is strongly suggestive that there is within-group variability that would benefit from a cluster effect because of the cluster. We also made a model comparison for four models as detailed and we found that model 3 was the best model for the final estimation which fits the data well . From , we found that the estimated odds of having antenatal care visits among pregnant women in age groups 25–29, 30–34, and 35–39 were 1.880, 2.143, and 2.279 times more likely to have recommended ANC contacts than women age group 15–19 respectively, considering the other variable constant. The estimated odds of pregnant women in age groups 20-24, 25–29, 30–34, and 35–39 were 1.542, 1.916, 2.362, and 2.641 were 1.542, 1.916, 2.362 and 2.641 times more likely than women age groups 15-19 years, respectively. The estimated odds of minimum recommended ANC visits and health facility delivery for married women were 1.452 and 1.508 times more likely than unmarried women respectively. The estimated odds of women who attained primary, secondary, and higher levels of education were 1.763, 2.823, and 3.803 times more likely than women who did not have formal education respectively. The estimated odds of health facility delivery for women who attained primary, secondary, and higher levels of education were 2.095, 4.379, and 8.406 times more likely than women who did not attain education. The estimated odds of women with middle and rich wealth status were 1.255 and 1.810 times more likely to have a minimum recommended ANC contact as compared with poor economic status women, respectively. The estimated odds of women who live in rural areas are 0.612 and 0.352 for ANC utilization and place of delivery and this shows that the odds of minimum recommended ANC utilization and facility utilization are 0.612 and 0.352 times less likely than women who live in urban areas. The odds of women for facility delivery who lived in male-headed households were 0.641 and this indicates that the likelihood of having facility-based delivery is 0.641 times (AOR = 0.641; 95% CI: 0.505, 0.814) less likely to deliver in a health facility than women in female-headed households. The estimated odds of place of delivery among women who parity of 2-4 and 5 and above were 0.392 and 0.357 and this indicates that the odds of facility delivery is 0.392 times (AOR = 0.392; 95% CI: 0.247–0.623) and 0.357 times (AOR = 0.357; 95% CI: 0.249–0.510) less likely to deliver institutionally compared to women who had one child. The estimated odds ratio between ANC utilization and place of delivery was 6.381 [OR = 6.381, 95% CI: (5.518, 7.81)] indicating that there is a statistically significant association between the two outcome variables . This study aims to examine the association between minimum recommended ANC visits and place of delivery among reproductive-age women based on the bivariate multilevel logistic models in Ethiopia using the 2019 EMDHS. The model is designed to evaluate the dependency between minimum recommended ANC visits and place of delivery as well as to estimate the clustering effect given other covariates. From this study, we found that 57% of reproductive-aged women have no recommended ANC, and 47.5% of births were attended at home. The mother’s age, mother’s marital status, mother’s education level, family wealth index, sex of household head, and place of residence are important determinant factors that have a significant effect on recommended ANC utilization. The mother’s age, mother’s marital status, mother’s religion, child’s born, mother’s education level, wealth index, sex of household head, and place of residence are important determinant factors that have a significant effect on place of delivery. The current study showed that the age of mothers was a major factor in antenatal care (ANC) utilization. As the age of the mother rises, the likelihood of receiving antenatal care (ANC) Services also increases. This finding supports previous studies done in different countries that showed the positive association between ANC contacts and increased age of women. This might be because health conditions and birth complications are higher in older women who tend to demand more contact . This study also revealed that the education level of mothers was significantly associated with recommended antenatal care contacts and it shows that the odds of attending ANC visits among mothers who attended with primary, secondary, and higher educational levels were 1.763, 2.823, and 3.803 times the odds that women with no formal education. This finding is consistent with the studies conducted in Pakistan , East African countries and Ethiopia . This could be explained by the fact that mothers with higher levels of education are more likely to use antenatal care, have a better understanding of information, and are more aware of the necessity of the service . Furthermore, educated women are more likely to improve their independence, self-confidence, and ability to make health-related decisions for themselves. The wealth index of the household was found to be significantly associated with ANC utilization in this study. The study shows that mothers who reside in middle and rich households were 25.5% and 81.0% more likely to utilize recommended ANC compared to mothers who reside in poor households’ income, respectively. This result is consistent with the findings in Georgia , India and Ghana . This might be due to the difficulty that women in poor households face with out-of-pocket expenditure associated with the service. Women are responsible for transportation and other indirect cost to receive the service, even though the service is offered free of charge . In this context, wealthier women can manage transportation and work commitments, facilitating timely ANC contacts. In contrast, women with low economic status might not attend ANC due to fear of unexpected payments and prioritizing daily expenses over their health . In this study marital status affects ANC utilization and married women were more likely to attend ANC compared with their unmarried counterparts consistent with a study conducted in Ghana . This could be due to the psychosocial and financial support received from their husbands, planning/desirability of their pregnancy, and the societal acceptability and support of their pregnant state when compared with their unmarried counterparts Place of residence was another factor for the utilization of recommended ANC in Ethiopia. The study showed that women who lived in rural areas were less likely to get recommended ANC than urban resident women [AOR = 0.612, 95% CI = 0.485–0.772]. The place of residence was significantly associated with the usage of antenatal care services. This shows that women who were living in rural areas were less likely to receive recommended ANC services than those living in urban areas. This finding is consistent with the study conducted in Nigeria and Kenya . This might be because the health infrastructures in the rural area are less developed and there are fewer trained health workers to give information and education about recommended ANC. In addition, place of residence also significantly affects place of delivery in which rural women were 64.8% less likely to deliver at a health facility compared to the odds that urban women deliver at a health facility. This finding was supported by studies conducted in Kenya , Guinea , Pakistan , and Ethiopia . The possible justification might be women in an urban area easily get access to health knowledge, have financial accessing institutional/skilled personnel assistance, and have proximity to health facilities. This study showed that middle and high-wealth status increased the likelihood of institutional delivery. This is consistent with previous studies conducted in Ethiopia , and Bangladesh . This might be because the economic capability of the households and costs related to transport might influence the preference of the place of delivery. Besides, mothers from higher wealth status might be more likely to utilize maternal health services compared to others. Moreover, women with a higher educational status had a higher likelihood of institutional delivery than women who have had no formal education and the finding is consistent with previous studies conducted in Ethiopia and Tanzania . This study showed that mothers who reside in middle and rich wealth status households increased the likelihood of health facility delivery compared to mothers who reside in poor households. This is consistent with previous studies conducted in Ethiopia , Bangladesh and Pakistan . It could be reasoned that the economic capability of households and the costs related to transport might influence the preference for health facility deliveries. Besides, these mothers from better economic status might be pursuing maternal health services and the capacity to make decisions about deliveries from health facilities than others . In contrast, financial problem leads to poor maternal health care. The findings of this study also showed that the educational level of mothers significantly affects the place of women’s delivery. It revealed that women who attained primary, secondary, and higher education had a higher likelihood of health facilities delivery than women who had no education. This finding is consistent with previous studies in Ethiopia , and Tanzania . This might be due to the optimistic attitude of educated women that might enhance women’s self-determination; and the knowledge of the drawbacks of home delivery. The finding of this study also indicated that the mother’s age is an important determinant of the place of deliveries. The likelihood of institutional delivery increased as women’s ages increased beyond the age range of 15–19 years, which is consistent with a study in Ethiopia . The reason might be due to the young women’s fear of complications during home delivery, and it might be also due to maturity and understanding of the safety and other benefits of giving birth in healthcare facilities. As age increases, there will be increased knowledge and ability to make beneficial decisions regarding maternal health services. Furthermore, mothers who live in the Somalia region had lower odds of health facility delivery compared to the reference category (Tigray region), and this result is consistent with other studies in Ethiopia . The possible explanation might be the inaccessibility of health facilities in Somalia regions and people might have difficulty having permanent residency access to the services. The health facility delivery decreased with a high parity of women, which is consistent with the study conducted in Bangladesh , Ghana , and Pakistan . This might be the service quality given in previous births. Even if the Ethiopian health system has improved in the previous decade, still there were critical shortages of health personnel, supplies of drugs, and equipment. This could discourage women from utilizing health services in later pregnancies for delivery. Similarly, Women who lived in male-headed households were less likely to deliver in a health facility than women in female-headed households. The findings of this study are in agreement with those of a study conducted in Tanzania. The strengths of the study are data from a nationally representative population-based study with appropriate weighting, as well as useful, high-quality data on mothers, households, and communities. Furthermore, the study has a large sample size drawn at random across the country, allowing results to be generalized to women of reproductive age. Furthermore, when using binary logistic regression, the relationship between ANC visits and the place of delivery is ignored. The researcher employs a model that accounts for the data’s hierarchical structure. This statistical model is used to simultaneously simulate two binary outcome variables and assess their relationship to other predictors. The major limitation of this study was a cross-sectional survey, which may not help establish a temporal relationship between the possible risk factors of pregnant women’s ANC visits and place of delivery. Moreover, the data was self-reported; there might also be a possibility of recall and social desirability biases that will result in underreporting and misreporting of events. This study aims to examine the association between ANC contacts and place of delivery among reproductive-age women based on the bivariate multilevel logistic models in Ethiopia using the 2019 Ethiopian min demographic and health survey data. The model applied in this study is designed to evaluate the effects of risk factors on ANC and place of delivery as well as to estimate the clustering effect. In this study, out of 3926 reproductive-age women, 57% of women had below the minimum recommended ANC contacts and 47.5% of the total women made their delivery at home. Of the total women considered in the study, about 36.7% of them delivered their new child at home and they didn’t utilize the recommended ANC during their pregnancy period. Only 32.2% of the women have made both facility-based deliveries and got the minimum WHO-recommended ANC contact during their pregnancy. Women’s age, women education level, marital status, wealth status, sex of household head, residence, and region were significant predictors of antenatal care and delivery care utilization simultaneously in Ethiopia. It would be useful to increase financial support strategies that enable pregnant women from poor households to use health services and enhance pregnant women’s understanding of the significance of recommended ANC and institutional delivery through health education targeting women with their level of education. Emphasis should also be placed on supporting unmarried pregnant women to have recommended ANC and institutional delivery. Ministry of Health, health facility professionals, and community health workers have an important task in raising mothers’ attitudes to ANC utilization and institutional delivery. S1 File Antenatal and delivery care weighted. (DTA) |
Medical interns’ training in family medicine at a district hospital and primary health care clinics in Middelburg, Mpumalanga | 6136350d-86f6-4347-9712-4b1f27d1706d | 11151463 | Family Medicine[mh] | For many years, Middelburg District Hospital in Mpumalanga serves as the clinical learning centre for the training of the students of health professions under the aegis of the University of Pretoria, Family Medicine Department. Family medicine in Middelburg is the only specialist unit in the hospital and also is the main clinical unit for the Mpumalanga clinical training centres. The unit has been training Bachelor of Clinical and Medical Practice (BCMP) (Clinical associates) students since 2010, in house programme. The unit also accommodates elective students from most universities in South Africa and a few students from abroad. There is a consistent regular elective from the physician’s assistants’ students from the University of Wisconsin in the United States (US). This has been ongoing since 2016 with the exception of 2020 and 2021 because of the coronavirus disease 2019 (COVID-19) restrictions. Middelburg clinical learning centre is also accredited for the in-house training of the registrars in family medicine. Five specialists have graduated from this site; they obtained their Master’s degrees in family medicine as well as the Fellowship from the College of Family Physicians (South Africa). Some of them are providing clinical leadership at the hospital and in the province of Mpumalanga. Since the implementation of the new medical interns training programme, Middelburg hospital and Witbank Hospital are considered as one training complex for family medicine in the Nkangala district of Mpumalanga. The interns spent 4 months in Witbank and 2 months in Middelburg. The latter allows for more exposure to the district health system through practice at the district hospital and the primary health care clinics. The latest assessment by the Health Professions Council of South Africa (HPCSA’s) interns training committee in August 2023, judged the family medicine training programme as excellent. The clinical learning centre’s experience and involvement in training of other categories of health professionals have made a seamless integration of the medical interns in its activities. The family physicians and other clinicians at the hospital have long time ties with the local University of Mpumalanga , in the prevention and management of non-communicable and communicable diseases. The 2 months’ rotation at the district hospital is embedded in the 6 months of family medicine training where the majority (4 months) of the time is spent in the sister department at Witbank regional/tertiary hospital. Clear guidelines are laid down by the HPCSA for the family medicine training in general and the district rotation in particular. The domain of family medicine offers the interns the opportunity to manage a wide spectrum of conditions in undifferentiated patients. Among other conditions are communicable and non-communicable chronic diseases, palliative care, acute and non-acute conditions, preventive medicine, clinical forensic medicine, an so forth, the list is non-exhaustive. Also created are collaboration and interaction with other primary care health workers such as clinical associates, allied professionals, pharmacists and nurses. Ultimately the medical intern is able to contribute to the management of patients who present at any primary care facility, for example primary health clinic, community health centre or the district hospital. In addition to assessing and managing patients at these levels, the interns are capacitated to identify patients whose conditions require a higher level of care and appropriately refer, in consultation with the supervisor. In case a patient is admitted to the district hospital for the treatment, the admitting intern follows up the patient’s progress until discharge, to ensure continuity of care. The same principle applies in the primary health care facilities where the medical interns see patients with acute or chronic conditions in a continuum of longitudinal care until the rotation ends. Local circumstances dictate the length of stay at the primary health care (PHC) clinic or the work in the district hospital. This report is based on the interaction the family physicians, the medical officers and the medical interns have had during the year 2021 and 2022 at Middelburg district hospital and the surrounding primary health care clinics. The following are the activities undertaken with each group of medical interns. Orientation On the first day, medical interns are orientated on the functioning and organisation of services at the district hospital and PHC clinics and their respective roles in the health system of South Africa. The interns are also introduced to the nursing units’ managers, the allied professionals, the pharmacists and other administration staff they may need help from during their rotation in the district hospital. Clinical exposure and learning The district hospital is a composite structure where the medical intern is exposed to all types of patients in different clinical units: emergency room, operating theatre, maternity, child health and general outpatient. Outreach services to PHC clinics are organised on a weekly basis for each intern. After hours’ work is mainly delivered in the emergency room, also called casualty unit. From time to time, the intern may be called to operating theatres to be an assistant for emergency surgical procedures (mainly obstetrical). Supervision and collaboration While the interns work under the supervision of senior medical officers and family physicians, they also work side by side with the clinical associates and the nursing personnel as well as the other elective students of healthcare sciences. To streamline the monitoring and evaluation of the work done by the medical interns during the 2 months, we have developed a few working tools using the 2020 guidelines set by the HPCSA regarding the domain of family medicine and primary care. The developed tools were made to be user-friendly; therefore, we framed them as checklists (copies available on request from the corresponding author). The following checklists are introduced to the intern at the beginning of the rotation, and they are monitored and evaluated throughout the rotation and also at the completion of their time at the district hospital: Daily output to report on the clinical work and exposure to skills during working hours and after hours as applicable. Hand over and clinical discussion participation on a daily basis. Continuing professional development (CPD) presentations on identified topics as per the list from the HPCSA. Signature of attendance registers at each academic and clinical meeting. Overall global formative assessment by the supervisor(s) using the intern’s logbook. Where appropriate, the data is also entered in the interns’ logbook provided by the HPCSA. On the first day, medical interns are orientated on the functioning and organisation of services at the district hospital and PHC clinics and their respective roles in the health system of South Africa. The interns are also introduced to the nursing units’ managers, the allied professionals, the pharmacists and other administration staff they may need help from during their rotation in the district hospital. The district hospital is a composite structure where the medical intern is exposed to all types of patients in different clinical units: emergency room, operating theatre, maternity, child health and general outpatient. Outreach services to PHC clinics are organised on a weekly basis for each intern. After hours’ work is mainly delivered in the emergency room, also called casualty unit. From time to time, the intern may be called to operating theatres to be an assistant for emergency surgical procedures (mainly obstetrical). While the interns work under the supervision of senior medical officers and family physicians, they also work side by side with the clinical associates and the nursing personnel as well as the other elective students of healthcare sciences. To streamline the monitoring and evaluation of the work done by the medical interns during the 2 months, we have developed a few working tools using the 2020 guidelines set by the HPCSA regarding the domain of family medicine and primary care. The developed tools were made to be user-friendly; therefore, we framed them as checklists (copies available on request from the corresponding author). The following checklists are introduced to the intern at the beginning of the rotation, and they are monitored and evaluated throughout the rotation and also at the completion of their time at the district hospital: Daily output to report on the clinical work and exposure to skills during working hours and after hours as applicable. Hand over and clinical discussion participation on a daily basis. Continuing professional development (CPD) presentations on identified topics as per the list from the HPCSA. Signature of attendance registers at each academic and clinical meeting. Overall global formative assessment by the supervisor(s) using the intern’s logbook. Where appropriate, the data is also entered in the interns’ logbook provided by the HPCSA. During the 12 months of 2021, the cumulative total of 27 medical interns rotated through Middelburg hospital for a 2 months’ district hospital and primary healthcare training. Almost the same number in 2022 where 29 medical interns were received in Middelburg. Exposure and responsibilities of the medical interns in the district Each intern had a logbook supplied by the HPCSA for the recording of the activities during this rotation. Additional to the logbook, the five checklists mentioned above were used to establish and evaluate work done by the interns under our supervision. Daily outputs The medical intern’s bulk of work in the district is mainly in the ambulatory care, primary health clinics, outpatients and the emergency room. The interns also have the opportunity to follow up patients they admitted to the wards for in-hospital treatment or for a short period of observation; this ensures the continuity of care. Admitted conditions vary from trauma-related injuries, interpersonal violence, complications of non-communicable diseases like hypertension, diabetes, asthma and epilepsy (see ). Human immunodeficiency virus (HIV)-related complications are also admitted. Sepsis, malnutrition and neonatal jaundice were found to be prominent in child health. The interns also documented high incidence of pregnancy-related complications especially bleeding in the first and the last trimesters. There are instances where interns took part in surgical operations as assistants to the surgeon (usually a senior medical officer or a family physician), mainly caesarean sections. In a few instances, the medial interns have assisted to the open reduction and internal fixation of fractures done by the visiting orthopaedic surgeons. Fast-forward to 2022 The year 2022 saw many of the health restrictions related to the COVID-19 pandemic being relaxed. The medical interns that started this year had more activities recorded than the previous ones, specifically in trauma-related care. It also included activities in line with community oriented primary care programme (COPC). The COPC activities were initiated by the family physicians with the support of the managers of the Hospital and PHC facilities. Similar initiatives have been reported by Singaram et al. A good balance is observed between the clinical exposure, learning and participation in the continuing professional development (see ). The latter includes monthly activities such as participation at the Perinatal Problems Identification Programme (PPIP), Child Problems Identification Programme (Child PIP) and other morbidity and mortality meetings. Interns are also encouraged to make presentations on topics of interest taking into account the diseases’ epidemiology of the area. A daily morning meeting takes place where interns and supervisors reflect on the previous day’s activities and also planning for the day ahead. A sizeable amount of learning takes place in these meetings. Interns have been exposed to the broad spectrum of patients and conditions. Common conditions seen by interns reflected the burden of diseases in South Africa, mainly consisting of trauma-related injuries, adult communicable chronic illnesses, adult non-communicable chronic illnesses, maternal health issues including complications of pregnancy and childhood diarrhoeal diseases. Similar findings have been reported by Ross et al. Regarding theatre exposure, the medical interns did mostly attend to the gynaecological and obstetrical procedures; the most common operations were caesarean sections followed by laparotomies for ectopic pregnancy. Miller made similar observations elsewhere. Each intern had a logbook supplied by the HPCSA for the recording of the activities during this rotation. Additional to the logbook, the five checklists mentioned above were used to establish and evaluate work done by the interns under our supervision. The medical intern’s bulk of work in the district is mainly in the ambulatory care, primary health clinics, outpatients and the emergency room. The interns also have the opportunity to follow up patients they admitted to the wards for in-hospital treatment or for a short period of observation; this ensures the continuity of care. Admitted conditions vary from trauma-related injuries, interpersonal violence, complications of non-communicable diseases like hypertension, diabetes, asthma and epilepsy (see ). Human immunodeficiency virus (HIV)-related complications are also admitted. Sepsis, malnutrition and neonatal jaundice were found to be prominent in child health. The interns also documented high incidence of pregnancy-related complications especially bleeding in the first and the last trimesters. There are instances where interns took part in surgical operations as assistants to the surgeon (usually a senior medical officer or a family physician), mainly caesarean sections. In a few instances, the medial interns have assisted to the open reduction and internal fixation of fractures done by the visiting orthopaedic surgeons. The year 2022 saw many of the health restrictions related to the COVID-19 pandemic being relaxed. The medical interns that started this year had more activities recorded than the previous ones, specifically in trauma-related care. It also included activities in line with community oriented primary care programme (COPC). The COPC activities were initiated by the family physicians with the support of the managers of the Hospital and PHC facilities. Similar initiatives have been reported by Singaram et al. A good balance is observed between the clinical exposure, learning and participation in the continuing professional development (see ). The latter includes monthly activities such as participation at the Perinatal Problems Identification Programme (PPIP), Child Problems Identification Programme (Child PIP) and other morbidity and mortality meetings. Interns are also encouraged to make presentations on topics of interest taking into account the diseases’ epidemiology of the area. A daily morning meeting takes place where interns and supervisors reflect on the previous day’s activities and also planning for the day ahead. A sizeable amount of learning takes place in these meetings. Interns have been exposed to the broad spectrum of patients and conditions. Common conditions seen by interns reflected the burden of diseases in South Africa, mainly consisting of trauma-related injuries, adult communicable chronic illnesses, adult non-communicable chronic illnesses, maternal health issues including complications of pregnancy and childhood diarrhoeal diseases. Similar findings have been reported by Ross et al. Regarding theatre exposure, the medical interns did mostly attend to the gynaecological and obstetrical procedures; the most common operations were caesarean sections followed by laparotomies for ectopic pregnancy. Miller made similar observations elsewhere. Medical interns’ training in the district hospital plays a pivotal role in that it provides an adequate platform for unselected patients and a variety of medical conditions. Both enhance learning and continuing professional development. The allocated period of 2 months for this rotation may not be enough to build confidence in many areas of the district health system. A recommendation to increase this period to 3 months would address this shortcoming. This report is a reflection from only one district hospital regarding a 2 months’ rotations in family medicine. It may not represent the situation in other hospitals in the province or in the country. Interns’ rotation in family medicine at the district hospital contributes to the strengthening of the district health systems especially supporting primary healthcare services in underserved areas. The rotation proved to develop competent clinicians who should contribute to the improvement of the population health and their wellbeing. Managers of health programmes and health facilities are called upon to support this rotation because it benefits the interns and the community at large. |
Photodocumentation in oculoplastic surgery: an up-to-date
overview | 45663a00-c65b-4d99-b5ec-dee3b681294b | 11826790 | Ophthalmology[mh] | The equipment used was as follows: a digital single lens reflex (DSLR) camera model EOS Rebel T6 (Canon, Inc., Tokyo, Japan) with an Advanced Photo System Type C (APS-C) sensor measuring 22.3 mm x 14.9 mm, EF 100 mm f/2.8 Macro USM Lens (Canon, Inc., Tokyo, Japan), a dedicated Canon Speedlite 430EX III-RT (Canon, Inc., Tokyo, Japan) with white diffuser reflector, and a tripod. The images were taken in the medical office of one of the authors (Barbi, JSF), whose roof is painted white and the walls light gray. For full face and periorbital region composition, the camera was used in manual (M) mode, with f/8 aperture, 1/125 shutter speed, ISO 100, dedicated flash used in the through-the-lens (TTL) automatic mode with 75-degree tilt, and white diffuser reflector. The white balance (WB) was set to “flash” mode, and the focus automatically centered in the patient´s eyes. In two particular situations, settings were modified: in cases of eyelid ptosis, images were taken using the flash from the front, directed toward the patient’s face and diaphragm closing by 1 stop (f/11), and in cases of macro photographs, the flash was directed backward toward a silver diffuser reflector. The patient was seated on a swivel stool without wheels. A floor mat with markings was used to signalize the front (anteroposterior), oblique (45 o ), and profile (90 o ) positions. The camera was positioned on a tripod, and the height was adjusted so that the camera lens was aligned with the patient’s eyes. The external flash was directed at the ceiling and a white diffuser reflector attached to the body of the flash . For full face-framing in the primary position gaze (PPG), the camera’s thirds grid was used so that the upper horizontal line passed through the pupils and the apex of the patient’s ears. For framing the face in the oblique position, the patient was requested to rotate the entire body to 45 ° until the tip of the nose aligned with the malar eminence, and the ear apex aligned with the lateral eyelid canthus. For lateral view framing, it is important to align the ear apex to the lateral eyelid canthus so that only the same side of Cupid’s bow can be seen . With an APS-C sensor camera and a 100-mm macro lens, the photographer’s was 3 m away from the patient. For periorbital framing, the working distance was 1 m. To ensure and standardize this distance, the patient and photographer were placed on marks on the floor. For macro photographs (e.g., eyelid tumors), the camera was set to manual focus, using a 0.38-m focus distance and 1:1 magnification ratio. In this setting, the photographer had to approach or move away from the lesion to be photographed to obtain a sharp focus. At the start of each set of photographs, the patient was asked to hold an 18% gray card next to his/her face, and a picture was taken for WB correction, followed by the full set of photographs. Photographs were taken in RAW format and processed in postproduction using Adobe Lightroom ® . All patients signed consent form for the use of their photographs. The included articles showed some common aspects: 10 of 19 articles suggested the use of a telelens (60-110 mm), and only 3 of them described the working distance . Background might vary between blue, white, black, and gray tones; however, in more than half of the included articles, the blue background was cited as preferential or alternative(1-3,6-13). Most of the articles(1-3,6,7,9,10,13-15) have suggested illumination with studio lights, four suggested speedlight only (ring flash and dedicated TTL flashes or unspecified speedlight , and two did not mention the illumination source. Only four articles mentioned the three parameters used: ISO, aperture, and shutter speed . More than half of the included articles did not mention these parameters . Most of the included articles cited the Frankfurt horizontal plane for the head position and positioning angles such as frontal, lateral, and oblique views . shows the results of the protocol adopted in this study for photographic documentation in oculoplasty. Illustrative photographs are shown to demonstrate the result of using this protocol. We considered special angles of view and some particularities in oculoplastic photography. In pre-and post-blepharoplasty or eyelid surgery photographs, we took photographs in the PPG, right and left oblique, and lateral views . In cases of ptosis, we photographed the patient in PPG, supraversion, and infraversion. We also used the frontal flash to demonstrate the margin reflex distance, as this reference is the main comparison parameter for the position of the upper eyelid in the postoperative period . We recorded the results of the 10% phenylephrine test , when performed, by placing a mark (white eye pencil or piece of white micropore) above the ipsilateral eyebrow where the drop was instilled. This way, when reviewing the photographs, it was clear which is the before photograph and the after-the-test photograph. In orbit photography, we recorded all gazes for the assessment and documentation of motility/restrictions of extraocular muscles. In addition, to demonstrate the anterior projection of the eyeball and proptosis, we took photographs with the camera tilted up and located below the patient’s chin, and we asked the patient to raise the chin . In the treatment of dynamic wrinkles and expression lines, we asked the patient to contract the facial muscle groups to be treated, as shown in , in the preand post-botulinum toxin treatment. We took macrophotographs with an external flash bent backward and bounced on a sheet of polystyrene or a white or silver diffuser reflector, as shown in . By using the protocol described in this article, we collected data and photographs from 324 patients and obtained consistent results with optimal standardization of facial photographs before and after the surgical/cosmetic procedures. There are three types of photographers: amateurs, professionals, and functional . Physicians who practice oculoplastic photography fall into the category of “functional photographers,” i.e., those who are not professional photographers but need to have the minimum basic knowledge of photographic recording for their medical practice. Photodocumentation is valuable for various purposes, such as medical record keeping, insurance and legal situations, creation of models for preoperative planning and clarifying it to the patient, assessment for self-improvement, and medical education, including teaching residents, sharing data with colleagues, and preparing presentations and publications . Notably, even though the 19 articles included proposed standardization of face photographs, only two reported information of all the characteristics analyzed. Guided by the literature review and a self-developed standardization protocol used by one of the authors (Barbi, JSF), we proposed a protocol for oculoplastic photodocumentation, which focuses on basic technical knowledge on photography, camera parameters, illumination, background, head position, and image size. DSLR cameras were recommended in several articles for their excellent cost-benefit ratio. They can offer the benefits of interchangeable lenses; thus, one can select the appropriate focal length, have complete control over camera settings, and obtain good image quality. Full-frame (24 mm x 35 mm) sensors are larger than APS-C sensors and can produce better image quality and less digital noise. However, full-frame cameras are generally more expensive, and APS-C sensor cameras are good enough for clinical photography. The focal length is quoted in millimeters. With an APS-C sensor camera, lenses are regarded as “wide angle” when the focal length is <35 mm. This type of lenses delivers a wider field of view and is ideal for panoramic and landscape photography. The so-called “normal” lenses provide a viewing angle similar to that of the human eye, which corresponds to 35-mm lenses in cameras with APS-C sensor. Telephoto lenses provide a smaller viewing angle and magnification of the image and are ideal for face and close-up photographs . Such lenses avoid the face distortions that can occur with the approach of normal or wide-angle lenses . The lenses can be fixed focal length (prime lenses) or zoom lenses with adjustable focal length. Fixed lenses are those with a single focal length, and to change the frame, the photographer must move back and forth. Given the same objective, zoom lenses allow several focal lengths that can be adjusted manually, there is no need for the photographer to move just to change the frame. In medical practice, fixed lenses are recommended for two reasons: they allow better image quality and easier standardization. The authors of this article recommend a 100-mm macro lens, according to personal experience and published articles . The authors recommend using this type of lenses because they allow greater distance between the photographer and the patient, which is favorable for lighting, as it bounces back, has a longer path to reach the patient, and delivers a softer and broader illumination. A dedicated macro lens is preferred, as it allows closer focusing on a 1:1 magnification ratio of the patient . This feature is advantageous for photographing small lesions at large magnification. Cameras interact with light by basically three components in the camera settings, the so-called light photographic triangle: diaphragm or aperture (f-number), shutter speed, and ISO . In an easy-to-understand and didactic way, we can compare these parameters with the physiology of the eye. The diaphragm of a camera can be compared to the pupil; with a larger aperture, more light enters and the shallower the depth of field and vice versa. The shutter speed could be compared with the act of “blinking” It works as a window that allows light to enter the eye depending on how long it is open or how fast it closes. The ISO is a measure of the sensitivity of the sensor to light, equivalent to the retina, and can be adjusted according to environmental brightness. In a very bright environment, the ISO should be reduced and vice versa. These three parameters can be adjusted automatically, semi-automatically, or manually on DSLR cameras. For correct standardization, the photographer should always choose the manual mode and determine these values for consistency in the photographs before and after surgical or cosmetic treatments. A camera’s aperture is quantified by the “f” number, which is a ratio of the lens focal length and the diameter of the aperture. The aperture of the diaphragm in a 100-mm f/2.8 macro lens ranges from 2.8 to 22; the smaller the value, the more open the diaphragm will be and the smaller the depth of field. In medical practice, it is recommended to work with apertures between f/5.6 and f/8 for two reasons: (1) to have sufficient depth of field to make the whole face of the patient sharp (very wide diaphragms can leave the eyes focused and the tip of the nose and ears blurred), and (2) in a 100-mm macro lens, these “f” values allow for better image quality (in very open or very closed diaphragms, there is a slight loss in image quality). The shutter speed is a measure of how long the camera’s shutter blades are open to expose the sensor to the light, and it is measured in seconds or fractions of a second. In the present study, the shutter speed of the camera ranges from 30 s (slowest) to 1/4000 seconds (fastest). When the flash is attached to the camera shoe, this command is limited to a maximum speed of 1/125 (corresponding to the maximum synchronization speed of the flash). At very high speeds, the second shutter curtain may close before the light hits the entire sensor, which generates a dark band on the photographs. To prevent this, the camera limits fast shutter speeds when the flash is attached. Since the ambient light in an ophthalmology outpatient clinic is not consistent in brightness and color temperature , the maximum flash sync speed should be used to exclude ambient light interference in the photographs. Thus, regardless of the time of the day, sunlight in windows, and ceiling lights (on or off), the photograph will have light consistency because the camera will record the light coming only from the flash. For the same reason mentioned above, the photographer should work with lower ISO values, and 100 is suggested. The ISO represents the sensitivity of the sensor to light, and lower values are desired so that there is minimal or no capture of ambient light. Although most articles encourage the use of studio lights , the authors suggest using a unique light source (TTL speedlight) (1) to simplify the photography equipment as studio lights may not be viable in terms of space in ophthalmology offices and 2) because a speedlight works in very effectively without requiring additional lights for full face and periorbital region photographs. In addition to the photographic triangle, the power level of the flash is the fourth element during flashed photographs. This parameter is set on the flash itself and can be placed in manual or automatic mode. This study agrees with Ong et al. , who suggested the use of the flash in automatic mode, that is, the camera will determine the power of the flash using the TTL metering function. In a TTL system, the camera fires a quick pre-flash to determine the amount of light needed to illuminate the subject followed by the correct flash output to achieve correct exposure. If photographs are always taken at the same working distance using the same lens, the photographs will have good consistency in this parameter even when working with the flash in automatic mode. However, as most of the oculoplastic surgeons are not using a studio, factors, such as furniture or patient clothing color, may cause small changes in automatic reading. To avoid handling the flash in manual mode, the authors suggest using an 18% gray card in the first photo in a series. The card helps with the WB in the editing software in postproduction, as it serves as a neutral color reference for adjustment of the color temperature of the photographs . Another important suggestion is the use of the flash directed toward the ceiling, thus using the bounced light to achieve smoother and broader lighting and greater idea of three-dimensionality. The frontal flash results in harsh light, which leaves the photo “flat”, that is, without any portrayal of relief and contours that are fundamental for medical photodocumentation . In addition, a white diffuser reflector attached to the flash is also recommended to direct part of the light in a straight line, which will reach the patient’s face much more smoothly, as it is bounced, not direct light. This reflector is used to eliminate shadows that can occur in the periorbital region when the lighting is exclusively done from “top to bottom” when using only light reflection from the ceiling. This is useful in men who have a very prominent forehead. These “shadows” in the periocular region can be eliminated or smoothened. This this problem can be also solved by the use of a small rectangular reflecting panel positioned horizontally against the patient’s chest, just under the collarbone, outside the framing, just to reflect the light from bottom to top, minimizing these shadows . For macrophotographs, the current literature suggests the use of a ring or twin flash, specific for this type of photographs. However, as this study aimed to simplify the photographic apparatus without loss of quality and consistency of images, the authors used the external flash aimed at a white reflector (which can be something as simple as a white card) or a 30-cm silver reflector located behind the photographer. We chose this method because the photographer-patient distance for macrophotography is approximately 0.38 m, which is very close. The use of the front flash will cause the flattening of lesions, removing the same important characteristics as shadows, and raised edges. Reflecting the light in the ceiling is not also a good option, as in macrophotography, the area framed reflecting light is very small and needs greater illumination than the light reflected from the ceiling, leaving the image dark. Directing the flash backward into a reflector allows the light to smoothly illuminate the lesions, without erasing its contour and relief, which should be demonstrated in the preoperative record or clinical follow-up. WB is a process of removing the unnatural color cast from an image . Different light sources can have different color temperatures. Fluorescent lamps can result in greenish images, just as incandescent sources can result in orange skin tones . In medical photography using the proposed protocol, since there will always be a flash, it is recommended to leave the WB in “flash” mode. As previously said, if the flash perhaps “misses” the light reading of the scene, the WB can be corrected by using an 18% gray card in the editing software. The need for a uniform background is well established, but the color suggested for this background varies between authors. Several articles of medical photography have suggested the use of a blue background because it is opposed to the yellow shade of the skin of most patients, generating a composition pleasant to the human eye. Some authors suggest a black background to eliminate shadows that could be generated by a white background, depending on the incident light. However, a black background may present challenges in patients with dark hair and cannot provide subjectbackground separation unless another source of light is used, making office photography more complex . The simplest and cheapest way is to paint the wall with a color close to 18% gray based on established photography books . This tone is intermediate between white and absolute black, and it helps with the photometer readings built in DSLR cameras. In addition, due to its “absence of color”, it prevents unwanted reflections of color by the flash on the patient’s face. The suggestion was to paint the office ceiling white, as it will be used to bounce the light from the dedicated flash. Moreover, matte paint should be used to avoid over-reflection of the light . The recommended positions for full-face photographs are anteroposterior, right anterior oblique, left anterior oblique, right profile, and left profile(5,6,8,9,11-16,18,26,28,30). Discrete marks on the floor are also essential to determine the distance between the photographer and the patient . The use of the thirds grid on the camera’s screen is highly recommended to help the photographer align the camera precisely . The authors agree with the statements of Rhee regarding a frame reference, i.e., a horizontal line that passes through the center of the pupil or eyelid canthus and the apex of the ear, because this reference can be used in frontal, oblique, and profile pictures. More than half of the included articles used Frankfurt’s horizontal plane (a line that passes through the external auditory canal and the inferior orbital rim) as a reference for head positioning . Nevertheless, it is a radiological reference, and it can be difficult to reproduce in photography. To frame the face in the oblique position, the patient rotated his/her entire body, facing markings on the floor corresponding to 45 ° on the right and then on the left . These markings correspond to the patient rotating his/her body until the patient’s tip of the nose aligns with the malar eminence and the ear apex with the lateral eyelid canthus. For lateral views, the ear apex should be aligned to the lateral eyelid canthus and only the ipsilateral Cupid’s bow should be seen. Although photographs of the face are recommended in a vertical orientation , the authors recommend the horizontal orientation for the ease of operation with the camera and flash. In orbital diseases, the apparent exophthalmos and enophthalmos must be documented. Thus, the authors recommend the “worm view” which is taken with the patient in neck extension and the camera viewing from below; the tip of the nose aligns with the glabella, right in the middle of the eyebrows, the focus is kept in the eyes, and both eyes appear in the frame. Regarding the format of the photographs, the suggestion was to shoot in RAW format , which is a type of file without processing in which color information, WB, and other parameters will be edited later in postproduction using the Adobe Lightroom® for editing, converting RAW to JPEG files (without loss of quality since the files will not be compressed), and storing them in the cloud, which allows for device synchronization and therefore easy sharing or exchange of files. A narrative review on photodocumentation in facial surgery was performed, and a self-developed, guided by the literature review, standardized protocol for facial photographic registration was described. Preand postphotographs of surgical and cosmetic procedures were collected from 324 patients using this protocol, and consistent results were obtained with optimal standardization of facial photographs. This protocol can be easily adapted to any oculoplastic surgeon’s practice, without the need to set up a studio in the office or spend on unnecessary extra photographic equipment. Essentially, there is not a gold standard protocol in facial photodocumentation. However, there is agreement on the importance of standardization and reduction in variables as much as possible to achieve consistency across photographs and register the patient’s condition and clinical evolution as accurately as possible. |
Peri‐operative identification and management of patients with unhealthy alcohol intake | b1534e5e-5dc7-4f46-9271-4d940ff29fb9 | 11825216 | Surgical Procedures, Operative[mh] | There are no other nationally or internationally agreed guidelines for screening, identification and standardised management of patients with excessive alcohol consumption who are having elective and emergency surgical procedures. Accepting that the focus of this consensus statement is on the peri‐operative management of patients with excessive alcohol consumption, we have also drawn on evidence related to hepatic disease and its anaesthetic implications. Patients with unhealthy alcohol intake represent a high‐risk surgical cohort, having an increased probability of peri‐operative complications, mortality and long‐term functional decline following surgical procedures. Many of the problems in this group of patients can be minimised by timely identification and pre‐emptive management, thereby improving outcomes and potentially reducing healthcare expenditure. In developing this document, we have incorporated elements from national guidelines relating to non‐alcoholic liver disease and alcohol issues in the non‐surgical population. These resources are primarily from the National Institute for Health and Care Excellence (NICE) that describe relevant interventions but do not frame these within a peri‐operative context. Patients should be screened during pre‐assessment for surgery using an appropriate tool such as the alcohol use disorders identification test consumption (AUDIT) questionnaire to triage risk relating to alcohol use. Those patients with harmful chronic alcohol intake or who are judged to be at high risk of alcohol‐related liver disease should have careful pre‐operative physical examination and additional relevant investigations. Harmful intake is suggested by an AUDIT score > 19 and/or consumption of > 35 units of alcohol per week (women) or > 50 units of alcohol per week (men). In pre‐operative patient populations, elevated AUDIT questionnaire scores should prompt a proportional response ranging from brief intervention to inpatient specialist referral. It should be borne in mind that patients who have previously demonstrated harmful alcohol consumption may still be at high risk of alcohol‐related liver disease irrespective of current screening scores. Risk assessment and decision‐making in established alcohol‐related liver disease should be guided via the use of validated scoring systems such as the Surgical Outcome Risk Tool (SORT). The peri‐operative clinician managing harmful alcohol intake should consider surgical urgency according to recognised classification systems. It may be appropriate to delay some surgical procedures to allow interventions that reduce the risk of alcohol‐related complications. Even emergency surgical populations may benefit from screening for alcohol‐related risk and commensurate intervention to mitigate this. Patients with chronic harmful alcohol intake may be appropriate for day‐case surgery. Potential barriers to suitability are organ dysfunction, social circumstances and concerns relating to immediate recommencement of alcohol consumption following day‐case discharge. These should be carefully investigated and discussed. Analgesic techniques that reduce the requirements for opioids, including regional methods if appropriate, should be considered. Patients with cirrhosis and alcoholic hepatitis are at elevated risk of complications when opioids are used. Specialist assistance from acute pain services should be considered, while the prescription of other commonly used analgesics may need to be modified or avoided altogether in these cases. Patients with harmful alcohol intake are at high risk of postoperative complications such as infections, arrhythmias, bleeding and delirium. The development of these complications should be actively monitored with due consideration for critical care admission. At‐risk patients should undergo regular peri‐operative assessment for alcohol withdrawal. Interventions should be guided by objective scoring systems, such as the Clinical Institute Withdrawal Assessment Alcohol—Revised (CIWA‐Ar) scale, within a defined care pathway. Patients with underlying liver cirrhosis and/or who are older are at risk of benzodiazepine‐related complications; short‐acting drugs should be employed to manage withdrawal in these cases. Elevated or harmful alcohol consumption is common in England. In 2019, 30% of men and 15% of women were documented as regularly drinking more than the recommended safe weekly limits for alcohol consumption . Such data are proportionally reflected in the 2% of all 2019–2020 NHS admissions for which alcohol was a primary reason for inpatient stay . Chronic harmful alcohol intake causes a spectrum of organ dysfunction of which alcohol‐related liver disease (including cirrhosis) remains the most common cause of death . This heterogeneity reflects the myriad systemic effects that may complicate management from both chronic alcohol intake and acute intoxication respectively, with a range of clinical implications (Fig. ) . These issues justify proactive identification and treatment of such patients, who may have clinically occult, yet potentially impactful, liver disease. A structured, multidisciplinary approach is appropriate throughout the peri‐operative period, with due modification according to surgical priority (Figs. and ) . We aimed to produce a multidisciplinary consensus statement directed by a working party with a diverse authorship who were invited based on their clinical and/or academic expertise in the area. Recommendations were formulated using a modified Delphi process. An initial list of recommendations was produced following targeted literature reviews for all relevant phases of patient care throughout the peri‐operative pathway. These recommendations were distributed among the authors who rated each as ‘include’, ‘exclude’ or ‘revise’, as well as providing anonymised comments onto a Microsoft Excel spreadsheet (Microsoft Inc., Redmond, WA, USA). Recommendations with ≥ 75% inclusion decision were included. Assessment Current literature recommends that all patients are screened for alcohol intake as part of routine practice . In the peri‐operative setting, this aligns with a recent target from the Commissioning for Innovation and Quality (CQUIN) framework advising that 80% of patients with a hospital stay of one night should be screened for alcohol use using a validated screening tool . Of the available tools, the abbreviated alcohol use disorders identification test – consumption (AUDIT‐C) (Fig. ) combines high sensitivity and specificity with rapid use . It comprises three self‐reported questions highlighting at‐risk patients (score ≥ 5), who would benefit from comprehensive assessment of alcohol consumption by a healthcare professional using seven further questions from the full abbreviated alcohol use disorders identification test (AUDIT) questionnaire (Fig. ) . Patients who score < 5 on the abbreviated questionnaire or < 8 on comprehensive assessment, are deemed at lower risk and can continue on the normal patient pathway. Patients who score 8–19 on comprehensive AUDIT assessment are at increased risk of alcohol‐related liver disease. In addition, a score of > 19 is consistent with harmful drinking and may indicate alcohol dependence. This is also suggested by consumption of > 35 units of alcohol per week (female sex) or > 50 units of alcohol per week (male sex). If there is a significant history of harmful drinking in the past in terms of units consumed per week, then patients may potentially still be high risk regardless of the current score on the questionnaire. Patients with alcohol‐related liver disease should be evaluated using established scoring systems to guide peri‐operative risk quantification and shared decision‐making. Externally validated risk calculators such as P‐POSSUM and SORT may be corroborated with disease‐specific systems such the Child‐Turcotte‐Pugh score or surgical risk metrics such as the Mayo Risk Score or the VOCAL‐Penn score . The VOCAL‐Penn score has been externally validated in patients with cirrhosis and found to have superior discrimination compared with other related tools ; it can be calculated online at http://www.vocalpennscore.com . Physical examination In addition to a standard assessment, at‐risk patients should be examined for stigmata of chronic liver disease. In particular, assessment for ascites or asterixis (as a sign of hepatic encephalopathy) should be undertaken as these represent decompensation of cirrhosis which may require specific management. Hypertension is common with alcohol excess, although patients with cirrhosis may have low blood pressure due to hyperdynamic circulation and reduced systemic vascular resistance. It should be noted that findings in liver disease may range from normal to overt decompensation depending on the severity of the disease . When considering regional techniques, it is prudent to document pre‐existing weakness or neurological abnormalities as this may be relevant in the assessment of post‐procedural complications such as nerve injury. Investigations A full blood count is useful since anaemia (defined in the peri‐operative setting as haemoglobin < 130 g.l ‐1 ) from a variety of causes is common in harmful drinking. Although this does not invariably represent a haematinic issue, macrocytic anaemia with a raised mean corpuscular volume may indicate folate/vitamin B12 deficiency or bone marrow toxicity . Iron deficiency may predominate, with reduced ferritin or transferrin/iron saturation levels; however, it should be remembered that ferritin is often artefactually elevated in patients with raised alcohol consumption. Reciprocally, the absence of microcytosis does not mean the patient will not be iron deficient as peripheral erythrocytes may exist in various stages of development. The red cell distribution width may be helpful in these cases as a diagnostic adjunct. Thrombocytopaenia may be present secondary to bone marrow suppression, folate deficiency or hypersplenism, possibly reflecting underlying portal hypertension. Leucocytosis may also be present due to hepatitis‐related leukaemoid reaction . As discussed, severity of liver disease may be assessed by the Child‐Turcotte‐Pugh score. Key components of this are serum albumin; serum bilirubin; and prothrombin time. Additionally, elevated (conjugated) bilirubin may suggest active alcoholic steatohepatitis, even in the absence of cirrhosis. Electrolyte disturbances, particularly hyponatraemia, are common with alcohol excess even in the absence of advanced or cirrhotic liver disease. Hypokalaemia and hypophosphataemia may cause muscle weakness, while hypomagnesaemia may worsen hypokalaemia and may cause seizures . Urea and creatinine can be elevated by pre‐renal causes of acute kidney injury more commonly seen in patients with cirrhosis, while hepatorenal syndrome only occurs in the setting of ascites or significant portal hypertension. Isolated uraemia may be due to active gastrointestinal bleeding . Biochemical markers for alcohol‐related liver disease have been employed historically to assess liver dysfunction; however, abnormal liver enzymes are poorly predictive of underlying liver disease with only 3.9% of patients with an abnormal value in this domain shown to develop significant liver disease within 5 years of testing . Liver enzyme values should, therefore, be interpreted in the context of wider clinical findings and other test results. Harmful alcohol intake increases the risk of cardiovascular disease ranging from hypertension and atrial arrhythmias to more serious problems such as cardiomyopathy. The incidence of stroke (haemorrhagic or ischaemic) is also increased . Diagnosis can be aided by using a 12‐lead ECG, while echocardiography can show structural damage to the heart caused by alcohol misuse such as a dilated left ventricle with decreased mass and wall thickness or systolic dysfunction. Echocardiography should be reserved for those patients with recognised criteria such as those displaying signs and symptoms of cardiac failure . In selected cases, cardiopulmonary exercise testing may be of value to identify systolic dysfunction that is masked by the hyperdynamic circulation of liver impairment . Other more technical investigations to identify a fatty or cirrhotic liver such as ultrasound, computed tomography or liver biopsy are best organised by a gastroenterologist after appropriate referral . The use of transient elastography to measure liver stiffness (e.g. Fibroscan®, Echosens, France) in patients who are alcohol‐dependent has gained particular diagnostic prominence as reflected by recent commissioning data and national guidelines . Depending on local availability, some institutions may employ alternatives to elastographic testing, such as the enhanced liver fibrosis test , in the assessment of hepatic fibrosis risk. Establishing the presence of underlying advanced fibrosis/cirrhosis is important due to its impact on relevant peri‐operative outcomes and the risk of hepatic decompensation. Additionally, transient elastography is now used to stratify the risk of complications from portal hypertension; in this context liver stiffness < 20 kPa and platelet count > 150 × 10 9 .l ‐1 are accepted thresholds to rule out the presence of large oesophageal varices in compensated cirrhosis . Harmful alcohol use and older patients Age‐related changes in organ function, combined with comorbidity and depleted physiologic reserve, render older patients particularly vulnerable to the effects of chronic excessive alcohol use and some related management interventions such as long‐acting benzodiazepines. This is compounded by the increasing frequency and volume of alcohol consumption observed in populations aged 65–74 y. Harmful alcohol intake is an independent risk factor for the development of frailty and is strongly associated with dementia, falls and delirium . Additional risk factors for harmful chronic alcohol intake may be observed in patients > 65 y including: male sex; single marital status; social isolation; insomnia; depression; dementia; chronic pain; and substance availability . Intervention In the community In elective settings, safe pre‐operative reduction in alcohol intake remains a core element of management. The level of intervention to achieve this is influenced by the degree of patient risk as determined by objective scoring or weekly alcohol intake. Pre‐operative assessment represents an ideal opportunity to make a brief intervention (a ‘teachable moment’) among those patients scoring 8–19 on the AUDIT score to discuss the potential harm from excessive alcohol intake and provide advice leaflets. Increased risk scores should prompt a recommendation that patients self‐refer to community alcohol support services . Clinicians should have lower intervention thresholds for patients with additional risk factors such as female sex, patients aged < 18 or > 65 y, and those from Black and minority ethnic backgrounds . Patients scoring > 19 on comprehensive AUDIT have possible alcohol dependency and should be referred for community‐based assisted withdrawal . This requires close liaison with local alcohol support services and primary care to ensure that it is carried out in a timely manner before admission for surgery. Those individuals who are judged to be vulnerable or are aged ≤ 16 y may need to be referred as an inpatient for specialist support and possible inpatient alcohol withdrawal by hospital alcohol support services (or gastroenterology teams where these are not available). A lower threshold for controlled in‐hospital detoxification may also be applied for cases where community management is superseded by operative urgency. Local service availability, expertise and referral pathways may vary such that a community withdrawal programme may be preferred rather than inpatient admission in some institutions. Ownership from named, contactable clinicians or services from the outset is important to promote continuity of care in all settings. Parent teams should review patient progress regularly and intervene as needed to ensure effective, safe management. Finally, the refusal of a competent patient to engage in discussed interventions should not be treated as an absolute contraindication to surgery provided there has been an appropriate discussion about related risks on a case‐by‐case basis. Inpatients Inpatient specialist referral may be warranted for patients admitted with harmful drinking or alcohol‐dependence who are experiencing acute withdrawal or at high risk of this (Box ) . This includes urgent/emergency cases and patients admitted for planned detoxification. Close liaison aids early awareness of any potential issues but also facilitates assessment of community withdrawal success and guides the decision for further support. Box 1 Clinical features of acute alcohol withdrawal syndrome. Adapted from Alcohol withdrawal syndrome Acute alcohol withdrawal syndrome describes a spectrum of clinical sequelae typically occurring after abrupt cessation of alcohol consumption in patients with chronic, harmful intake. It has been attributed to impaired neurotransmitter regulation in the central nervous system, whereby chronically elevated cerebral ethanol concentrations decrease sensitivity to inhibitory gamma‐amino butyric acid (GABA) and interfere with glutamate receptor binding, causing compensatory upregulation of excitatory N‐methyl‐D‐Aspartate (NMDA) receptors. The sudden absence of ethanol in this context results in a pro‐excitatory imbalance that causes many of the observed symptoms and signs during withdrawal states. Minor symptoms typically occur within 6 h and include insomnia; tremor; anxiety; gastrointestinal upset; headache; diaphoresis; and palpitations. More severe features, described below, may occur in some patients. Withdrawal seizures and delirium tremens Generalised tonic–clonic seizures may occur, most commonly in patients aged 40–50 years with prolonged harmful intake. Seizure activity is most common 12–48 h following cessation of alcohol intake. If untreated, one third of alcoholic seizures may progress to delirium tremens, characterised by hypertension; tachycardia; confusion; hallucinations; agitation; and diaphoresis. It is distinct from isolated alcoholic hallucinations, which are not usually accompanied by confusion or deranged vital signs. With appropriate recognition and management, the mortality from delirium tremens is now < 5%. Metabolic abnormalities Patients may be hypovolaemic and demonstrate a metabolic acidosis, frequently accompanied by electrolyte abnormalities such as hypokalaemia; hypophosphataemia; and hypomagnesaemia. Deranged metabolic findings in this group are often multifactorial and interrelated, and themselves may result in organ dysfunction and an increased risk of cardiac failure, arrhythmias, reduced seizure threshold and mortality. On admission to hospital, the revised Clinical Institute Withdrawal Assessment – Alcohol (CIWA‐Ar) score should be measured in at‐risk patients . This is a 10‐point scale (Fig. ) used to assess alcohol withdrawal that can be used to direct pharmacological interventions; benzodiazepines such as chlordiazepoxide are a mainstay of therapy, while alternatives include carbamazepine and clomethiazole . The required interval between medication doses will vary but is often every 90 min with administration of a longer‐acting benzodiazepine, such as diazepam 20 mg, triggered at a CIWA‐Ar score of > 11 . This is repeated until there are three consecutive CIWA‐Ar scores < 11, indicating complete detoxification. The patient should then continue with standard vital and neurological observations unless there is a re‐emergence of symptoms. This strategy may warrant modification in established liver cirrhosis due to the concurrent risk of encephalopathy; in this setting the exclusive use of short‐acting benzodiazepines such as oxazepam is warranted. This may also apply to other groups who can have greater sensitivity to sedative drugs, such as patients who are older. Seizures are a common sequel of harmful alcohol intake, particularly in the context of withdrawal. Other differentials, such as hypoglycaemia or electrolyte deficiencies, should be considered as part of management. Alcohol‐related seizures occurring as an inpatient should be dealt with according to established guidelines . Benzodiazepines are a mainstay of therapy; phenytoin should be avoided . Following a seizure the patient should continue, or recommence, CIWA‐Ar scoring as appropriate. Type of surgery Day‐case surgery There are few published data relating to day surgery in patients with harmful alcohol intake. However, day‐case procedures are potentially advantageous in this group as they can avoid the withdrawal states associated with inpatient hospital stays. Many general aspects of management are encompassed within pre‐existing guidelines . At all stages, appropriate involvement of the multidisciplinary team is vital for decision‐making regarding day‐case suitability and timing of other elements, such as controlled withdrawal programmes. Case‐by‐case decision making may vary depending on individual patient characteristics, local resources and personnel, particularly when planning safe postoperative discharge. In high‐volume, nurse‐led pre‐assessment settings, the addition of the AUDIT‐C to other commonly used scoring systems provides a valuable, rapidly applicable measure for seeking further clinician involvement such as discussion with an anaesthetist. Day surgery‐specific pre‐operative assessment should otherwise be undertaken by practitioners competent in eliciting symptoms and signs indicative of complications from chronic alcohol use. Cardiovascular and hepatic manifestations such as dilated cardiomyopathy and cirrhosis are particularly relevant to day‐case suitability. The presence of coagulopathy or thrombocytopenia, with their attendant bleeding risks, should also influence decision‐making. It may be safer for some patients to be managed as inpatients, depending on the nature of the planned surgical intervention and outcomes of appropriate multidisciplinary discussions. As for inpatient surgery, consideration should be given to deferring surgery for planned community withdrawal under appropriate supervision for all high‐risk patients. Potential benefits from abstinence for at least 2 weeks include improved platelet function and therefore reduced bleeding risk, while cessation of alcohol intake for 8 weeks or longer may improve wound healing . Patients may refuse to engage with a pre‐operative intervention as well as advice for the postoperative period, such as non‐consumption of alcohol for 24 h. This may affect the patients' appropriateness for day‐case procedures and should carefully be explored, with appropriate risk/benefit discussions and development of an individualised management plan in consultation with the wider team. Patients who present acutely intoxicated on the day of surgery should have surgery postponed with appropriate follow up. There may be a particular role for regional techniques, including peripheral nerve blockade if feasible and safe, to facilitate same‐day discharge while potentially reducing home analgesic requirements. The use of short‐acting anaesthetic drugs provides an additional margin of safety by minimising the risk of pharmacological interaction with alcohol consumed after surgery. Caution should be exercised when prescribing take‐home postoperative systemic drugs where there is evidence of alcohol‐related liver disease (Table ). All patients with harmful alcohol intake should have a clearly documented plan for safe discharge following day surgery. Consideration should be given to social circumstances in high‐risk patients, given the potential for home alcohol withdrawal and the interaction of alcohol consumed at home with peri‐operative medications. Such patients may not be suitable for ‘home alone’ discharge as described elsewhere and should return home with a responsible adult . Nurse‐led discharge on the day of surgery should otherwise follow explicit criteria according to local protocols, with additional vigilance for signs of withdrawal in the recovery area. Provision of written patient discharge material, with details of surgical procedure, follow‐up planning and key contacts such as community alcohol liaison services as relevant, is critical for ongoing patient safety. Emergency surgery It may not be practical or safe to implement some of the pre‐operative measures described above, such as controlled alcohol withdrawal, in patients who require time‐critical or emergency surgery. However, it should be noted that acute presentation does not preclude judicious pre‐operative intervention, depending on clinical context. Furthermore, the identification of patients with excessive alcohol intake also influences acute management within the operating theatre and after surgery (Fig. ). If feasible, screening patients rapidly with the AUDIT‐C questionnaire during anaesthetic assessment is therefore important for prompt recognition, early involvement of other key specialities and tailoring of subsequent therapy. Optimisation General considerations Consent should include details of the additional risks of alcohol withdrawal and postoperative complications. Patients who are intoxicated, encephalopathic or have chronic alcohol‐related brain damage may not have sufficient mental capacity to provide consent specific to the intended procedure. Management should incorporate the patient's best interests and, if appropriate, an attempt at substituted judgement regarding their wishes. These elements should align with the principles of the Mental Capacity Act (2005) in keeping with existing guidance published by the Association of Anaesthetists . Patients with cirrhosis should be discussed with gastroenterology or hepatology specialists where possible. The severity of cirrhosis correlates with the risk of postoperative mortality and a careful history should be taken for features of hepatic decompensation. The risk calculators discussed previously may assist with planning for these cases. However, such tools should be treated as adjuncts to clinical decision‐making; it may be most appropriate to refer complex and/or decompensated patients to a liver centre where there is immediately available subspecialty expertise. In the elective setting, appropriately timed optimisation for this cohort may include treatment of ascites; screening and prophylaxis of varices; treatment of hepatic encephalopathy; and nutritional intervention. Balanced correction of nutritional, metabolic, fluid and haematological abnormalities will require multidisciplinary input, with referral to relevant specialist teams for organ‐specific complications. The detailed peri‐operative management of deranged physiology in liver disease is covered in a recent publication . Issues most relevant to patients with harmful alcohol intake are discussed below. Nutritional issues The direct effects of harmful alcohol intake may compound with secondary liver disease to profoundly disrupt energy and nutritional homeostasis. Resultant malnutrition is common, occurring in 60–85% of patients with cirrhosis and encompassing a spectrum of conditions including sarcopenia; frailty; electrolyte abnormalities; and vitamin deficiencies. Underlying mechanisms are multiple and interdependent, including appetite loss; intestinal mucosal/microbiome disruption with resultant malabsorption; and hypercatabolic states secondary to oxidative alcohol metabolism and systemic, low grade bacterial translocation . Nutritional status should therefore be considered in all patients with harmful alcohol intake. There is no gold standard tool for screening this population for malnutrition. The Malnutrition Universal Screening Tool is recommended by the European Society for Clinical Nutrition and Metabolism, although BMI calculations may be confounded by the presence of ascites. Alternative screening tools adapted for liver disease, such as the Royal Free Hospital Nutritional Prioritisation Tool, may present a useful alternative if there is diagnostic uncertainty . Although detailed discussion regarding management is beyond the scope of this document, general considerations include energy requirements; protein intake; electrolyte replacement; vitamin supplementation; and micronutrients. Associated hepatic complications, such as alcoholic hepatitis, ascites or encephalopathy may alter individual needs in these domains . Harmful alcohol intake is also an independent risk factor for the development of refeeding syndrome , marked by potentially life‐threatening fluid and electrolyte shifts precipitated by over‐supplementation in susceptible individuals. Appropriate involvement of a dietician for complex cases is advised throughout the peri‐operative period. It should also be noted that infusions of ferric carboxymaltose (Ferinject®, Vifor Pharma UK Ltd, Staines, UK) for iron replacement may present a particular risk of hypophosphataemia for some patients in this group due to pre‐existing phosphate depletion/concurrent refeeding syndrome. Regular thiamine‐containing supplements, such as Pabrinex® (Kyowa Kirin Ltd, Galashiels, UK), should be prescribed at admission to patients admitted with chronic harmful alcohol consumption to prevent Wernicke's encephalopathy, a serious complication of alcohol misuse comprising ophthalmoplegia, ataxia and acute delirium precipitated by vitamin B deficiency. Regimens may vary by institution but would typically include intravenous therapy for 72 h followed by conversion to oral supplementation (Fig. ). Haematological issues The treatment of coagulation abnormalities in patients with cirrhosis before surgery has become increasingly complex. For low‐risk procedures, routine administration of blood products such as fresh frozen plasma to achieve a specific laboratory coagulation value has little benefit and may cause harm. Where bleeding is considered likely, such as in those undergoing major surgery, it is reasonable to target specific endpoints such as a platelet count of > 50 × 10 9 cells.l ‐1 or a fibrinogen concentration of > 1 g.l ‐1 in decompensated disease. Appropriate haematological involvement, supplemented by investigations such as viscoelastic testing, is critical for decision‐making in complex cases . The literature on the efficacy and safety of peri‐operative tranexamic acid in patients with chronic harmful alcohol intake and associated liver disease is incomplete. Recent recommendations from NICE advise routine prophylaxis in particular circumstances, in keeping with several large meta‐analyses . This contrasts with a recent, large multicentre randomised controlled trial indicating that tranexamic acid use in acute upper gastrointestinal haemorrhage, a known complication of cirrhosis, does not reduce 5‐day mortality from bleeding and may slightly increase the risk of venous thromboembolism and seizures . The authors highlighted the divergent findings in this study compared with a preceding Cochrane review, attributing this in part to the potential for erroneous results in meta‐analyses of small studies . Literature examining tranexamic acid use for the management of bleeding in other emergency settings predominantly consists of single‐centre studies and has not demonstrated an increased thrombotic risk, although high doses may increase the risk of seizures . These studies do not specifically focus on patients with harmful alcohol intake or liver disease and the true risk/benefit of antifibrinolytic therapy in this group therefore remains controversial. In the absence of definitive evidence or a specific contraindication, it remains reasonable to give tranexamic acid for the prevention or management of acute haemorrhage in most circumstances. Avoidance of high doses is advised where seizure risk is elevated, such as acute withdrawal states or when patients are at risk of this. Peri‐operative management Intra‐operative The systemic effects of harmful alcohol intake, in combination with other potential issues such as acute illness and secondary organ dysfunction, may warrant modification of anaesthesia techniques (Table ). It should be noted that there are few outcome data supporting neuraxial techniques over general anaesthesia in patients with established liver disease . Given the challenges encountered in patients with harmful alcohol consumption, it may be more appropriate to consider performing surgery using regional or local anaesthetic techniques. Potential intra‐ and postoperative benefits include avoidance of the cardiorespiratory and neurological effects of general anaesthesia in a high‐risk cohort; enhanced seizure detection in patients who are awake; and mitigation of pharmacological effects of systemic medications. Due consideration should be given to factors influencing the suitability of regional techniques in these patients, including appropriateness of techniques for surgical procedures and their duration; reduced patient cooperation or reduction in mental capacity due to intoxication or chronic neurological impairment; presence of coagulopathy and/or thrombocytopenia; and the influence of hypoproteinaemia (secondary to malnutrition and/or hepatic synthetic dysfunction) on local anaesthetic distribution and dosing limits. Pharmacological issues Chronic harmful alcohol intake and related organ dysfunction, as well as acute intoxication, may significantly influence the efficacy and safety of commonly used medications, with a commensurate requirement to modify or avoid their use. Where there is alcohol‐related secondary organ dysfunction, judicious administration of intra‐operative anaesthetic drugs is warranted (Table ). This reflects vulnerability to the cardiorespiratory effects of such medications, as well as the widely recognised changes in drug pharmacokinetics seen in alcohol‐related liver disease. Postoperative Unhealthy alcohol intake contributes significantly to postoperative complications, in particular alcohol withdrawal syndrome, postoperative infections and delirium . This leads to an increased duration of hospital stay, ICU admissions and mortality . This picture is exacerbated by the presence of related comorbidities, decompensation of which may precipitate hepatic encephalopathy, cardiac ischaemia and arrhythmias . Patients should therefore be monitored to detect complications, with a low threshold for postoperative admission to critical care. The decision to refer for critical care support should be guided using a peri‐operative risk calculator, such as P‐POSSUM, which should be appropriately repeated at the end of surgery to account for determinant intra‐procedural events such as major haemorrhage. It should be noted that liver‐specific scoring systems outlined earlier have primary utility in pre‐operative discussion and referral; they are not suitable for dynamic assessment of risk and determination of postoperative location for patients undergoing procedures. The incidence of alcohol withdrawal syndrome is estimated to be 2–5 times higher in surgical patients compared with other inpatients and has a higher associated morbidity . Features range from minor symptoms such as headache, tremor, and insomnia, to more serious manifestations including withdrawal seizures, hallucinations and delirium tremens. Approximately 5% of patients who undergo alcohol withdrawal suffer from delirium tremens and mortality is significant if left untreated . At‐risk patients should be monitored using the described CIWA‐Ar scale or equivalent, with ongoing consideration for thiamine supplementation . Where available, continued oversight from alcohol support services is advised. Patients with established alcohol‐related liver disease are at increased risk of postoperative morbidity and mortality, a picture complicated by the potential impact of both hepatic dysfunction and continued harmful alcohol intake on postoperative analgesia (Table ) . Hepatic encephalopathy can be precipitated by hypoxia, hypovolaemia or acid–base/electrolyte disturbance, which should be avoided during and after surgery . Infections may also provoke hepatic decompensation, warranting vigilance and aggressive treatment when present. Patients with cirrhosis are also at high risk of renal dysfunction, necessitating caution when using nephrotoxic agents and judicious fluid balance, aiming to preserve renal perfusion while avoiding volume overload. Current literature recommends that all patients are screened for alcohol intake as part of routine practice . In the peri‐operative setting, this aligns with a recent target from the Commissioning for Innovation and Quality (CQUIN) framework advising that 80% of patients with a hospital stay of one night should be screened for alcohol use using a validated screening tool . Of the available tools, the abbreviated alcohol use disorders identification test – consumption (AUDIT‐C) (Fig. ) combines high sensitivity and specificity with rapid use . It comprises three self‐reported questions highlighting at‐risk patients (score ≥ 5), who would benefit from comprehensive assessment of alcohol consumption by a healthcare professional using seven further questions from the full abbreviated alcohol use disorders identification test (AUDIT) questionnaire (Fig. ) . Patients who score < 5 on the abbreviated questionnaire or < 8 on comprehensive assessment, are deemed at lower risk and can continue on the normal patient pathway. Patients who score 8–19 on comprehensive AUDIT assessment are at increased risk of alcohol‐related liver disease. In addition, a score of > 19 is consistent with harmful drinking and may indicate alcohol dependence. This is also suggested by consumption of > 35 units of alcohol per week (female sex) or > 50 units of alcohol per week (male sex). If there is a significant history of harmful drinking in the past in terms of units consumed per week, then patients may potentially still be high risk regardless of the current score on the questionnaire. Patients with alcohol‐related liver disease should be evaluated using established scoring systems to guide peri‐operative risk quantification and shared decision‐making. Externally validated risk calculators such as P‐POSSUM and SORT may be corroborated with disease‐specific systems such the Child‐Turcotte‐Pugh score or surgical risk metrics such as the Mayo Risk Score or the VOCAL‐Penn score . The VOCAL‐Penn score has been externally validated in patients with cirrhosis and found to have superior discrimination compared with other related tools ; it can be calculated online at http://www.vocalpennscore.com . In addition to a standard assessment, at‐risk patients should be examined for stigmata of chronic liver disease. In particular, assessment for ascites or asterixis (as a sign of hepatic encephalopathy) should be undertaken as these represent decompensation of cirrhosis which may require specific management. Hypertension is common with alcohol excess, although patients with cirrhosis may have low blood pressure due to hyperdynamic circulation and reduced systemic vascular resistance. It should be noted that findings in liver disease may range from normal to overt decompensation depending on the severity of the disease . When considering regional techniques, it is prudent to document pre‐existing weakness or neurological abnormalities as this may be relevant in the assessment of post‐procedural complications such as nerve injury. A full blood count is useful since anaemia (defined in the peri‐operative setting as haemoglobin < 130 g.l ‐1 ) from a variety of causes is common in harmful drinking. Although this does not invariably represent a haematinic issue, macrocytic anaemia with a raised mean corpuscular volume may indicate folate/vitamin B12 deficiency or bone marrow toxicity . Iron deficiency may predominate, with reduced ferritin or transferrin/iron saturation levels; however, it should be remembered that ferritin is often artefactually elevated in patients with raised alcohol consumption. Reciprocally, the absence of microcytosis does not mean the patient will not be iron deficient as peripheral erythrocytes may exist in various stages of development. The red cell distribution width may be helpful in these cases as a diagnostic adjunct. Thrombocytopaenia may be present secondary to bone marrow suppression, folate deficiency or hypersplenism, possibly reflecting underlying portal hypertension. Leucocytosis may also be present due to hepatitis‐related leukaemoid reaction . As discussed, severity of liver disease may be assessed by the Child‐Turcotte‐Pugh score. Key components of this are serum albumin; serum bilirubin; and prothrombin time. Additionally, elevated (conjugated) bilirubin may suggest active alcoholic steatohepatitis, even in the absence of cirrhosis. Electrolyte disturbances, particularly hyponatraemia, are common with alcohol excess even in the absence of advanced or cirrhotic liver disease. Hypokalaemia and hypophosphataemia may cause muscle weakness, while hypomagnesaemia may worsen hypokalaemia and may cause seizures . Urea and creatinine can be elevated by pre‐renal causes of acute kidney injury more commonly seen in patients with cirrhosis, while hepatorenal syndrome only occurs in the setting of ascites or significant portal hypertension. Isolated uraemia may be due to active gastrointestinal bleeding . Biochemical markers for alcohol‐related liver disease have been employed historically to assess liver dysfunction; however, abnormal liver enzymes are poorly predictive of underlying liver disease with only 3.9% of patients with an abnormal value in this domain shown to develop significant liver disease within 5 years of testing . Liver enzyme values should, therefore, be interpreted in the context of wider clinical findings and other test results. Harmful alcohol intake increases the risk of cardiovascular disease ranging from hypertension and atrial arrhythmias to more serious problems such as cardiomyopathy. The incidence of stroke (haemorrhagic or ischaemic) is also increased . Diagnosis can be aided by using a 12‐lead ECG, while echocardiography can show structural damage to the heart caused by alcohol misuse such as a dilated left ventricle with decreased mass and wall thickness or systolic dysfunction. Echocardiography should be reserved for those patients with recognised criteria such as those displaying signs and symptoms of cardiac failure . In selected cases, cardiopulmonary exercise testing may be of value to identify systolic dysfunction that is masked by the hyperdynamic circulation of liver impairment . Other more technical investigations to identify a fatty or cirrhotic liver such as ultrasound, computed tomography or liver biopsy are best organised by a gastroenterologist after appropriate referral . The use of transient elastography to measure liver stiffness (e.g. Fibroscan®, Echosens, France) in patients who are alcohol‐dependent has gained particular diagnostic prominence as reflected by recent commissioning data and national guidelines . Depending on local availability, some institutions may employ alternatives to elastographic testing, such as the enhanced liver fibrosis test , in the assessment of hepatic fibrosis risk. Establishing the presence of underlying advanced fibrosis/cirrhosis is important due to its impact on relevant peri‐operative outcomes and the risk of hepatic decompensation. Additionally, transient elastography is now used to stratify the risk of complications from portal hypertension; in this context liver stiffness < 20 kPa and platelet count > 150 × 10 9 .l ‐1 are accepted thresholds to rule out the presence of large oesophageal varices in compensated cirrhosis . Age‐related changes in organ function, combined with comorbidity and depleted physiologic reserve, render older patients particularly vulnerable to the effects of chronic excessive alcohol use and some related management interventions such as long‐acting benzodiazepines. This is compounded by the increasing frequency and volume of alcohol consumption observed in populations aged 65–74 y. Harmful alcohol intake is an independent risk factor for the development of frailty and is strongly associated with dementia, falls and delirium . Additional risk factors for harmful chronic alcohol intake may be observed in patients > 65 y including: male sex; single marital status; social isolation; insomnia; depression; dementia; chronic pain; and substance availability . In the community In elective settings, safe pre‐operative reduction in alcohol intake remains a core element of management. The level of intervention to achieve this is influenced by the degree of patient risk as determined by objective scoring or weekly alcohol intake. Pre‐operative assessment represents an ideal opportunity to make a brief intervention (a ‘teachable moment’) among those patients scoring 8–19 on the AUDIT score to discuss the potential harm from excessive alcohol intake and provide advice leaflets. Increased risk scores should prompt a recommendation that patients self‐refer to community alcohol support services . Clinicians should have lower intervention thresholds for patients with additional risk factors such as female sex, patients aged < 18 or > 65 y, and those from Black and minority ethnic backgrounds . Patients scoring > 19 on comprehensive AUDIT have possible alcohol dependency and should be referred for community‐based assisted withdrawal . This requires close liaison with local alcohol support services and primary care to ensure that it is carried out in a timely manner before admission for surgery. Those individuals who are judged to be vulnerable or are aged ≤ 16 y may need to be referred as an inpatient for specialist support and possible inpatient alcohol withdrawal by hospital alcohol support services (or gastroenterology teams where these are not available). A lower threshold for controlled in‐hospital detoxification may also be applied for cases where community management is superseded by operative urgency. Local service availability, expertise and referral pathways may vary such that a community withdrawal programme may be preferred rather than inpatient admission in some institutions. Ownership from named, contactable clinicians or services from the outset is important to promote continuity of care in all settings. Parent teams should review patient progress regularly and intervene as needed to ensure effective, safe management. Finally, the refusal of a competent patient to engage in discussed interventions should not be treated as an absolute contraindication to surgery provided there has been an appropriate discussion about related risks on a case‐by‐case basis. Inpatients Inpatient specialist referral may be warranted for patients admitted with harmful drinking or alcohol‐dependence who are experiencing acute withdrawal or at high risk of this (Box ) . This includes urgent/emergency cases and patients admitted for planned detoxification. Close liaison aids early awareness of any potential issues but also facilitates assessment of community withdrawal success and guides the decision for further support. Box 1 Clinical features of acute alcohol withdrawal syndrome. Adapted from Alcohol withdrawal syndrome Acute alcohol withdrawal syndrome describes a spectrum of clinical sequelae typically occurring after abrupt cessation of alcohol consumption in patients with chronic, harmful intake. It has been attributed to impaired neurotransmitter regulation in the central nervous system, whereby chronically elevated cerebral ethanol concentrations decrease sensitivity to inhibitory gamma‐amino butyric acid (GABA) and interfere with glutamate receptor binding, causing compensatory upregulation of excitatory N‐methyl‐D‐Aspartate (NMDA) receptors. The sudden absence of ethanol in this context results in a pro‐excitatory imbalance that causes many of the observed symptoms and signs during withdrawal states. Minor symptoms typically occur within 6 h and include insomnia; tremor; anxiety; gastrointestinal upset; headache; diaphoresis; and palpitations. More severe features, described below, may occur in some patients. Withdrawal seizures and delirium tremens Generalised tonic–clonic seizures may occur, most commonly in patients aged 40–50 years with prolonged harmful intake. Seizure activity is most common 12–48 h following cessation of alcohol intake. If untreated, one third of alcoholic seizures may progress to delirium tremens, characterised by hypertension; tachycardia; confusion; hallucinations; agitation; and diaphoresis. It is distinct from isolated alcoholic hallucinations, which are not usually accompanied by confusion or deranged vital signs. With appropriate recognition and management, the mortality from delirium tremens is now < 5%. Metabolic abnormalities Patients may be hypovolaemic and demonstrate a metabolic acidosis, frequently accompanied by electrolyte abnormalities such as hypokalaemia; hypophosphataemia; and hypomagnesaemia. Deranged metabolic findings in this group are often multifactorial and interrelated, and themselves may result in organ dysfunction and an increased risk of cardiac failure, arrhythmias, reduced seizure threshold and mortality. On admission to hospital, the revised Clinical Institute Withdrawal Assessment – Alcohol (CIWA‐Ar) score should be measured in at‐risk patients . This is a 10‐point scale (Fig. ) used to assess alcohol withdrawal that can be used to direct pharmacological interventions; benzodiazepines such as chlordiazepoxide are a mainstay of therapy, while alternatives include carbamazepine and clomethiazole . The required interval between medication doses will vary but is often every 90 min with administration of a longer‐acting benzodiazepine, such as diazepam 20 mg, triggered at a CIWA‐Ar score of > 11 . This is repeated until there are three consecutive CIWA‐Ar scores < 11, indicating complete detoxification. The patient should then continue with standard vital and neurological observations unless there is a re‐emergence of symptoms. This strategy may warrant modification in established liver cirrhosis due to the concurrent risk of encephalopathy; in this setting the exclusive use of short‐acting benzodiazepines such as oxazepam is warranted. This may also apply to other groups who can have greater sensitivity to sedative drugs, such as patients who are older. Seizures are a common sequel of harmful alcohol intake, particularly in the context of withdrawal. Other differentials, such as hypoglycaemia or electrolyte deficiencies, should be considered as part of management. Alcohol‐related seizures occurring as an inpatient should be dealt with according to established guidelines . Benzodiazepines are a mainstay of therapy; phenytoin should be avoided . Following a seizure the patient should continue, or recommence, CIWA‐Ar scoring as appropriate. In elective settings, safe pre‐operative reduction in alcohol intake remains a core element of management. The level of intervention to achieve this is influenced by the degree of patient risk as determined by objective scoring or weekly alcohol intake. Pre‐operative assessment represents an ideal opportunity to make a brief intervention (a ‘teachable moment’) among those patients scoring 8–19 on the AUDIT score to discuss the potential harm from excessive alcohol intake and provide advice leaflets. Increased risk scores should prompt a recommendation that patients self‐refer to community alcohol support services . Clinicians should have lower intervention thresholds for patients with additional risk factors such as female sex, patients aged < 18 or > 65 y, and those from Black and minority ethnic backgrounds . Patients scoring > 19 on comprehensive AUDIT have possible alcohol dependency and should be referred for community‐based assisted withdrawal . This requires close liaison with local alcohol support services and primary care to ensure that it is carried out in a timely manner before admission for surgery. Those individuals who are judged to be vulnerable or are aged ≤ 16 y may need to be referred as an inpatient for specialist support and possible inpatient alcohol withdrawal by hospital alcohol support services (or gastroenterology teams where these are not available). A lower threshold for controlled in‐hospital detoxification may also be applied for cases where community management is superseded by operative urgency. Local service availability, expertise and referral pathways may vary such that a community withdrawal programme may be preferred rather than inpatient admission in some institutions. Ownership from named, contactable clinicians or services from the outset is important to promote continuity of care in all settings. Parent teams should review patient progress regularly and intervene as needed to ensure effective, safe management. Finally, the refusal of a competent patient to engage in discussed interventions should not be treated as an absolute contraindication to surgery provided there has been an appropriate discussion about related risks on a case‐by‐case basis. Inpatient specialist referral may be warranted for patients admitted with harmful drinking or alcohol‐dependence who are experiencing acute withdrawal or at high risk of this (Box ) . This includes urgent/emergency cases and patients admitted for planned detoxification. Close liaison aids early awareness of any potential issues but also facilitates assessment of community withdrawal success and guides the decision for further support. Box 1 Clinical features of acute alcohol withdrawal syndrome. Adapted from Alcohol withdrawal syndrome Acute alcohol withdrawal syndrome describes a spectrum of clinical sequelae typically occurring after abrupt cessation of alcohol consumption in patients with chronic, harmful intake. It has been attributed to impaired neurotransmitter regulation in the central nervous system, whereby chronically elevated cerebral ethanol concentrations decrease sensitivity to inhibitory gamma‐amino butyric acid (GABA) and interfere with glutamate receptor binding, causing compensatory upregulation of excitatory N‐methyl‐D‐Aspartate (NMDA) receptors. The sudden absence of ethanol in this context results in a pro‐excitatory imbalance that causes many of the observed symptoms and signs during withdrawal states. Minor symptoms typically occur within 6 h and include insomnia; tremor; anxiety; gastrointestinal upset; headache; diaphoresis; and palpitations. More severe features, described below, may occur in some patients. Withdrawal seizures and delirium tremens Generalised tonic–clonic seizures may occur, most commonly in patients aged 40–50 years with prolonged harmful intake. Seizure activity is most common 12–48 h following cessation of alcohol intake. If untreated, one third of alcoholic seizures may progress to delirium tremens, characterised by hypertension; tachycardia; confusion; hallucinations; agitation; and diaphoresis. It is distinct from isolated alcoholic hallucinations, which are not usually accompanied by confusion or deranged vital signs. With appropriate recognition and management, the mortality from delirium tremens is now < 5%. Metabolic abnormalities Patients may be hypovolaemic and demonstrate a metabolic acidosis, frequently accompanied by electrolyte abnormalities such as hypokalaemia; hypophosphataemia; and hypomagnesaemia. Deranged metabolic findings in this group are often multifactorial and interrelated, and themselves may result in organ dysfunction and an increased risk of cardiac failure, arrhythmias, reduced seizure threshold and mortality. On admission to hospital, the revised Clinical Institute Withdrawal Assessment – Alcohol (CIWA‐Ar) score should be measured in at‐risk patients . This is a 10‐point scale (Fig. ) used to assess alcohol withdrawal that can be used to direct pharmacological interventions; benzodiazepines such as chlordiazepoxide are a mainstay of therapy, while alternatives include carbamazepine and clomethiazole . The required interval between medication doses will vary but is often every 90 min with administration of a longer‐acting benzodiazepine, such as diazepam 20 mg, triggered at a CIWA‐Ar score of > 11 . This is repeated until there are three consecutive CIWA‐Ar scores < 11, indicating complete detoxification. The patient should then continue with standard vital and neurological observations unless there is a re‐emergence of symptoms. This strategy may warrant modification in established liver cirrhosis due to the concurrent risk of encephalopathy; in this setting the exclusive use of short‐acting benzodiazepines such as oxazepam is warranted. This may also apply to other groups who can have greater sensitivity to sedative drugs, such as patients who are older. Seizures are a common sequel of harmful alcohol intake, particularly in the context of withdrawal. Other differentials, such as hypoglycaemia or electrolyte deficiencies, should be considered as part of management. Alcohol‐related seizures occurring as an inpatient should be dealt with according to established guidelines . Benzodiazepines are a mainstay of therapy; phenytoin should be avoided . Following a seizure the patient should continue, or recommence, CIWA‐Ar scoring as appropriate. Day‐case surgery There are few published data relating to day surgery in patients with harmful alcohol intake. However, day‐case procedures are potentially advantageous in this group as they can avoid the withdrawal states associated with inpatient hospital stays. Many general aspects of management are encompassed within pre‐existing guidelines . At all stages, appropriate involvement of the multidisciplinary team is vital for decision‐making regarding day‐case suitability and timing of other elements, such as controlled withdrawal programmes. Case‐by‐case decision making may vary depending on individual patient characteristics, local resources and personnel, particularly when planning safe postoperative discharge. In high‐volume, nurse‐led pre‐assessment settings, the addition of the AUDIT‐C to other commonly used scoring systems provides a valuable, rapidly applicable measure for seeking further clinician involvement such as discussion with an anaesthetist. Day surgery‐specific pre‐operative assessment should otherwise be undertaken by practitioners competent in eliciting symptoms and signs indicative of complications from chronic alcohol use. Cardiovascular and hepatic manifestations such as dilated cardiomyopathy and cirrhosis are particularly relevant to day‐case suitability. The presence of coagulopathy or thrombocytopenia, with their attendant bleeding risks, should also influence decision‐making. It may be safer for some patients to be managed as inpatients, depending on the nature of the planned surgical intervention and outcomes of appropriate multidisciplinary discussions. As for inpatient surgery, consideration should be given to deferring surgery for planned community withdrawal under appropriate supervision for all high‐risk patients. Potential benefits from abstinence for at least 2 weeks include improved platelet function and therefore reduced bleeding risk, while cessation of alcohol intake for 8 weeks or longer may improve wound healing . Patients may refuse to engage with a pre‐operative intervention as well as advice for the postoperative period, such as non‐consumption of alcohol for 24 h. This may affect the patients' appropriateness for day‐case procedures and should carefully be explored, with appropriate risk/benefit discussions and development of an individualised management plan in consultation with the wider team. Patients who present acutely intoxicated on the day of surgery should have surgery postponed with appropriate follow up. There may be a particular role for regional techniques, including peripheral nerve blockade if feasible and safe, to facilitate same‐day discharge while potentially reducing home analgesic requirements. The use of short‐acting anaesthetic drugs provides an additional margin of safety by minimising the risk of pharmacological interaction with alcohol consumed after surgery. Caution should be exercised when prescribing take‐home postoperative systemic drugs where there is evidence of alcohol‐related liver disease (Table ). All patients with harmful alcohol intake should have a clearly documented plan for safe discharge following day surgery. Consideration should be given to social circumstances in high‐risk patients, given the potential for home alcohol withdrawal and the interaction of alcohol consumed at home with peri‐operative medications. Such patients may not be suitable for ‘home alone’ discharge as described elsewhere and should return home with a responsible adult . Nurse‐led discharge on the day of surgery should otherwise follow explicit criteria according to local protocols, with additional vigilance for signs of withdrawal in the recovery area. Provision of written patient discharge material, with details of surgical procedure, follow‐up planning and key contacts such as community alcohol liaison services as relevant, is critical for ongoing patient safety. Emergency surgery It may not be practical or safe to implement some of the pre‐operative measures described above, such as controlled alcohol withdrawal, in patients who require time‐critical or emergency surgery. However, it should be noted that acute presentation does not preclude judicious pre‐operative intervention, depending on clinical context. Furthermore, the identification of patients with excessive alcohol intake also influences acute management within the operating theatre and after surgery (Fig. ). If feasible, screening patients rapidly with the AUDIT‐C questionnaire during anaesthetic assessment is therefore important for prompt recognition, early involvement of other key specialities and tailoring of subsequent therapy. There are few published data relating to day surgery in patients with harmful alcohol intake. However, day‐case procedures are potentially advantageous in this group as they can avoid the withdrawal states associated with inpatient hospital stays. Many general aspects of management are encompassed within pre‐existing guidelines . At all stages, appropriate involvement of the multidisciplinary team is vital for decision‐making regarding day‐case suitability and timing of other elements, such as controlled withdrawal programmes. Case‐by‐case decision making may vary depending on individual patient characteristics, local resources and personnel, particularly when planning safe postoperative discharge. In high‐volume, nurse‐led pre‐assessment settings, the addition of the AUDIT‐C to other commonly used scoring systems provides a valuable, rapidly applicable measure for seeking further clinician involvement such as discussion with an anaesthetist. Day surgery‐specific pre‐operative assessment should otherwise be undertaken by practitioners competent in eliciting symptoms and signs indicative of complications from chronic alcohol use. Cardiovascular and hepatic manifestations such as dilated cardiomyopathy and cirrhosis are particularly relevant to day‐case suitability. The presence of coagulopathy or thrombocytopenia, with their attendant bleeding risks, should also influence decision‐making. It may be safer for some patients to be managed as inpatients, depending on the nature of the planned surgical intervention and outcomes of appropriate multidisciplinary discussions. As for inpatient surgery, consideration should be given to deferring surgery for planned community withdrawal under appropriate supervision for all high‐risk patients. Potential benefits from abstinence for at least 2 weeks include improved platelet function and therefore reduced bleeding risk, while cessation of alcohol intake for 8 weeks or longer may improve wound healing . Patients may refuse to engage with a pre‐operative intervention as well as advice for the postoperative period, such as non‐consumption of alcohol for 24 h. This may affect the patients' appropriateness for day‐case procedures and should carefully be explored, with appropriate risk/benefit discussions and development of an individualised management plan in consultation with the wider team. Patients who present acutely intoxicated on the day of surgery should have surgery postponed with appropriate follow up. There may be a particular role for regional techniques, including peripheral nerve blockade if feasible and safe, to facilitate same‐day discharge while potentially reducing home analgesic requirements. The use of short‐acting anaesthetic drugs provides an additional margin of safety by minimising the risk of pharmacological interaction with alcohol consumed after surgery. Caution should be exercised when prescribing take‐home postoperative systemic drugs where there is evidence of alcohol‐related liver disease (Table ). All patients with harmful alcohol intake should have a clearly documented plan for safe discharge following day surgery. Consideration should be given to social circumstances in high‐risk patients, given the potential for home alcohol withdrawal and the interaction of alcohol consumed at home with peri‐operative medications. Such patients may not be suitable for ‘home alone’ discharge as described elsewhere and should return home with a responsible adult . Nurse‐led discharge on the day of surgery should otherwise follow explicit criteria according to local protocols, with additional vigilance for signs of withdrawal in the recovery area. Provision of written patient discharge material, with details of surgical procedure, follow‐up planning and key contacts such as community alcohol liaison services as relevant, is critical for ongoing patient safety. It may not be practical or safe to implement some of the pre‐operative measures described above, such as controlled alcohol withdrawal, in patients who require time‐critical or emergency surgery. However, it should be noted that acute presentation does not preclude judicious pre‐operative intervention, depending on clinical context. Furthermore, the identification of patients with excessive alcohol intake also influences acute management within the operating theatre and after surgery (Fig. ). If feasible, screening patients rapidly with the AUDIT‐C questionnaire during anaesthetic assessment is therefore important for prompt recognition, early involvement of other key specialities and tailoring of subsequent therapy. General considerations Consent should include details of the additional risks of alcohol withdrawal and postoperative complications. Patients who are intoxicated, encephalopathic or have chronic alcohol‐related brain damage may not have sufficient mental capacity to provide consent specific to the intended procedure. Management should incorporate the patient's best interests and, if appropriate, an attempt at substituted judgement regarding their wishes. These elements should align with the principles of the Mental Capacity Act (2005) in keeping with existing guidance published by the Association of Anaesthetists . Patients with cirrhosis should be discussed with gastroenterology or hepatology specialists where possible. The severity of cirrhosis correlates with the risk of postoperative mortality and a careful history should be taken for features of hepatic decompensation. The risk calculators discussed previously may assist with planning for these cases. However, such tools should be treated as adjuncts to clinical decision‐making; it may be most appropriate to refer complex and/or decompensated patients to a liver centre where there is immediately available subspecialty expertise. In the elective setting, appropriately timed optimisation for this cohort may include treatment of ascites; screening and prophylaxis of varices; treatment of hepatic encephalopathy; and nutritional intervention. Balanced correction of nutritional, metabolic, fluid and haematological abnormalities will require multidisciplinary input, with referral to relevant specialist teams for organ‐specific complications. The detailed peri‐operative management of deranged physiology in liver disease is covered in a recent publication . Issues most relevant to patients with harmful alcohol intake are discussed below. Nutritional issues The direct effects of harmful alcohol intake may compound with secondary liver disease to profoundly disrupt energy and nutritional homeostasis. Resultant malnutrition is common, occurring in 60–85% of patients with cirrhosis and encompassing a spectrum of conditions including sarcopenia; frailty; electrolyte abnormalities; and vitamin deficiencies. Underlying mechanisms are multiple and interdependent, including appetite loss; intestinal mucosal/microbiome disruption with resultant malabsorption; and hypercatabolic states secondary to oxidative alcohol metabolism and systemic, low grade bacterial translocation . Nutritional status should therefore be considered in all patients with harmful alcohol intake. There is no gold standard tool for screening this population for malnutrition. The Malnutrition Universal Screening Tool is recommended by the European Society for Clinical Nutrition and Metabolism, although BMI calculations may be confounded by the presence of ascites. Alternative screening tools adapted for liver disease, such as the Royal Free Hospital Nutritional Prioritisation Tool, may present a useful alternative if there is diagnostic uncertainty . Although detailed discussion regarding management is beyond the scope of this document, general considerations include energy requirements; protein intake; electrolyte replacement; vitamin supplementation; and micronutrients. Associated hepatic complications, such as alcoholic hepatitis, ascites or encephalopathy may alter individual needs in these domains . Harmful alcohol intake is also an independent risk factor for the development of refeeding syndrome , marked by potentially life‐threatening fluid and electrolyte shifts precipitated by over‐supplementation in susceptible individuals. Appropriate involvement of a dietician for complex cases is advised throughout the peri‐operative period. It should also be noted that infusions of ferric carboxymaltose (Ferinject®, Vifor Pharma UK Ltd, Staines, UK) for iron replacement may present a particular risk of hypophosphataemia for some patients in this group due to pre‐existing phosphate depletion/concurrent refeeding syndrome. Regular thiamine‐containing supplements, such as Pabrinex® (Kyowa Kirin Ltd, Galashiels, UK), should be prescribed at admission to patients admitted with chronic harmful alcohol consumption to prevent Wernicke's encephalopathy, a serious complication of alcohol misuse comprising ophthalmoplegia, ataxia and acute delirium precipitated by vitamin B deficiency. Regimens may vary by institution but would typically include intravenous therapy for 72 h followed by conversion to oral supplementation (Fig. ). Haematological issues The treatment of coagulation abnormalities in patients with cirrhosis before surgery has become increasingly complex. For low‐risk procedures, routine administration of blood products such as fresh frozen plasma to achieve a specific laboratory coagulation value has little benefit and may cause harm. Where bleeding is considered likely, such as in those undergoing major surgery, it is reasonable to target specific endpoints such as a platelet count of > 50 × 10 9 cells.l ‐1 or a fibrinogen concentration of > 1 g.l ‐1 in decompensated disease. Appropriate haematological involvement, supplemented by investigations such as viscoelastic testing, is critical for decision‐making in complex cases . The literature on the efficacy and safety of peri‐operative tranexamic acid in patients with chronic harmful alcohol intake and associated liver disease is incomplete. Recent recommendations from NICE advise routine prophylaxis in particular circumstances, in keeping with several large meta‐analyses . This contrasts with a recent, large multicentre randomised controlled trial indicating that tranexamic acid use in acute upper gastrointestinal haemorrhage, a known complication of cirrhosis, does not reduce 5‐day mortality from bleeding and may slightly increase the risk of venous thromboembolism and seizures . The authors highlighted the divergent findings in this study compared with a preceding Cochrane review, attributing this in part to the potential for erroneous results in meta‐analyses of small studies . Literature examining tranexamic acid use for the management of bleeding in other emergency settings predominantly consists of single‐centre studies and has not demonstrated an increased thrombotic risk, although high doses may increase the risk of seizures . These studies do not specifically focus on patients with harmful alcohol intake or liver disease and the true risk/benefit of antifibrinolytic therapy in this group therefore remains controversial. In the absence of definitive evidence or a specific contraindication, it remains reasonable to give tranexamic acid for the prevention or management of acute haemorrhage in most circumstances. Avoidance of high doses is advised where seizure risk is elevated, such as acute withdrawal states or when patients are at risk of this. Consent should include details of the additional risks of alcohol withdrawal and postoperative complications. Patients who are intoxicated, encephalopathic or have chronic alcohol‐related brain damage may not have sufficient mental capacity to provide consent specific to the intended procedure. Management should incorporate the patient's best interests and, if appropriate, an attempt at substituted judgement regarding their wishes. These elements should align with the principles of the Mental Capacity Act (2005) in keeping with existing guidance published by the Association of Anaesthetists . Patients with cirrhosis should be discussed with gastroenterology or hepatology specialists where possible. The severity of cirrhosis correlates with the risk of postoperative mortality and a careful history should be taken for features of hepatic decompensation. The risk calculators discussed previously may assist with planning for these cases. However, such tools should be treated as adjuncts to clinical decision‐making; it may be most appropriate to refer complex and/or decompensated patients to a liver centre where there is immediately available subspecialty expertise. In the elective setting, appropriately timed optimisation for this cohort may include treatment of ascites; screening and prophylaxis of varices; treatment of hepatic encephalopathy; and nutritional intervention. Balanced correction of nutritional, metabolic, fluid and haematological abnormalities will require multidisciplinary input, with referral to relevant specialist teams for organ‐specific complications. The detailed peri‐operative management of deranged physiology in liver disease is covered in a recent publication . Issues most relevant to patients with harmful alcohol intake are discussed below. The direct effects of harmful alcohol intake may compound with secondary liver disease to profoundly disrupt energy and nutritional homeostasis. Resultant malnutrition is common, occurring in 60–85% of patients with cirrhosis and encompassing a spectrum of conditions including sarcopenia; frailty; electrolyte abnormalities; and vitamin deficiencies. Underlying mechanisms are multiple and interdependent, including appetite loss; intestinal mucosal/microbiome disruption with resultant malabsorption; and hypercatabolic states secondary to oxidative alcohol metabolism and systemic, low grade bacterial translocation . Nutritional status should therefore be considered in all patients with harmful alcohol intake. There is no gold standard tool for screening this population for malnutrition. The Malnutrition Universal Screening Tool is recommended by the European Society for Clinical Nutrition and Metabolism, although BMI calculations may be confounded by the presence of ascites. Alternative screening tools adapted for liver disease, such as the Royal Free Hospital Nutritional Prioritisation Tool, may present a useful alternative if there is diagnostic uncertainty . Although detailed discussion regarding management is beyond the scope of this document, general considerations include energy requirements; protein intake; electrolyte replacement; vitamin supplementation; and micronutrients. Associated hepatic complications, such as alcoholic hepatitis, ascites or encephalopathy may alter individual needs in these domains . Harmful alcohol intake is also an independent risk factor for the development of refeeding syndrome , marked by potentially life‐threatening fluid and electrolyte shifts precipitated by over‐supplementation in susceptible individuals. Appropriate involvement of a dietician for complex cases is advised throughout the peri‐operative period. It should also be noted that infusions of ferric carboxymaltose (Ferinject®, Vifor Pharma UK Ltd, Staines, UK) for iron replacement may present a particular risk of hypophosphataemia for some patients in this group due to pre‐existing phosphate depletion/concurrent refeeding syndrome. Regular thiamine‐containing supplements, such as Pabrinex® (Kyowa Kirin Ltd, Galashiels, UK), should be prescribed at admission to patients admitted with chronic harmful alcohol consumption to prevent Wernicke's encephalopathy, a serious complication of alcohol misuse comprising ophthalmoplegia, ataxia and acute delirium precipitated by vitamin B deficiency. Regimens may vary by institution but would typically include intravenous therapy for 72 h followed by conversion to oral supplementation (Fig. ). The treatment of coagulation abnormalities in patients with cirrhosis before surgery has become increasingly complex. For low‐risk procedures, routine administration of blood products such as fresh frozen plasma to achieve a specific laboratory coagulation value has little benefit and may cause harm. Where bleeding is considered likely, such as in those undergoing major surgery, it is reasonable to target specific endpoints such as a platelet count of > 50 × 10 9 cells.l ‐1 or a fibrinogen concentration of > 1 g.l ‐1 in decompensated disease. Appropriate haematological involvement, supplemented by investigations such as viscoelastic testing, is critical for decision‐making in complex cases . The literature on the efficacy and safety of peri‐operative tranexamic acid in patients with chronic harmful alcohol intake and associated liver disease is incomplete. Recent recommendations from NICE advise routine prophylaxis in particular circumstances, in keeping with several large meta‐analyses . This contrasts with a recent, large multicentre randomised controlled trial indicating that tranexamic acid use in acute upper gastrointestinal haemorrhage, a known complication of cirrhosis, does not reduce 5‐day mortality from bleeding and may slightly increase the risk of venous thromboembolism and seizures . The authors highlighted the divergent findings in this study compared with a preceding Cochrane review, attributing this in part to the potential for erroneous results in meta‐analyses of small studies . Literature examining tranexamic acid use for the management of bleeding in other emergency settings predominantly consists of single‐centre studies and has not demonstrated an increased thrombotic risk, although high doses may increase the risk of seizures . These studies do not specifically focus on patients with harmful alcohol intake or liver disease and the true risk/benefit of antifibrinolytic therapy in this group therefore remains controversial. In the absence of definitive evidence or a specific contraindication, it remains reasonable to give tranexamic acid for the prevention or management of acute haemorrhage in most circumstances. Avoidance of high doses is advised where seizure risk is elevated, such as acute withdrawal states or when patients are at risk of this. Intra‐operative The systemic effects of harmful alcohol intake, in combination with other potential issues such as acute illness and secondary organ dysfunction, may warrant modification of anaesthesia techniques (Table ). It should be noted that there are few outcome data supporting neuraxial techniques over general anaesthesia in patients with established liver disease . Given the challenges encountered in patients with harmful alcohol consumption, it may be more appropriate to consider performing surgery using regional or local anaesthetic techniques. Potential intra‐ and postoperative benefits include avoidance of the cardiorespiratory and neurological effects of general anaesthesia in a high‐risk cohort; enhanced seizure detection in patients who are awake; and mitigation of pharmacological effects of systemic medications. Due consideration should be given to factors influencing the suitability of regional techniques in these patients, including appropriateness of techniques for surgical procedures and their duration; reduced patient cooperation or reduction in mental capacity due to intoxication or chronic neurological impairment; presence of coagulopathy and/or thrombocytopenia; and the influence of hypoproteinaemia (secondary to malnutrition and/or hepatic synthetic dysfunction) on local anaesthetic distribution and dosing limits. Pharmacological issues Chronic harmful alcohol intake and related organ dysfunction, as well as acute intoxication, may significantly influence the efficacy and safety of commonly used medications, with a commensurate requirement to modify or avoid their use. Where there is alcohol‐related secondary organ dysfunction, judicious administration of intra‐operative anaesthetic drugs is warranted (Table ). This reflects vulnerability to the cardiorespiratory effects of such medications, as well as the widely recognised changes in drug pharmacokinetics seen in alcohol‐related liver disease. Postoperative Unhealthy alcohol intake contributes significantly to postoperative complications, in particular alcohol withdrawal syndrome, postoperative infections and delirium . This leads to an increased duration of hospital stay, ICU admissions and mortality . This picture is exacerbated by the presence of related comorbidities, decompensation of which may precipitate hepatic encephalopathy, cardiac ischaemia and arrhythmias . Patients should therefore be monitored to detect complications, with a low threshold for postoperative admission to critical care. The decision to refer for critical care support should be guided using a peri‐operative risk calculator, such as P‐POSSUM, which should be appropriately repeated at the end of surgery to account for determinant intra‐procedural events such as major haemorrhage. It should be noted that liver‐specific scoring systems outlined earlier have primary utility in pre‐operative discussion and referral; they are not suitable for dynamic assessment of risk and determination of postoperative location for patients undergoing procedures. The incidence of alcohol withdrawal syndrome is estimated to be 2–5 times higher in surgical patients compared with other inpatients and has a higher associated morbidity . Features range from minor symptoms such as headache, tremor, and insomnia, to more serious manifestations including withdrawal seizures, hallucinations and delirium tremens. Approximately 5% of patients who undergo alcohol withdrawal suffer from delirium tremens and mortality is significant if left untreated . At‐risk patients should be monitored using the described CIWA‐Ar scale or equivalent, with ongoing consideration for thiamine supplementation . Where available, continued oversight from alcohol support services is advised. Patients with established alcohol‐related liver disease are at increased risk of postoperative morbidity and mortality, a picture complicated by the potential impact of both hepatic dysfunction and continued harmful alcohol intake on postoperative analgesia (Table ) . Hepatic encephalopathy can be precipitated by hypoxia, hypovolaemia or acid–base/electrolyte disturbance, which should be avoided during and after surgery . Infections may also provoke hepatic decompensation, warranting vigilance and aggressive treatment when present. Patients with cirrhosis are also at high risk of renal dysfunction, necessitating caution when using nephrotoxic agents and judicious fluid balance, aiming to preserve renal perfusion while avoiding volume overload. The systemic effects of harmful alcohol intake, in combination with other potential issues such as acute illness and secondary organ dysfunction, may warrant modification of anaesthesia techniques (Table ). It should be noted that there are few outcome data supporting neuraxial techniques over general anaesthesia in patients with established liver disease . Given the challenges encountered in patients with harmful alcohol consumption, it may be more appropriate to consider performing surgery using regional or local anaesthetic techniques. Potential intra‐ and postoperative benefits include avoidance of the cardiorespiratory and neurological effects of general anaesthesia in a high‐risk cohort; enhanced seizure detection in patients who are awake; and mitigation of pharmacological effects of systemic medications. Due consideration should be given to factors influencing the suitability of regional techniques in these patients, including appropriateness of techniques for surgical procedures and their duration; reduced patient cooperation or reduction in mental capacity due to intoxication or chronic neurological impairment; presence of coagulopathy and/or thrombocytopenia; and the influence of hypoproteinaemia (secondary to malnutrition and/or hepatic synthetic dysfunction) on local anaesthetic distribution and dosing limits. Chronic harmful alcohol intake and related organ dysfunction, as well as acute intoxication, may significantly influence the efficacy and safety of commonly used medications, with a commensurate requirement to modify or avoid their use. Where there is alcohol‐related secondary organ dysfunction, judicious administration of intra‐operative anaesthetic drugs is warranted (Table ). This reflects vulnerability to the cardiorespiratory effects of such medications, as well as the widely recognised changes in drug pharmacokinetics seen in alcohol‐related liver disease. Unhealthy alcohol intake contributes significantly to postoperative complications, in particular alcohol withdrawal syndrome, postoperative infections and delirium . This leads to an increased duration of hospital stay, ICU admissions and mortality . This picture is exacerbated by the presence of related comorbidities, decompensation of which may precipitate hepatic encephalopathy, cardiac ischaemia and arrhythmias . Patients should therefore be monitored to detect complications, with a low threshold for postoperative admission to critical care. The decision to refer for critical care support should be guided using a peri‐operative risk calculator, such as P‐POSSUM, which should be appropriately repeated at the end of surgery to account for determinant intra‐procedural events such as major haemorrhage. It should be noted that liver‐specific scoring systems outlined earlier have primary utility in pre‐operative discussion and referral; they are not suitable for dynamic assessment of risk and determination of postoperative location for patients undergoing procedures. The incidence of alcohol withdrawal syndrome is estimated to be 2–5 times higher in surgical patients compared with other inpatients and has a higher associated morbidity . Features range from minor symptoms such as headache, tremor, and insomnia, to more serious manifestations including withdrawal seizures, hallucinations and delirium tremens. Approximately 5% of patients who undergo alcohol withdrawal suffer from delirium tremens and mortality is significant if left untreated . At‐risk patients should be monitored using the described CIWA‐Ar scale or equivalent, with ongoing consideration for thiamine supplementation . Where available, continued oversight from alcohol support services is advised. Patients with established alcohol‐related liver disease are at increased risk of postoperative morbidity and mortality, a picture complicated by the potential impact of both hepatic dysfunction and continued harmful alcohol intake on postoperative analgesia (Table ) . Hepatic encephalopathy can be precipitated by hypoxia, hypovolaemia or acid–base/electrolyte disturbance, which should be avoided during and after surgery . Infections may also provoke hepatic decompensation, warranting vigilance and aggressive treatment when present. Patients with cirrhosis are also at high risk of renal dysfunction, necessitating caution when using nephrotoxic agents and judicious fluid balance, aiming to preserve renal perfusion while avoiding volume overload. Patients with harmful alcohol intake represent a complex and challenging population, mandating a multidisciplinary team‐based approach at all stages in a cohort whose disease may be clinically silent until unmasked by surgical stressors. Safety depends on an understanding of the direct and indirect effects of alcohol intake on physiology and pharmacology, with the use of validated tools to identify those who would benefit from withdrawal management and targeted critical care involvement in the immediate peri‐operative phase. |
Invention of a new percutaneous closure technique for vascular haemostasis in percutaneous veno-arterial extracorporeal membrane oxygenation | 07a679fa-39bf-4e0b-8750-80d87dfccbcd | 11837478 | Surgical Procedures, Operative[mh] | Veno-arterial extracorporeal membrane oxygenation (VA-ECMO) is a form of temporary mechanical circulatory support frequently used to provide respiratory and cardiac support to critically ill patients . Current studies have shown that VA-ECMO may reduce the short-term mortality risk of patients with refractory acute respiratory distress syndrome, pulmonary embolism and cardiogenic shock . VA-ECMO has been used to treat severely ill COVID-19 patients . For VA-ECMO, peripheral arterial access is established by either surgical arterial cutdown or percutaneously via a modified Seldinger technique. Deployed arterial cannulas typically range from 8 to 21F. And the common femoral artery is most frequently utilized given its size and ease of access . Surgical arterial cutdown is the most traditional and reliable technique. However, this technique has disadvantages such as a long operation time, multiple complications and a long hospital stay . Compared with the surgical arterial cutdown, the percutaneous approach to obtain vascular access for peripheral VA-ECMO was associated with fewer local infections and improved 30-day survival outcomes . However, rapid haemostasis at the vascular puncture point after weaning from VA-ECMO as well as reducing the risk of vascular complications are the most notable challenges associated with percutaneous approaches. The Perclose ProGlide (Abbott Vascular) is a percutaneous vascular suture device. It can be used to percutaneously suture the puncture site with polypropylene material and therefore achieve immediate and lasting haemostasis. However, the ProGlide may not be able to close the puncture site if the size of the sheath is larger than 8F . It has been reported that the deployment of the ProGlide device prior to the insertion of large-bore devices is referred to as the preclosure technique (Pre-CT). Pre-CT allows rapid arteriotomy closure after percutaneous interventions, as it is performed immediately after the procedure is completed. They allow early ambulation and discharge after groin puncture and mitigate patient discomfort caused by extended manual compression . However, pre-CT prolongs the overall duration of the VA-ECMO procedure, adds additional postoperative care to workloads and increases the potential for ProGlide stitch infection . Based on clinical experience, we propose a new technique, the parallel closure technique (Par-CT), to suture large aperture puncture points. The accuracy of ProGlide suturing can be improved by replacing the larger vascular sheath with a smaller vascular sheath to decrease the size of the puncture point of the femoral artery. The advantage of percutaneous puncture was preserved, and the haemostatic effect was improved by using the Par-CT. Technical principles of the Par-CT Arterial cannulas range from 8 to 21F for VA-ECMO. If the cannula is larger than 8F, it is difficult to effectively achieve vascular haemostasis with only ProGlide sutures. Therefore, Par-CT was required to close the vascular puncture site (as shown in Fig. ). After puncture of the proximal portion of the arterial cannula by using the Seldinger technique, a 0.035-in. guidewire (guidewire 1) was cautiously advanced into the aorta. The arterial cannula was withdrawn and removed under manual compression during access. The appropriately sized arterial sheath was selected according to the diameter of the arterial cannula, and the diameter of the vascular puncture outside the arterial sheath was reduced to approximately 2.55 mm (8F sheath diameter). The arterial sheath was placed along guidewire 1 to reduce the area of the vascular puncture point, and a short guidewire (guidewire 2) was introduced through the arterial sheath (two short guidewires may be introduced if necessary). The ProGlide was advanced along guidewire 2 to allow suturing of the puncture point of the vessel outside the sheath. The second ProGlide was advanced along guidewire 1 after removing the sheath to allow suturing of the remaining puncture site. After the suture was tightened, two ProGlide stitches were tied to enhance haemostasis. Selection of vascular sheath size The puncture point was defined as an oval puncture hole with a perimeter of 17 mm. Using the software, the perimeter of the remaining space after the insertion of vascular sheaths of different sizes was calculated. Considering that the remaining parts were irregular semicircles, the maximum suture perimeter of the second ProGlide suture device was measured by changing different sheaths (as shown in Fig. ). Materials and methods The study protocol was approved by the Animal Ethics Committee of The Fourth Affiliated Hospital of Soochow University, and the study was conducted in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The study was conducted in compliance with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines . Five male Boer goats (9 months of age, mean body weight 54.17 ± 3.43 kg) were treated under general anaesthesia, and all efforts were made to minimize pain. On the day of the procedure, the animals were anaesthetized with zolazepam (2.5 mg/kg) and xylazine (0.02 ml/10 kg) and anaesthesia was maintained with isoflurane (1–3%). The animals were intubated and received mechanical ventilatory support for the duration of the procedure. After the site was disinfected, an incision was made in the median abdomen of the goats. The abdominal wall tissue was separated layer by layer, and the abdominal aorta was carefully exposed and separated. Based on body weight, administered heparin at a dosage of 1 mg/kg to prevent thrombosis. A 17F vascular sheath (Terumo Corporation) was used to simulate VA-ECMO arterial intubation. Vascular sheaths of different sizes were introduced to reduce the puncture site area, and the Par-CT was used to suture the puncture site. After checking that the surgical wound had no obvious bleeding and counting the surgical instruments to ensure none were missing, the abdominal incision was sutured (as shown in Fig. ). The animals were fed for 2 weeks after the operation. Arterial cannulas range from 8 to 21F for VA-ECMO. If the cannula is larger than 8F, it is difficult to effectively achieve vascular haemostasis with only ProGlide sutures. Therefore, Par-CT was required to close the vascular puncture site (as shown in Fig. ). After puncture of the proximal portion of the arterial cannula by using the Seldinger technique, a 0.035-in. guidewire (guidewire 1) was cautiously advanced into the aorta. The arterial cannula was withdrawn and removed under manual compression during access. The appropriately sized arterial sheath was selected according to the diameter of the arterial cannula, and the diameter of the vascular puncture outside the arterial sheath was reduced to approximately 2.55 mm (8F sheath diameter). The arterial sheath was placed along guidewire 1 to reduce the area of the vascular puncture point, and a short guidewire (guidewire 2) was introduced through the arterial sheath (two short guidewires may be introduced if necessary). The ProGlide was advanced along guidewire 2 to allow suturing of the puncture point of the vessel outside the sheath. The second ProGlide was advanced along guidewire 1 after removing the sheath to allow suturing of the remaining puncture site. After the suture was tightened, two ProGlide stitches were tied to enhance haemostasis. The puncture point was defined as an oval puncture hole with a perimeter of 17 mm. Using the software, the perimeter of the remaining space after the insertion of vascular sheaths of different sizes was calculated. Considering that the remaining parts were irregular semicircles, the maximum suture perimeter of the second ProGlide suture device was measured by changing different sheaths (as shown in Fig. ). The study protocol was approved by the Animal Ethics Committee of The Fourth Affiliated Hospital of Soochow University, and the study was conducted in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The study was conducted in compliance with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines . Five male Boer goats (9 months of age, mean body weight 54.17 ± 3.43 kg) were treated under general anaesthesia, and all efforts were made to minimize pain. On the day of the procedure, the animals were anaesthetized with zolazepam (2.5 mg/kg) and xylazine (0.02 ml/10 kg) and anaesthesia was maintained with isoflurane (1–3%). The animals were intubated and received mechanical ventilatory support for the duration of the procedure. After the site was disinfected, an incision was made in the median abdomen of the goats. The abdominal wall tissue was separated layer by layer, and the abdominal aorta was carefully exposed and separated. Based on body weight, administered heparin at a dosage of 1 mg/kg to prevent thrombosis. A 17F vascular sheath (Terumo Corporation) was used to simulate VA-ECMO arterial intubation. Vascular sheaths of different sizes were introduced to reduce the puncture site area, and the Par-CT was used to suture the puncture site. After checking that the surgical wound had no obvious bleeding and counting the surgical instruments to ensure none were missing, the abdominal incision was sutured (as shown in Fig. ). The animals were fed for 2 weeks after the operation. The abdominal aorta of 5 goats was punctured with a 17F sheath, and a vascular sheath of 6 to 10F (TERUMO, Japan) was introduced to reduce the area of the vascular tear. The abdominal aorta vascular tear was successfully sutured by using the Par-CT. The puncture sites in goats 1 and 2 were too large to suture after the first Par-CT, so an 8F vascular sheath (TERUMO, Japan) was reintroduced to reduce the remaining area and successfully suture the puncture site. The Par-CT was used to effectively suture the puncture sites in goats 3, 4, and 5, consistent with the predicted model results. After the operation, there was smooth blood flow in the distal abdominal aorta with no obvious blood seepage at the puncture point. After a 2 week postoperative feeding period, no complications, such as massive haemorrhage, abdominal cavity or incision infection, were observed, nor were there any ischaemic manifestations, such as lower limb ischaemia. There were no accidental deaths during or after surgery. VA-ECMO has emerged as the preferred treatment option for patients suffering from cardiogenic shock or respiratory failure . The most common catheterization approach involves the femoral artery. Generally, simple compression can achieve an ideal haemostatic effect if small-calibre arterial catheters (8F) are used. However, to maintain adequate blood flow, a larger diameter vascular sheath is often needed, which increases the incidence of peripheral vascular complications . The establishment of and weaning from VA-ECMO often require surgical operation to cut and suture the peripheral artery. This method is effective but time-consuming and delays the treatment of some critically ill patients. Exposure of blood vessels also leads to a greater risk of infection. An increasing number of centres are adopting the percutaneous technique for establishing VA-ECMO. Compared to surgery, this method offers advantages such as rapid catheterization, reduced tissue damage, and a lower wound infection rate. However, surgical suturing is still necessary after weaning and extubation. Compared with traditional surgery, the use of the ProGlide vascular suture device has improved this situation by reducing transfusion rates, length of hospital stay, and overall hospital costs . However, ProGlide is indicated for suturing vessels after removal of conduits with sheaths less than 8F. For some puncture points with larger conduits, suturing may still be effective . Lee proposed the use of a ProGlide vascular suture device for haemostasis after the extubation of vascular sheaths of up to 24F using PCT technology . Since then, this technology has been widely used in interventional procedures requiring large-diameter vascular sheaths and has been gradually applied to the extubation of VA-ECMO patients. A study compared the ProGlide suture device with surgery for femoral artery suturing after ECMO weaning and revealed that although limb ischaemia and infection were more common in the surgical incision group, the overall complication rate did not significantly differ between the two groups, which may be related to the poor performance of the suture device in suturing large-calibre puncture sites . Later, Martin et al. performed percutaneous extubation in 20 patients using PCT technology, with a technical success rate of 95% (19/20) . Chandel et al. compared the PCT technique with surgery and reported that compared with surgical resection, PCT was feasible and safe and reduced the likelihood of both limb complications and bleeding events by approximately 80% . However, this technique prolongs the duration of VA-ECMO support, requires additional postoperative care, and increases the likelihood of ProGlide suture infection . For emergency patients, preimplantation may fail due to factors such as a small diameter of the access vessel and calcification of the anterior wall. VA-ECMO cannulas often have large side holes and lack hydrophilic coatings. This special structure causes the preset ProGlide suture to easily become entangled with the tube at the time of extubation, resulting in difficult sheath removal if the suture is tightened prematurely. To avoid these situations, some scholars have proposed adopting a minimally invasive approach for weaning patients from VA-ECMO support as well as balloon dilation, which can be performed percutaneously to remove the arterial cannula of VA-ECMO in a short time . However, this method requires patient transportation and therefore has higher risks than those of bedside extubation . Li et al. compared the ProGlide suture technique with open arterial repair for VA-ECMO decannulation and reported that the ProGlide suture technique had a shorter operative time, shorter ICU stay after decannulation, and less bleeding events and pain. However, two ProGlide sutures are introduced, which makes both compression and the operation relatively difficult . For the above reasons, we have proposed an innovative approach and protocol for the post-closure technique that is considered suitable for VA-ECMO extubation. To minimize the need for additional operations and reduce the risk of infection, a stapler was not used during VA-ECMO catheterization. At the time of extubation, the suture area was reduced by replacing the inappropriately sized vascular sheath with an appropriately sized vascular sheath, and the arterial access was adequately sutured. In animal models, this method is simple and effective for suturing the puncture point without affecting distal blood flow. In summary, the results of this study demonstrated that the Par-CT can be safely and effectively used for suturing the arterial access after weaning from VA-ECMO support. However, the following limitations were found during the study: 1. This study was a nonrandomized observational study and lacked a control group. 2. The low incidence of vascular complications in this study may be due to the small number of samples enrolled. 3. In addition, we did not provide long-term follow-up data, which prevented us from determining long-term vascular complication rates. |
An Overview of SARS-CoV-2 Molecular Diagnostics in Europe | 57fb7ee7-f166-47e4-8abf-91b0880b96ae | 8901364 | Pathology[mh] | • The COVID-19 pandemic has led not only to an influx of new molecular diagnostics but also a drive to modify existing technologies to allow the testing of thousands of patients daily over a variety of settings. • The need for rapid turn-around times for severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) testing for public health actions and patient care has led to the necessity for synchronously using multiple assays and platforms. • Testing solutions exist for any scale of SARS-CoV-2 testing strategy. • Overall SARS-CoV-2 molecular diagnostics seem to perform well; however, market saturation has left peer-reviewed real-world data lacking. • With these new developments, diagnostic testing regulations for SARS-CoV-2 are paramount to aid manufacturers in achieving assay performance and for laboratories to use as a tool alongside local verification to determine the suitability of assays and platforms for use in future epidemics. An emerging viral pneumonia of unknown etiology was detected in patients from several health care facilities in the city of Wuhan in China on 30 December 2019. A novel coronavirus was identified initially termed “2019-nCoV” and designated as severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) with the clinical disease termed “coronavirus infectious disease-19” (COVID-19). , , , It has overwhelmed health care systems globally due to rapid asymptomatic spread and lethality leading the World Health Organization (WHO) to declare a COVID-19 pandemic on 11 March 2020. , , SARS-CoV-2 is a betacoronavirus and one of the seven known members of the Coronaviridae family. , It is an enveloped positive-strand RNA virus (single linear RNA segment) with a genome length of 29,881 bp (GenBank no. MN908947). Its genome has 14 open reading frames (ORFs), which encode for 28 different proteins—4 structural proteins such as the S (spike), E (envelope), M (membrane), and N (nucleocapsid) proteins; 16 nonstructural proteins (NSP 1–16); and 8 accessory proteins as shown in . The genome commences with a 5′ untranslated region (UTR), then the replication complex (ORF1a and ORF1b) followed by the four structural proteins and 3′ UTR, ending with nonstructural ORFs and a poly(A) tail. , ORF1a contains 10 NSPs, while ORF1b contains 16 NSPs. The combination of ORF1a and ORF1b codes for polyproteins pp1a and pp1b that form the viral replication complex. , Structurally, the RNA genome is bound by the N protein, while the S, E, and M proteins together create the double-layered lipid viral envelope. The principle genes of diagnostic significance are the RdRp (NSP-12), various ORF1ab regions, and the viral structural proteins (S, E, and N). The early sequencing of the SARS-CoV-2 genome and subsequent distribution of the genome sequence via Global Initiative on Sharing Avian Influenza Data (GISAID) enabled the development of nucleic acid amplification tests (NAATs), which became the cornerstone for the diagnosis of SARS-CoV-2. Although that is not the only molecular diagnostic technique, real-time polymerase chain reaction (RT-PCR) has become the mainstay across Europe with only limited use of other molecular techniques such as transcription-mediated amplification (TMA) or CRISPR. , One of the first published RT-PCR assays originated from Europe in January 2020 with primer probe sets targeting the E, N, and RdRp genes. The RdRp assay included a Pan Sarbecco probe that detected SARS-CoV-1, SARS-CoV-2, and Bat-SARS-related-CoV with a second probe specific to SARS-CoV-2 leading to the recommendation of using the E gene assay as the first-line screening tool, followed by confirmatory testing with the RdRp gene assay. A further assay was quickly developed by the Centers for Disease Control and Prevention (CDC) targeting multiple regions of the N gene, which has become the baseline assay for several commercially available molecular diagnostic tests. , , , At the start of the COVID pandemic, in vitro medical devices (IVD), including NAAT-based systems and assays, needed to comply with European Union Directive 98/79/EC In Vitro Diagnostic Directive (IVDD) and bear a Conformitè Europëenne ( CE) symbol as proof, to be marketed in European Union (EU) and European Free Trade Association countries and Turkey and the United Kingdom. , CE marking required the manufacturer to have verified compliance with legal requirements and prepared an EC declaration of conformity containing the device performance and safety data. This allowed the device to be CE marked if it was intended for use by health care professionals although specific national requirements may also have been required. Although the United Kingdom left the EU in 2020, it will still accept CE-marked kits until 2023 when the UK Conformity Assessed mark will be required to market IVDs in the United Kingdom. Under Directive 98/79/EC, devices could also be granted emergency market access in the interest of health protection, such as in the COVID-19 pandemic; this required a derogation to be issued by the competent authority of a country allowing temporary marketing of a device without a full declaration of conformity, which was valid only for that nation. , As of May 2021, Directive 98/79/EC was replaced in the EU by Regulation (EU) 2017/746, which expands the risk-based device classification system alongside a requirement for device assessment by independent third parties and confirmation of test performance by EU reference laboratories before a CE mark is awarded. All products currently on the market that comply with the old legislation will have to recertify according to the new regulations. , Regulation (EU) 2017/746 still allows the national emergency market access of IVDs in the interest of protection of health if the derogation is issued by the country’s competent authority. This change in regulation brings CE marking more in line with the more stringent Food and Drug Administration ( FDA ) approval process, which requires devices to be tested by clinical trial and licensed only for use in specific circumstances. On 17 June 2021, the UK government announced the intention to introduce a mandatory validation scheme initially for COVID-19 diagnostics to expand to cover all devices sold in the United Kingdom. This process would require manufacturers to provide a minimum set of standard performance data, which would undergo independent verification by specially commissioned laboratories. If successfully introduced, it would be a criminal offense to market devices that have failed or not undergone this mandatory validation in the United Kingdom under the Medicines and Medical Devices Act 2021. The above pieces of legislation along with the European Commission’s guidelines for the Current Performance of COVID-19 Test Methods and Devices and Proposed Performance Criteria state the performance characteristics for IVDs, which includes but is not limited to analytical and diagnostic sensitivity and specificity, limits of detection (LODs), and expected values in normal and affected populations. , , No required values for these characteristics are published in these documents although common specifications are planned. A list of CE-marked COVID-19 IVDs is maintained at the European commission’s Joint Research Centre In Vitro Diagnostic Devices and Test Methods Database. As of 08/06/2021 325 CE-marked NAATs exist in this database originating from 240 unique manufacturers with 31 countries of origin. This database lacks key performance criteria for a significant number of entries including 120 tests with no stated LOD, 226 with no analytical sensitivity, 209 with no analytical specificity, and 200 with no clinical accuracy data. The entrance of many nontraditional manufacturers to the market has fueled a lack of peer-reviewed publications that make assessment of real-world performance difficult. An improved and standardized approach to market regulations would be welcomed as at present local validations/verifications of diagnostics are hugely important in ensuring the suitability of test selection for the intended purpose. In addition to CE marking, the WHO and national bodies such as the UK Medicines and Healthcare products Regulatory Agency (MHRA) have published target product profiles (TPPs) that outline performance characteristics that a test must meet to be considered successful for its intended use. , , , WHO and MHRA TPPs outline “acceptable” and “desirable” characteristics including ranges for parameters such as analytical sensitivity/LOD and clinical sensitivity. , , These documents are not legally binding but were developed to aid manufacturers in achieving assay performance that would be desired for use in the field. Equally these documents can be used by laboratories as a tool alongside local verification to determine the suitability of an assay for use. A selection of characteristics for NAAT-based tests is listed in and with the MHRA TPP showing much stricter acceptable criteria than the WHO criteria recommended for adoption by European Centre for Disease Prevention and Control (ECDC). , , , The scale of testing required to manage the SARS-COV-2 pandemic has been unprecedented with extensive yet flexible testing strategies being key to protecting public health through prompt isolation of cases. , The United Kingdom has undertaken a dual-arm approach to testing with twice weekly at home rapid antigen tests being freely available and actively encouraged in the asymptomatic general population and in laboratory NAAT being used for more sensitive screening of all hospital admissions including day case and those with symptoms consistent with COVID-19. , The ECDC not only recommends the use of NAAT for all symptomatic cases but also acknowledges the role for rapid antigen tests in population screening. , The use of sensitive molecular diagnostic assays is important to the control of transmission. If SARS-CoV-2 infection is allowed to spread unchecked, the emergence of novel variants is likely to be enhanced as mutations in key genes continue to accumulate as part of the natural error-prone replication of RNA viruses. As mutations accumulate, it is not only possible that they can lead to increased pathogenicity or vaccine escape, but that they may also lead to detection failures in well-established diagnostic assays. It is now recommended that the presence of SARS-CoV-2 in clinical samples is determined through the detection of at least two distinct targets to mitigate this risk. The observation of the ThermoFisher S gene PCR assay failure in the United Kingdom for the B.1.1.7 Alpha variant, which would have led to significant numbers of false-negative tests being reported if this was being used as a single target assay, highlights the importance of a multi-target approach. To achieve testing on such an immense scale testing, a diverse approach has been required with laboratories often using multiple assays and platforms in unison. The following is by no means an extensive review of all diagnostic assays used in Europe but aims to provide an overview of some of the most common. Rapid antigen near patient point of care and isothermal amplification techniques are outside the scope of this review but will be covered elsewhere in this Clinics edition. Rapid, commercial, cartridge-based sample-to-answer molecular diagnostic platforms for the detection of SARS-CoV-2 have fulfilled an important niche in point-of-care settings and clinical laboratories. They are simple to use, provide accurate results within 1–2 h, have minimal hands-on time, and permit on-demand testing of urgent specimens. An overview of the main sample-to-answer platforms is presented in . These single-use tests often automate nucleic acid extraction, purification, amplification, detection, and interpretation of results. All the platforms presented are internally controlled yet only three use an endogenous sample control, which monitors for an adequately taken sample and sample degradation. Independent studies evaluating the performance of rapid RT-PCR tests have varied with few head-to-head comparisons although evaluations of these platforms are more extensively published due to their widespread use in non-specialist laboratories. Unlike other applications, the rapid testing platforms exhibit significant variation in the technologies used. Cepheid Xpert Xpress, QiaStatDx, and VitaPCR SARS-CoV-2 rely on classic multiplex RT-PCR. Novodiag COVID-19 is unique in its use of qPCR and microarray technology for the detection of SARS-CoV-2. GenomEra SARS-CoV-2 and GenomEra SARS-CoV-2 with Flu A/B+ RSV use multiplex RT-PCR performed on chips. BioFire Respiratory Panel 2.1 plus (RP2.1plus) achieves extensive multiplexing through an initial RT-PCR step before target amplification using numerous monoplex PCR reactions, which are detected using endpoint melt curve analysis. GenMark ePlex SARS-CoV-2 and GenMark ePlex Respiratory Pathogen Panel 2 (RP2) use RT-PCR in combination with electrowetting and GenMark’s eSensor technology involving electrochemical detection rather than optical detection of fluorescence. Aside from the variation in technologies, the rapid testing platforms also offer detection of the widest range of pathogens. With the exception of Luminex Aries, SARS-CoV-2 can be detected in isolation or in combination with influenza as a minimum. , BioFire RP2.1plus detects 23 respiratory pathogens, GenMark ePlex RP2 detects 25 respiratory pathogens, and the QIAstat-Dx Respiratory SARS-CoV-2 Panel detects 22 respiratory pathogens. Xpert Xpress SARS-CoV-2 is the most widely evaluated rapid test with a recent systematic review and meta-analysis encompassing 1734 subjects determining a pooled sensitivity of 99% (97–99, 95% CI) and a specificity of 97% (95–98, 95% CI). Reported sensitivities for other platforms range from 90 to 100% with particular issues noted for samples with high cycle threshold (Ct) values in some studies. , , , , Fitoussi and colleagues (2021) found a VitaPCR SARS-CoV-2 sensitivity of 60% for samples that were positive at Ct > 33 using a comparator N gene assay; however, VitaPCR involves no formal RNA extraction and purification that may account for this poor performance. All tests in were shown to be near 100% specific except for the VitaPCR SARS-CoV-2 and QIAstat-Dx. , The VitaPCR gave a specificity of 94.7% in one study due to its increased sensitivity over the comparator assay, and a second study showed an improved sensitivity of 99%. , The QIAStat-Dx gave a specificity of 93% compared with a WHO-recommended RT-PCR. Evaluations often used small sample sets, due to a limited availability of reagents and used various SARS-CoV-2 reference controls, making LOD comparisons difficult. Reported LODs varied from 100 copies/ml for Xpert Xpress SARS-CoV-2 to 3000 genome copy equivalents for the Aries SARS-CoV-2. Several platforms fail to achieve the MHRA TPP “acceptable” LOD criteria of 1000 copies/ml; GenomEra SARS-CoV-2, Flu A/B+ RSV at 2857 copies/mL, Novodiag COVID-19 at 1815 copies/mL when using collection devices other than the provided medium nucleic acid amplification test; and both the GenMark ePlex SARS-CoV-2 and the QIAstat-Dx Respiratory SARS-CoV-2 Panel at 1000 copies/ml. The main limitations of the rapid sample-to-answer platforms include their high cost per test and low sample throughput. Moreover, despite their low complexity, rapid platforms are not infallible, and they are sensitive molecular tests that can be compromised without meticulous sample processing and good laboratory practice. Notably, BioFire and ePlex platforms do not output Ct values, meaning there is no indication of SARS-CoV-2 viral burden that can be of interest to the clinician as higher viral loads have been associated with increased SARS-CoV-2 mortality. One of the biggest barriers to the implementation of SARS-CoV-2 testing in non-specialist laboratories early in the pandemic was the availability of the correct equipment to enable the rapid introduction of testing. The solution to this problem for many manufacturers was the rapid introduction to the market of stand-alone assays encompassing kits, which include the reagents necessary for reverse-transcription PCR, including controls, but that are not tied to a specific extraction or PCR platform. They offer flexibility over more “closed” systems as they can potentially be run on existing instrumentation, precluding the requirement for purchasing new and often expensive equipment. Use of such reagents requires more extensive validation than end-to-end systems, and the onus on providing this validation, including sample preparation and the compatibility of any instrumentation with a particular kit, will fall on the individual laboratory. Some suppliers provide details of compatible platforms, but many do not, and it is this lack of data that have allowed many substandard kits to enter the market. Over 200 CE-marked manual RT-PCR kits are listed on the COVID-19 In Vitro Diagnostic Medical Devices database, a selection of which are shown in along with some of their main attributes. , , , , , , , , , , , , , , , , , , Kit formats are broadly similar and include minimal necessary reagents (primer/probe mixes, controls). Reagents may be provided either lyophilized or “wet” most commonly in tubes but also as eight-well strips. Although earlier kits relied on a single viral gene target, these have now been largely superseded by dual or triple target assays that focus on some combination of the E, N, S, and Orf1a genes. Although this has made the assays more robust in dealing with the emergence of novel SARS-CoV-2 variants, it has also complicated the interpretation of results when some gene targets fail to amplify. Furthermore, most kits supply an internal control (IC), which may be either endogenous (eg RNase P) , , , , or exogenous (eg MS2), , , which can be used either as full process controls or solely as PCR controls. Some kits include both endogenous and exogenous ICs although some fail to disclose the IC origin. , , , , , , , , , The number of tests per kit ranges from 48 to 4800 allowing for a wide range of throughputs although this will also depend on the number of wells required per sample and whether they are being tested in 96- or 384-well format. Many assays exploiting RT-PCR can typically use up to four different fluorescent reporter dyes, including the IC, but others are not so comprehensively multiplexed and require two or even three wells for each sample. At least one kit (Menarini) uses melt curve analysis in preference to hydrolysis probes, negating the requirement for multiple fluorescent reporter dyes. Although not shown in , many SARS-CoV-2 kits are also formulated as multiplexes with other respiratory viruses, most commonly influenza and respiratory syncytial virus (RSV), for example, Altona, Viasure, and ThermoFisher. This will usually require the addition of an extra well for each sample and/or the use of a single dye for multiple gene targets of the same virus. The actual throughput for these assays will depend heavily on the extraction and PCR equipment chosen for use and the level of automation. Use of an automated end-to-end system like the Roche FLOW could produce in excess of 1000 results in a 24hr period from experience in our local laboratory. Owing to the pressure to manufacture diagnostic kits rapidly as the pandemic took hold, much of the technical and clinical validation data used minimal data sets. Unlike the rapid platforms that are in widespread use, peer-reviewed literature is sparse for many stand-alone kits and in some cases completely absent. For those referenced assays in , the LOD was most commonly in the range of 1–20 copies/reaction although this was liable to small variations depending on the extraction and eluate volume and the volume of eluate used in the PCR. When comparisons between kits using clinical samples or External Quality Assurance (EQA) samples were performed, most kits performed comparably with only small variations in results between the Altona, , , , , , , Integrated DNA Technologies (IDT), , Seegene, , , , TaqPath, , , Viasure, and Tib MolBiol kits. , Specificity was 100% in virtually all cases. Stand-alone kits offer a convenient alternative to more closed systems allowing rapid implementation on existing equipment. However, despite a broad agreement in the performance of these assays on clinical specimens, the sheer number of kits available means that in-house validation is essential before implementation as a clinical service. The use of stand-alone PCR kits is not always an attractive option for laboratories, particularly if the existing molecular diagnostic infrastructure is not in place. Manufacturers identified a niche in the market for automated low-to-medium input end-to-end solutions, which could be easily introduced to laboratories with minimal molecular diagnostic experience. All platforms assessed here use multiplex RT-PCR with all assays containing an IC except the Virokey SARS-CoV-2, which contains neither an endogenous nor manufacturer-provided IC . False-negative results will not be identified by the failure to include an IC to demonstrate either sample adequacy or PCR failure. The Qiagen NeuMoDx has the best throughput of these systems at 435 samples in 24hr and also has the advantage of being a true random access platform with a quick time to result of only 1hr 25 min. Peer-reviewed literature for these platforms is significantly lacking over all other investigated areas with most performance data presented here being sourced from the manufacturer’s literature. The BD MAX system can use a variety of kits from different manufacturers including SARS-CoV-2 in isolation or with other respiratory pathogens such as influenza. The BD MAX SARS-CoV-2 assays, including the ViaSure SARS-CoV-2 N1 + N2 assay, have repeatedly shown 100% sensitivity but the specificity of greater than 95% both in manufacturers post-market surveillance and in real-world data. Fears around the production of false-positive results led the FDA to release a product notice recommending confirmation of all positive results generated by the BD MAX; however, both of the assessed assays are based on the CDC N gene assay, which has been shown to be highly sensitive. , , The Amplidiag COVID-19 assay was highly sensitive showing greater than 98% agreement compared directly with Cobas 6800 SARS-CoV-2. All other assessed platforms as shown in were also found to have acceptable sensitivity and specificity of greater than 96% based on manufacturer’s data only. , , , , , All assessed platforms were shown to have good analytical sensitivity as outlined in with the exception of Aus Diagnostics SARS-CoV-2, influenza, and RSV, which has an LOD on 2150 to 4325 copies/ml. Real-world testing of the Amplidiag COVID-19 also highlighted a failure to detect an EQA sample at 3300 copies/ml suggesting the manufacturer published LOD of 313 copies/ml may not be reliable. Local verification of the manufacturer’s claims is important before the introduction of any test into routine use to ensure discrepancies such as this are detected. The expected 24hr throughput for these systems is modest, and these systems are likely to be sited in laboratories that do not undertake 24/7 working meaning their full potential cannot be met. Although this may be the case, these automated solutions can offer easy-to-use solutions for laboratories with limited molecular experience. This has been important in providing the ability to decrease time to result over sending samples to specialist reference laboratories for testing, which in turn can reduce transmission risk particularly in health care settings. Several high-throughput platforms have been introduced for the detection of SARS-CoV-2 RNA offering end-to-end automated testing of samples from nucleic acid extraction through to amplification and detection. The introduction of high-throughput screening platforms into laboratories can improve laboratory efficiency and turnaround times while reducing staff hands-on time and facilitating a substantial increase in a testing capacity. The main high-throughput testing platforms and associated assays are listed in . All are RT-PCR-based assays except the Hologic Aptima SARS-CoV-2 assay that use TMA. All assays listed use a minimum of two different SARS-CoV-2 targets to reduce the risk of false negatives due to primer/probe mismatches caused by sequence variability. Multiple comparisons between the high-throughput platforms and standard RT-PCR demonstrate a high level of diagnostic performance. The Panther Fusion had an overall agreement of 96.4% compared with the Roche Cobas 6800 SARS-CoV-2 assay with a similar finding in a separate study. An agreement of 98.3% was found when comparing the Cobas to the Abbott Alinity M SARS-CoV-2 AMP, and in a three-way comparison between these platforms and the Panther Fusion, the overall agreement was 99.7%. When the TMA-based Aptima assay was compared with the Panther Fusion and rapid low-throughput BioFire Defense COVID-19 test, it produced a positive percent agreement of 98.7% compared with the consensus and a 100% agreement for negative results. Comparing analytical sensitivity is difficult due to differences in methods between studies, but generally all have high analytical sensitivities with LODs of 200 copies/ml or below, as collated from several studies and listed in . The TMA-based Aptima assay was shown to have a lower LOD when compared with standard RT-PCR, although when compared directly against the Roche Cobas and Abbott m2000, the Cobas test had the lowest LOD, a similar finding when the Cobas was directly compared with the Abbott m2000 and Panther Fusion. All systems offer a throughput of 1000 samples or more in a 24hr period. The highest throughput systems are the Roche Cobas 8800 system and the recently introduced Thermofisher Amplitude running the Taqpath COVID-19 assay, which claims a very high throughput of 8000 samples from a single platform over 24 hours. The Taqpath COVID-19 assay has been evaluated as a standard RT-PCR assay, but no published data exist for the diagnostic performance of the complete Amplitude system. Assays for these high-throughput platforms are being updated to include additional respiratory targets to meet the predicted increases in RSV and seasonal influenza infections once nonpharmaceutical interventions for COVID are removed. These include the Roche Cobas SARS-CoV-2 and Influenza A/B for the 6800/800 systems, the Aptima SARS-CoV-2/Flu Assay for the Hologic Panther system, and the m RESP-4-PLEX ASSAY for the Abbott Alinity system. The Cepheid GeneXpert infinity platform can give users the option to run up to 80 Xpert Xpress SARS-CoV-2 cartridges simultaneously with no increase in run time over the smaller cepheid instruments making this a high-throughput low complexity solution for laboratory settings. All viruses mutate, particularly RNA viruses, and the infection rate of SARS-Cov-2 on a large susceptible population has greatly increased the opportunity for mutations to occur. These mutations have led to variants of concern (VOCs) emerging with the potential of enhanced fitness, specifically toward increased transmissibility , and vaccine evasion. , , , , The first VOC (B.1.1.7—Alpha) was detected in the south of England and sequenced in September 2020. Soon after, new VOCs were identified from various locations across the world, each VOC becoming a prominent strain within their area of origin. Genomic sequencing is an invaluable tool in managing the pandemic due to its ability to detect unknown variations, which may indicate the emergence of a new VOC and the need for the development of new diagnostic assays. The United Kingdom currently sequences all SARS-CoV-2-positive samples where it is technically achievable; however, it can be slow, technically demanding, and currently has limited global availability. One solution to identifying known SARS-CoV-2 lineages without the need for genomic sequencing is the development of real-time genotyping PCR assays. Rapid real-time genotyping PCR assays usually target a single nucleotide polymorphism (SNP), with the most discriminatory targets often located within the S-gene. These types of mutations invariably lead to nonsynonymous amino acid substitutions. SNPs within this region can cause changes in the receptor-binding motif with successful variants retaining an increased affinity of the S-protein to the human angiotensin 2 receptor (ACE2). , , , Identification of these distinct mutations can be used as markers to detect specific VOC lineages. It is often the case that one distinct mutation may be present in several VOCs. For example, the presence of the N501Y mutation alone can be distinctive of the B.1.1.7 lineage, but the N501Y is also present in the B.1.351 and P1 VOC alongside the E484 K and K417 N or K417 T mutations, respectively; although the E484 K mutation is also occasionally seen in the B.1.1.7 lineage. It is often necessary to assay multiple targets to reliably determine the likely SARS-CoV-2 lineage. The range of SNP assays used will need to be modified as the new VOC are identified through whole-genome sequencing strategies. Public Health England currently uses the Applied Biosystems (Waltham, Massachusetts, USA) RT-PCR genotyping assay for the rapid detection of variants. This genotyping assay has a sufficient repertoire of target mutations to reliably cover all the major VOC currently recognized by the WHO and most of the variants of interest. , The current selection consists of 32 assays that can detect 30 SNPs and 2 deletions. Each assay is duplex in format detecting the mutant and the original SARS-CoV-2 reference/wild-type sequence on two different fluorescent dye layers. The high specificity of each assay target results in a significant reduction in the sensitivity, and it is advised by the manufacturer to only use extracted RNA from specimens with a CT of ≤30 where this information is available. There are several VOC assays in development or in early stages of marketing as shown in , many of which exist in stand-alone format to allow a reactive and rapid introduction of new SNP assays to the market as dictated by circulating variants. Agena Bioscience has developed the MassARRAY SARS-CoV-2 Variant Panel capable of detecting 15 variants over 36 gene targets in a two-well multiplex end-point RT-PCR assay. The use of SNP genotyping assays for the detection of SARS-CoV-2 VOC can be an effective early warning system for emerging VOC within a population, with quicker turnaround times compared with genomic sequencing. Data produced from this method can help scientists to quickly predict the prevalence of a VOC within a given population and may provide evidence toward vaccine effectiveness for new variants when collated with data regarding new infections or hospitalizations. The COVID-19 pandemic will have a long-reaching impact on molecular diagnostic testing. The speed at which molecular diagnostics entered the market has been unrivaled with strategies suitable for all desired testing throughputs available within a few short months. The overall analytical and clinical accuracy data for solutions marketed within Europe have generally been found to be satisfactory although published LODs can be variable. At the outset of the pandemic manufacturers, claims were not required to be independently verified in Europe, and outside the most used rapid or high-throughput testing platforms, peer-reviewed real-world data are sparse. Welcome changes to regulations for devices in Europe are on the horizon, but local laboratory validations will still play a key role in the future. With the increasing prevalence of new SARS-CoV-2 VOC and the need for enhanced surveillance, there is still potential for new developments in SARS-CoV-2 molecular diagnostics. • evere acute respiratory syndrome coronavirus-2 (SARS-CoV-2) required the rapid expansion of virological diagnostic techniques to ensure adequate testing capacity in the pandemic settings. • Rapid, molecular diagnostic platforms fulfill an important niche in point-of-care settings and clinical laboratories. They provide quick accurate results require minimal hands-on time and permit on-demand testing of urgent specimens, which is pertinent for non-COVID patient care. • High-throughput platforms improve laboratory efficiency and turnaround times while reducing staff hands-on time. This leads to an increase in the testing capacity of diagnostic laboratories to help meet the clinical demand throughout pandemics. • The use of SNP genotyping assays for the detection of SARS-CoV-2 VOCs can be an effective early warning system for emerging VOCs within a population, with faster turnaround times compared with genomic sequencing. This can assist with public health surveillance and provide high-quality evidence toward vaccine effectiveness. |
Factors Contributing to Uptake of Stillbirth Evaluations: A Qualitative Analysis | b5dc6be8-7e4c-4f43-89b9-94920d817628 | 11879755 | Forensic Medicine[mh] | Introduction Approximately two million stillbirths occur around the world each year . Determining stillbirth etiology is frequently done based on clinical history and observation, such as an external examination of the body. However, due to a lack of a single systematically applied protocol, clinical diagnostic approaches vary across institutions and often do not include standardised evaluation metrics . This lack of uniformity can lead to a misdiagnosis based on preliminary clinical presentation. Understanding the cause of stillbirth is important not only to help researchers and physicians reduce incidence but also to help facilitate bereavement and decrease emotional duress . Foetal autopsy, placental histology, and genetic testing are the most useful evaluations for assessing stillbirth . Yet, despite strong recommendations from the American Congress of Obstetricians and Gynaecologists (ACOG) , only about one fifth of stillbirths in the U.S. undergo perinatal autopsy . Identifying factors contributing to a stillbirth not only helps focus care for subsequent pregnancies and target prevention strategies; it improves mental health . Stillbirth often leaves parents with increased anxiety, depression, and feelings of guilt or shame surrounding their loss . Parents are also at increased risk for experiencing anxiety during pregnancies that follow a loss . The RESPECT study , a large multi‐country study, identified the most important factors in quality bereavement care. They stressed, “Make every effort to investigate and identify contributory factors to provide an acceptable explanation to women and families for the death of their baby.” as one of the top priciples . The objective of this study was to explore individuals' beliefs, values, and experiences surrounding stillbirth evaluation decisions. Here we interviewed parents about their stillbirth experience and identified the barriers and facilitators to uptake of stillbirth evaluations. Methods This descriptive research used semi‐structured interviews that were analysed using content analysis to gather in‐depth information about the stillbirth experience, as well as factors surrounding stillbirth evaluation decisions. Part of the stillbirth experience is deciding whether to have any evaluations conducted to determine the cause of death, such as foetal autopsy, placental pathology, or genetic testing. A semi‐structured interview guide was created based on published data and the clinical expertise of the research team (Table ). 2.1 Participants Participants were identified through medical record abstraction within a national research consortium on stillbirth (SL). Ninety‐four patients who met the following inclusion criteria were sent an invitation via email from the obstetrics clinic where they received care: (1) experienced a stillbirth within the last 5 years (mothers or fathers individually); (2) the stillbirth occurred at the University of Utah; (3) the patient had previously consented to being contacted for future research; (4) was at least 18 years old at the time of the interview; and (5) able to communicate in English. Two weeks after the email invitation was sent, patients were contacted by a member of the research team (SL) via telephone to determine their interest in the study if they had not already responded to the email invitation. Details about the study and its voluntary nature were reiterated over the phone prior to the interview. Enrollment for interviews concluded when data saturation was reached (i.e., no new information was added from additional interviews). 2.2 Data Analysis Interviews were conducted over the phone by a single interviewer (SL or NR), audio‐recorded (February to May 2021), professionally transcribed, and the transcripts were uploaded to the software Dedoose 9.0.17 . Inductive content analysis was conducted by identifying codes from within the transcripts and systematically designating data segments that contain similar material or themes to the remaining transcripts. This coding methodology was based on prior work . One member of the research team (NR) generated the original codes (e.g., stillbirth evaluation uptake barriers). These codes were systematically applied to the remaining transcripts, with additional codes added as necessary. An independent coder (ER) reviewed data for accuracy. Discrepancies were resolved through discussion until a consensus was reached. ER, a qualitative research expert, trained SL and NR in conducting interviews and conducting the content analysis. Our study follows suggested standards for reporting qualitative research . Participants Participants were identified through medical record abstraction within a national research consortium on stillbirth (SL). Ninety‐four patients who met the following inclusion criteria were sent an invitation via email from the obstetrics clinic where they received care: (1) experienced a stillbirth within the last 5 years (mothers or fathers individually); (2) the stillbirth occurred at the University of Utah; (3) the patient had previously consented to being contacted for future research; (4) was at least 18 years old at the time of the interview; and (5) able to communicate in English. Two weeks after the email invitation was sent, patients were contacted by a member of the research team (SL) via telephone to determine their interest in the study if they had not already responded to the email invitation. Details about the study and its voluntary nature were reiterated over the phone prior to the interview. Enrollment for interviews concluded when data saturation was reached (i.e., no new information was added from additional interviews). Data Analysis Interviews were conducted over the phone by a single interviewer (SL or NR), audio‐recorded (February to May 2021), professionally transcribed, and the transcripts were uploaded to the software Dedoose 9.0.17 . Inductive content analysis was conducted by identifying codes from within the transcripts and systematically designating data segments that contain similar material or themes to the remaining transcripts. This coding methodology was based on prior work . One member of the research team (NR) generated the original codes (e.g., stillbirth evaluation uptake barriers). These codes were systematically applied to the remaining transcripts, with additional codes added as necessary. An independent coder (ER) reviewed data for accuracy. Discrepancies were resolved through discussion until a consensus was reached. ER, a qualitative research expert, trained SL and NR in conducting interviews and conducting the content analysis. Our study follows suggested standards for reporting qualitative research . Results Nineteen parents were interviewed. The average age of participants at the time of their stillbirth was 31.1 years (Table ). The number of children participants reported having, including losses, ranged from 1 to 13, with an average of 4.4 children. Participants tended to be well‐educated, with 48.0% having a bachelor's degree or higher, and at least 48.0% of participants had an income higher than the median annual Utah household income of $71.6 k . Seventeen of the 19 participants were offered one or more stillbirth evaluations. Of those, 11 reported that they chose to undergo an autopsy, three, placenta histology, eight, genetic testing, three declined all examinations, and two were not offered any. 3.1 Facilitators to Stillbirth Evaluation We asked participants why they did or did not consent to fetal autopsy, placental histology, and genetic testing. The most commonly reported reason was due to personal values and beliefs. For example, having a strong belief in science, wanting the information to inform future pregnancies, altruism, or they simply wanted to know why. The following are examples, with quotes from participant interviews, of the values and beliefs that contributed as facilitators to the stillbirth evaluation decision. "I really wanted the autopsy. For me that wasn't weird, to be offered that, just because I have more of a medical‐based occupation. I have dissected cadavers, and I like knowing the reasons why. I was like, Why did this happen?" Participant 4 "I worked in medical field for almost 20 years, and my husband is very much into finding out as much weak information we could if we—for future pregnancies, so we definitely wanted to know what had gone wrong and answer some questions that we had in our own minds about what had gone wrong." Participant 5 "I remember that I pretty enthusiastically agreed because I believe in science, and having more information is helpful to me on a personal level. Also, HELLP syndrome in particular is still being researched." Participant 6 Parents who chose placental histology or genetic testing but declined fetal autopsy also stated how they desired some understanding of the cause, yet felt protective of their baby. "Ultimately, my husband and I decided that we didn't want an autopsy done on her and that, due to the genetic testing coming back with no real answers, there would be no need to find out what was wrong with her." Participant 13 The information parents received also contributed as a facilitator to consenting to stillbirth evaluation(s). When asked how evaluations were offered, responses ranged from not receiving any information to an in‐depth conversation supplemented with educational reading material. Those who chose to have at least one of the stillbirth evaluations ( n = 14) remembered more about the evaluation options offered and how the information was presented to them. For example, three participants received an informational pamphlet, and six were enrolled in a foetal autopsy study not related to this work. The six participants who opted to get an autopsy as part of another study were presented with the most information, and several expressed altruistic reasons for participating. "You know what? They had a clipboard, and they had some papers, and they pretty much talked me through the choices that I had. They explained to me about the study that they were doing and whether or not I wanted the autopsy." Participant 7 Additionally, one of the participants saw medical providers treat their baby with respect, which was expressed as a facilitator for choosing autopsy. "We had the time that we needed. They treated him just like a baby, even though he was the size of my hand and was dead. That was really helpful. That made the choice to do the autopsy that much easier because it didn't feel like, Oh yeah, okay, here; we're just gonna treat this like we're dissecting a frog in biology." Participant 9 3.2 Barriers to Stillbirth Evaluation Personal values and beliefs also were cited as the main reasons for declining evaluations, typically foetal autopsy. Participants said they declined one or more evaluations to protect their baby from the harm they imagined was caused by the procedure, to spend more time with their baby, because of cost, or because they believed they already knew the cause of death prior to being offered the evaluations. "She's my angel looking after me. I didn't want to put her through that." Participant 14 "I needed to spend the time with him and have the keepsakes of him taken care of, and I don't think I needed anything outside of that." Participant 17 "They did offer it [autopsy & genetic testing]. They said it was not covered by my insurance." Participant 15 "It was my cervix, so I had that answer. I didn't want to disturb her little body, so we decided not to." Participant 16 Medical providers were sometimes the barrier to obtaining a stillbirth evaluation. In several cases, the participant received a diagnosis before being offered any of the evaluations, which contributed to their decision to decline an evaluation. Additionally, some participants said their provider recommend against an evaluation, even if the participant asked for one. These participants also expressed lingering resentment towards their provider for not supporting their wishes. "It seemed like we had to push and pry to get testing done. ‘Well, if you guys are really worried about it, we could do an autopsy, but it's gonna cost money.’ It's like, well, how much? It's like $100. It's like, you kidding me? In the grand scheme of medical expenses, that's nothing." Participant 12 The last barrier identified was not recieving information about the evaluations. Two participants were not offered evaluation options. Nonetheless, they expressed that they would have liked to have been told about their options. "I think any knowledge, any option, is good because a lot of times you could go to a doctor if you have any questions, like now. Would you want to know? I don't know unless I'm offered." Participant 19 A summary of the facilitators and barriers can be seen in Table . 3.3 Satisfaction or Regret in Stillbirth Evaluation(s) Decision Sixteen of the 19 participants expressed satisfaction about their stillbirth evaluation decision regardless of their choice. The main reasons most participants gave for being satisfied with their decision was knowledge about the cause or because they felt they did everything they could. Several also expressed that the results helped them cope with the loss or that they felt relief from the results. Representative quotes from participants who opted for one or more evaluations: "Absolutely, yeah. Yeah, definitely. I'm now pregnant again, and I think it is helpful to have that information." Participant 1 "I feel like that one, we handled it the best. We did everything we could." Participant 3 "I would have regretted not having had that information." Participant 6 The following is a representative quote from a participant who declined evaluations: "Just because we already knew her condition and what she was diagnosed with. I truly just believed it was just a very rare situation. …I didn't want to put her through that." Participant 14 Most participants did not regret their stillbirth evaluation decision. "I guess, personally, I don't see the downside of the autopsy. It didn't seem like there was much of any noticeable expense to it. I wish I could say that, yeah, we got all this useful information from it. I don't know that we did, but at least I can look back on it and say, ‘You know what? At least we tried,’ or at least if there was something super obvious as to what happened, we would have found out." Participant 12 The only regret that was expressed was from one participant whose decision was not realised and from the two participants who did not receive information about their stillbirth options. "It definitely was [frustrating] to leave the hospital and not really have a definite answer on why it happened and how it could happen if I happen to get pregnant again." Participant 15 "I think at least—I would have like to know about the histology of the placenta. Even the genetic testing, I think that—I don't know if that would be able to tell me more or not, or the doctors more or not. We don't know unless we find out, unless we look." Participant 19 Facilitators to Stillbirth Evaluation We asked participants why they did or did not consent to fetal autopsy, placental histology, and genetic testing. The most commonly reported reason was due to personal values and beliefs. For example, having a strong belief in science, wanting the information to inform future pregnancies, altruism, or they simply wanted to know why. The following are examples, with quotes from participant interviews, of the values and beliefs that contributed as facilitators to the stillbirth evaluation decision. "I really wanted the autopsy. For me that wasn't weird, to be offered that, just because I have more of a medical‐based occupation. I have dissected cadavers, and I like knowing the reasons why. I was like, Why did this happen?" Participant 4 "I worked in medical field for almost 20 years, and my husband is very much into finding out as much weak information we could if we—for future pregnancies, so we definitely wanted to know what had gone wrong and answer some questions that we had in our own minds about what had gone wrong." Participant 5 "I remember that I pretty enthusiastically agreed because I believe in science, and having more information is helpful to me on a personal level. Also, HELLP syndrome in particular is still being researched." Participant 6 Parents who chose placental histology or genetic testing but declined fetal autopsy also stated how they desired some understanding of the cause, yet felt protective of their baby. "Ultimately, my husband and I decided that we didn't want an autopsy done on her and that, due to the genetic testing coming back with no real answers, there would be no need to find out what was wrong with her." Participant 13 The information parents received also contributed as a facilitator to consenting to stillbirth evaluation(s). When asked how evaluations were offered, responses ranged from not receiving any information to an in‐depth conversation supplemented with educational reading material. Those who chose to have at least one of the stillbirth evaluations ( n = 14) remembered more about the evaluation options offered and how the information was presented to them. For example, three participants received an informational pamphlet, and six were enrolled in a foetal autopsy study not related to this work. The six participants who opted to get an autopsy as part of another study were presented with the most information, and several expressed altruistic reasons for participating. "You know what? They had a clipboard, and they had some papers, and they pretty much talked me through the choices that I had. They explained to me about the study that they were doing and whether or not I wanted the autopsy." Participant 7 Additionally, one of the participants saw medical providers treat their baby with respect, which was expressed as a facilitator for choosing autopsy. "We had the time that we needed. They treated him just like a baby, even though he was the size of my hand and was dead. That was really helpful. That made the choice to do the autopsy that much easier because it didn't feel like, Oh yeah, okay, here; we're just gonna treat this like we're dissecting a frog in biology." Participant 9 Barriers to Stillbirth Evaluation Personal values and beliefs also were cited as the main reasons for declining evaluations, typically foetal autopsy. Participants said they declined one or more evaluations to protect their baby from the harm they imagined was caused by the procedure, to spend more time with their baby, because of cost, or because they believed they already knew the cause of death prior to being offered the evaluations. "She's my angel looking after me. I didn't want to put her through that." Participant 14 "I needed to spend the time with him and have the keepsakes of him taken care of, and I don't think I needed anything outside of that." Participant 17 "They did offer it [autopsy & genetic testing]. They said it was not covered by my insurance." Participant 15 "It was my cervix, so I had that answer. I didn't want to disturb her little body, so we decided not to." Participant 16 Medical providers were sometimes the barrier to obtaining a stillbirth evaluation. In several cases, the participant received a diagnosis before being offered any of the evaluations, which contributed to their decision to decline an evaluation. Additionally, some participants said their provider recommend against an evaluation, even if the participant asked for one. These participants also expressed lingering resentment towards their provider for not supporting their wishes. "It seemed like we had to push and pry to get testing done. ‘Well, if you guys are really worried about it, we could do an autopsy, but it's gonna cost money.’ It's like, well, how much? It's like $100. It's like, you kidding me? In the grand scheme of medical expenses, that's nothing." Participant 12 The last barrier identified was not recieving information about the evaluations. Two participants were not offered evaluation options. Nonetheless, they expressed that they would have liked to have been told about their options. "I think any knowledge, any option, is good because a lot of times you could go to a doctor if you have any questions, like now. Would you want to know? I don't know unless I'm offered." Participant 19 A summary of the facilitators and barriers can be seen in Table . Satisfaction or Regret in Stillbirth Evaluation(s) Decision Sixteen of the 19 participants expressed satisfaction about their stillbirth evaluation decision regardless of their choice. The main reasons most participants gave for being satisfied with their decision was knowledge about the cause or because they felt they did everything they could. Several also expressed that the results helped them cope with the loss or that they felt relief from the results. Representative quotes from participants who opted for one or more evaluations: "Absolutely, yeah. Yeah, definitely. I'm now pregnant again, and I think it is helpful to have that information." Participant 1 "I feel like that one, we handled it the best. We did everything we could." Participant 3 "I would have regretted not having had that information." Participant 6 The following is a representative quote from a participant who declined evaluations: "Just because we already knew her condition and what she was diagnosed with. I truly just believed it was just a very rare situation. …I didn't want to put her through that." Participant 14 Most participants did not regret their stillbirth evaluation decision. "I guess, personally, I don't see the downside of the autopsy. It didn't seem like there was much of any noticeable expense to it. I wish I could say that, yeah, we got all this useful information from it. I don't know that we did, but at least I can look back on it and say, ‘You know what? At least we tried,’ or at least if there was something super obvious as to what happened, we would have found out." Participant 12 The only regret that was expressed was from one participant whose decision was not realised and from the two participants who did not receive information about their stillbirth options. "It definitely was [frustrating] to leave the hospital and not really have a definite answer on why it happened and how it could happen if I happen to get pregnant again." Participant 15 "I think at least—I would have like to know about the histology of the placenta. Even the genetic testing, I think that—I don't know if that would be able to tell me more or not, or the doctors more or not. We don't know unless we find out, unless we look." Participant 19 Discussion 4.1 Main Findings In the present study, the reasons expressed for consenting to an evaluation varied by type of examination. Parents who consented to foetal autopsy wanted to understand why their baby passed, approaching their decision with more deductive reasoning than emotions. Those who chose placental histology or genetic testing over foetal autopsy often didn't want to harm their baby, but desired information about the cause of their stillbirth. Other reasons for consenting to one or more of the stillbirth evaluations were a desire to understand whether they were to blame, to inform possible future pregnancies, and the respect shown by the medical team towards the stillborn and parents. Those who chose one or more of the evaluations often expressed a desire to contribute to the scientific knowledge about preventing future stillbirths and to help others who may have suffered a loss. Decision‐making for stillbirth evaluations is often impacted by emotions and parental readiness . This decision comes at a time when parents are grieving the loss of their baby, maternal physical exhaustion from the birth, or maternal impairment from anaesthesia or opioid medications for pain. This level of distress decreases one's ability to make decisions , while conflicting desires and needs (e.g., protecting their baby vs. wanting answers) further complicate decision‐making. Among participants in this study, this decision‐making aligned with the knowledge, values, and existing parental beliefs prior to the event. Unfortunately, misconceptions about medical evaluation contributed to declining one or more evaluations, most often foetal autopsy. 4.2 Limitations Patients were recruited from a single hospital; the majority were non‐Hispanic White, well educated, wealthier, and at least 40% belonged to The Church of Jesus Christ of Latter‐Day Saints. Results from this cohort may not reflect the general population; stillbirth occurs at higher rates among minoritized and socioeconomically disadvantaged groups . However, our participants shared experiences similar to two other research cohorts from around the world . Additionally, there may be recall bias due to the impact of the highly distressing nature of stillbirth. 4.3 Interpretation The reasons for declining stillbirth evaluations are varied and complex, with several notable differences across populations. Previous research identified several reasons parents decline evaluations including complexity of the consent, emotional stress of the situation, lack of information for families and providers, mistrust, protective parenting, and belief that no new information will be found . However, culture also may contribute to stillbirth evaluation decisions. For instance, a study in Malaysia found that among the Muslim population, religious tenets stipulated that autopsies can only be carried out on a stillborn less than 120 days gestation . Women interviewed in tertiary care centers in Blantyre, Malawi, Mansa, Mwanza, Tanzania, and Zambia expressed fear of blame for the stillbirth and the concequences of denying cultural traditions . In the US, participants from this study who chose to consent to stillbirth evaluations approached their decision with more deductive reasoning than emotions and tended to note how a science or medical background made them feel comfortable with their decision. On the other hand, some findings may apply across cultures. In this study, the most common reason stated for declining a stillbirth evaluation was having already received a probable cause of stillbirth before any of the examinations were offered. They believed that no new information would be found with further testing. However, even in the case of inconclusive evaluation results, our participants expressed that actively doing something for their baby or confirming that they were not responsible for the death was valuable to them and their healing process, consistent with other findings . Our participants were more often misinformed about foetal autopsy than other tests. Several participants expressed a desire to protect their baby from the harm inflicted by an autopsy without knowing what the procedure entails or the other evaluation options. Meaney et al. indicated that parents' misperceptions about the invasiveness of autopsies were based on the dramatisation seen on television . Even among a gr Italian physicians, half believed the baby would be dismembered . However, the autopsy exam may be done at varying levels of invasiveness, with targeted options that assess part of the body or radiographic‐only exams, which do not require any incisions . The PURPOSe study in India and Pakistan determined that using minimally invasive tissue sampling was reliable and found acceptable by the Muslim community . Genetic testing only requires blood or a small tissue sample, identifies copy number variants, and can be used to inform care in subsequent pregnancies . Even with a complete foetal autopsy, the incisions are created similarly to a surgery and stitched afterward . Clothing and a cap can cover the incisions if the patient desires an open casket funeral. These misconceptions can be addressed through better communication or educational materials. Decision aids are tools that support informed decision‐making in medical situations where there is often no “best” option . A decision aid for stillbirth evaluations, tailored for particular cultures, could explain the level of invasiveness of each examination, and present alternatives to complete autopsy, such as partial autopsy, computed tomography, ultrasonography, or magnetic resonance imaging . The creation and utilization of this decision aid would provide unbiased information, increase capacity for shared decision‐making between patients and their providers, and reduce decisional conflict of those faced with uncertainty. Hospital‐level factors contributing to stillbirth evaluation decision‐making include the necessity to make many other decisions concerning the stillbirth, a lack of provider training, limited perinatal pathologists, cost, and limited time available for medical providers to communicate with bereaved parents . Hospitals could increase the uptake of evaluations by educating providers on evaluation options, the procedures involved, and the importance of a correct diagnosis . Participants in our study unveiled several examples of how they were misinformed. For instance, some parents chose to decline evaluations because they erroneously thought that it precluded them from spending time with their baby. Another key area that could facilitate decision‐making, is educating clinicians and hospital staff on best practices for interacting with grieving parents . Within our cohort, some women shared gratitude about the care they received in the hospital. Yet others expressed frustration concerning interactions with providers, which lingered long after the stillbirth. Physician communication training programs have been successfully created in other medical fields, such as primary care . By giving clinicians the skills they need to communicate with patients about difficult topics, the patient is more likely to receive care that supports their values and needs. Finally, parents are asked to make numerous decisions in a small timeframe that they, and sometimes hospital personnel, are not adequately prepared to deal with. Some healthcare providers are not confident in the utility of examinations, such as autopsy, and therefore do not take the time to discuss this option . Parents in this cohort who were not offered or they were denied stillbirth evaluations expressed dissatisfaction with the medical system. Negative experiences like these can colour the stillbirth experience for parents for years afterwards. Main Findings In the present study, the reasons expressed for consenting to an evaluation varied by type of examination. Parents who consented to foetal autopsy wanted to understand why their baby passed, approaching their decision with more deductive reasoning than emotions. Those who chose placental histology or genetic testing over foetal autopsy often didn't want to harm their baby, but desired information about the cause of their stillbirth. Other reasons for consenting to one or more of the stillbirth evaluations were a desire to understand whether they were to blame, to inform possible future pregnancies, and the respect shown by the medical team towards the stillborn and parents. Those who chose one or more of the evaluations often expressed a desire to contribute to the scientific knowledge about preventing future stillbirths and to help others who may have suffered a loss. Decision‐making for stillbirth evaluations is often impacted by emotions and parental readiness . This decision comes at a time when parents are grieving the loss of their baby, maternal physical exhaustion from the birth, or maternal impairment from anaesthesia or opioid medications for pain. This level of distress decreases one's ability to make decisions , while conflicting desires and needs (e.g., protecting their baby vs. wanting answers) further complicate decision‐making. Among participants in this study, this decision‐making aligned with the knowledge, values, and existing parental beliefs prior to the event. Unfortunately, misconceptions about medical evaluation contributed to declining one or more evaluations, most often foetal autopsy. Limitations Patients were recruited from a single hospital; the majority were non‐Hispanic White, well educated, wealthier, and at least 40% belonged to The Church of Jesus Christ of Latter‐Day Saints. Results from this cohort may not reflect the general population; stillbirth occurs at higher rates among minoritized and socioeconomically disadvantaged groups . However, our participants shared experiences similar to two other research cohorts from around the world . Additionally, there may be recall bias due to the impact of the highly distressing nature of stillbirth. Interpretation The reasons for declining stillbirth evaluations are varied and complex, with several notable differences across populations. Previous research identified several reasons parents decline evaluations including complexity of the consent, emotional stress of the situation, lack of information for families and providers, mistrust, protective parenting, and belief that no new information will be found . However, culture also may contribute to stillbirth evaluation decisions. For instance, a study in Malaysia found that among the Muslim population, religious tenets stipulated that autopsies can only be carried out on a stillborn less than 120 days gestation . Women interviewed in tertiary care centers in Blantyre, Malawi, Mansa, Mwanza, Tanzania, and Zambia expressed fear of blame for the stillbirth and the concequences of denying cultural traditions . In the US, participants from this study who chose to consent to stillbirth evaluations approached their decision with more deductive reasoning than emotions and tended to note how a science or medical background made them feel comfortable with their decision. On the other hand, some findings may apply across cultures. In this study, the most common reason stated for declining a stillbirth evaluation was having already received a probable cause of stillbirth before any of the examinations were offered. They believed that no new information would be found with further testing. However, even in the case of inconclusive evaluation results, our participants expressed that actively doing something for their baby or confirming that they were not responsible for the death was valuable to them and their healing process, consistent with other findings . Our participants were more often misinformed about foetal autopsy than other tests. Several participants expressed a desire to protect their baby from the harm inflicted by an autopsy without knowing what the procedure entails or the other evaluation options. Meaney et al. indicated that parents' misperceptions about the invasiveness of autopsies were based on the dramatisation seen on television . Even among a gr Italian physicians, half believed the baby would be dismembered . However, the autopsy exam may be done at varying levels of invasiveness, with targeted options that assess part of the body or radiographic‐only exams, which do not require any incisions . The PURPOSe study in India and Pakistan determined that using minimally invasive tissue sampling was reliable and found acceptable by the Muslim community . Genetic testing only requires blood or a small tissue sample, identifies copy number variants, and can be used to inform care in subsequent pregnancies . Even with a complete foetal autopsy, the incisions are created similarly to a surgery and stitched afterward . Clothing and a cap can cover the incisions if the patient desires an open casket funeral. These misconceptions can be addressed through better communication or educational materials. Decision aids are tools that support informed decision‐making in medical situations where there is often no “best” option . A decision aid for stillbirth evaluations, tailored for particular cultures, could explain the level of invasiveness of each examination, and present alternatives to complete autopsy, such as partial autopsy, computed tomography, ultrasonography, or magnetic resonance imaging . The creation and utilization of this decision aid would provide unbiased information, increase capacity for shared decision‐making between patients and their providers, and reduce decisional conflict of those faced with uncertainty. Hospital‐level factors contributing to stillbirth evaluation decision‐making include the necessity to make many other decisions concerning the stillbirth, a lack of provider training, limited perinatal pathologists, cost, and limited time available for medical providers to communicate with bereaved parents . Hospitals could increase the uptake of evaluations by educating providers on evaluation options, the procedures involved, and the importance of a correct diagnosis . Participants in our study unveiled several examples of how they were misinformed. For instance, some parents chose to decline evaluations because they erroneously thought that it precluded them from spending time with their baby. Another key area that could facilitate decision‐making, is educating clinicians and hospital staff on best practices for interacting with grieving parents . Within our cohort, some women shared gratitude about the care they received in the hospital. Yet others expressed frustration concerning interactions with providers, which lingered long after the stillbirth. Physician communication training programs have been successfully created in other medical fields, such as primary care . By giving clinicians the skills they need to communicate with patients about difficult topics, the patient is more likely to receive care that supports their values and needs. Finally, parents are asked to make numerous decisions in a small timeframe that they, and sometimes hospital personnel, are not adequately prepared to deal with. Some healthcare providers are not confident in the utility of examinations, such as autopsy, and therefore do not take the time to discuss this option . Parents in this cohort who were not offered or they were denied stillbirth evaluations expressed dissatisfaction with the medical system. Negative experiences like these can colour the stillbirth experience for parents for years afterwards. Conclusion Stillbirth evaluations improve etiological understanding for parents, and the care providers and researchers trying to identify risk factors to prevent stillbirth. Two major barriers to autopsy consent were a misconception regarding prohibited time spent with the baby and that no new information would be identified from an evaluation. Providers could potentially improve uptake by educating and offereing stillbirth evaluations, supporting parents' wishes, and treating their baby with respect. Nathan Blue: writing – review and editing (equal). Erin P. Johnson: supervision; writing – review and editing (equal). Sarah Lopez: data curation (equal); writing – review and editing (equal). Jessica Page: writing – review and editing (equal). Naomi O. Riches: formal analysis (lead), writing – original draft (lead); writing – review and editing (equal). Erin Rothwell: conceptualisation (lead); formal analysis; writing – original draft; writing – review and editing (equal). Robert M. Silver: conceptualisation (equal); writing – review and editing (equal). Tsegaselassie Workalemahu: writing – review and editing (equal). Ethical approval for this study was granted by the University of Utah Institutional Review Board (IRB_00133359). Conflicts of Interest The authors declare no conflicts of interest. Table S1. Semi‐structured interview guide. |
Postgraduate medical education in obstetrics and gynaecology: Where are we now and what do we need for the future? A study on postgraduate training in obstetrics and gynaecology in Germany, Austria and Switzerland | daa785c1-9091-4ed5-8965-29ff413ca4bf | 9585411 | Gynaecology[mh] | High quality training is essential to become a specialist in obstetrics and gynaecology (OBGYN) and to ensure optimal patient care in the future. Every country in Europe defined specific goals and qualifications of training for graduate and postgraduate medical education (PGME) . Being individually defined and implemented in the countries, the curricula considerably differ in most parts of Europe, e.g. in the catalogue of requirements. Differences in training are often caused by differing medical infrastructure of most countries, diverging clinical responsibilities of subspecialties and the lack of protected timeslots for practical and theoretical training. For example, breast surgery which is mandatory content of training curricula in some countries, is not even being trained in others. In 2009, Rodriguez et al. found major differences in definition of training content and outcome among European trainees in OBGYN of every level of experience . The resulting need for harmonization of European training outcomes is reflected by the establishment of the pan-European curriculum for training in OBGYN by the European Board and College of Obstetrics and Gynaecology (EBCOG). The neighbouring German-speaking countries Germany, Austria and Switzerland vary substantially regarding their training curricula; however, they are highly comparable regarding medical infrastructure and health care systems. Therefore, these countries are ideal models to study the effects of different training curricula on educational outcomes and satisfaction of trainees. In this study, we aim to provide a representative picture of the current situation of training in OBGYN in Germany, Austria and Switzerland. Furthermore, we intend to determine transferable advantages of the different systems. We considered a survey to be the most appropriate method to get a comprehensive overview of the general situation and satisfaction of trainees, since it allows a large group of people to participate, even in case of limited time and financial resources and even if they are geographically diversified. Participants were enrolled anonymously. We performed the survey through a digital questionnaire with a total of 30 questions (see attachment 1 ). The questionnaire was designed by consensus of experts, all board members of a cooperation, including representatives of trainee networks of the German (DGGG), Swiss (SGGG) and Austrian (OEGGG) society for obstetrics and gynaecology. Advice was given by eight gynaecologists (FMW, KW, GB, MF, PF, MF, BK, MW) from Germany, Austria and Switzerland, who are partially involved in the development of national OBGYN curricula. The experts chose to focus on certain topics (such as simulation programs, training regulations or training in subspecialties) in order to provide a representative picture of the current situation of trainees but prevent a question overload of the survey. The questionnaire was set-up in German as online questionnaire allowing wide accessibility. The questions were presented sequentially. A new question was shown only if the preceding was answered. Thus, for completion of the survey the respondent had to answer all 30 questions. The survey was controlled and circulated by the Swiss Federal Institute of Technology in Zurich (ETH), Department for Health Sciences and Technology, Consumer Behavior and was advertised via communication channels of the societies. Data collection was performed from August to September 2020. Multiple participations by one person were excluded by an anonymous IP address check. The evaluation of data was carried out with SPSS 26.0 (IBM Corp.). Characteristics of participants A total of 422 trainees took part in the survey. Of these 209 (49.5%) were currently trained in Germany, 116 (27.5%) in Switzerland and 97 (23%) in Austria. 88.9% were female and the median age was 32 years. Around three-fourths of the study participants were in the third year of training or more. 42% worked at a large hospital and training facility with at least 500 beds. 77% of participants worked full-time. Part-time work and work sharing Although the largest number of participants worked full time, the majority (70%) rated part-time workloads between 80 and 95 % as the most attractive employment obligation. However, this did not seem to reflect a general desire to work less, as part-time work with less than 40% pensum was classified as “not attractive or not at all attractive” by 82% and 40-55% pensum were perceived as considerably less attractive by 45% of participants. Overall, part-time working models seemed to be widespread already, as 94% of the participants confirmed that their hospital offered some form of reduced pensum (see table 1 ). Workload Notably, we found distinct country-specific differences in the performance of medical procedures by non-medical healthcare professionals. Certain invasive interventions no longer need to be carried out by doctors only (see figure 1 ). For example, 98% of the trainees in Switzerland stated that they “never or rarely” insert an intravenous line, whereas 85% of German trainees “always to often” performed this intervention. Generally, the trainees described to spend the majority of their daily working hours on documentation. 27% perceived this as “not or not at all efficient” and only a quarter of the trainees indicated to have an assistant available (e.g. ward assistant), supporting them in documentation and organization tasks. Training of sub-specialties in OBGYN Regarding the training of specific sub-areas in OBGYN, 76% of participants rated obstetrics as “well to very well” represented in all participating countries (see figure 2 ). Major differences between the countries were observed in gynaecological oncology and senology as well as in prenatal care. Interestingly, only 5% of participants stated Endocrinology being “well or very well” covered within their curricula. Paediatric Gynaecology, as well as Sexual Health and Reproductive Medicine were seen as underrepresented sub-areas in OBGYN training by a vast majority of participants (see attachment 2 ). Assessment and implementation of current training regulations Interestingly, only 22% of participants felt “well to very well” prepared for their work as a specialist in the hospital setting and only 11% felt “well to very well” prepared for working in private practice. The majority of trainees assessed themselves only “moderately prepared” for the further work in a hospital (66%) or private practice (57%). In addition, this was associated with the fact that only 47% of trainees stated that they regularly fulfilled the required obligatory numbers of self-performed interventions (see figure 3 ). 53% of participants stated that they were actually faced with notable to serious difficulties to fulfil the required obligatory numbers of self-performed interventions being documented. As a result, two-thirds of the participants agreed that an electronically kept logbook is useful when documenting the obligatory interventions. However, the survey showed substantial, country-specific differences in the actual implementation of an electronically kept logbook. Whereas 86% of the participants from Switzerland answered that their logbook is kept electronically, only 5% from Germany and 1% from Austria currently kept their logbook electronically. Furthermore, an annual evaluation interview being documented in written form was offered to only 48% of participants. 54% of the trainees stated that they had a supervisor giving advice on questions with medical content or about career planning. Confidence during intervention To determine if the lack of self-performed interventions and the corresponding documentation have an impact on the feeling of security among the trainees, participants were asked how confident they feel in standard situations and interventions in OBGYN. Thereby, the intrinsic feeling of safety when they perform standard surgery is an important parameter for determining the quality of training. Interestingly, around two-thirds of trainees felt “confident to very confident” during standard interventions like curettage, Caesarean section and hysteroscopy. Among other interventions like simple laparoscopy and vacuum extraction as well as the management of emergencies in obstetrics like postpartum bleeding, or shoulder dystocia more than half of participants felt only "moderately" confident. When it comes to rare situations like breech birth or forceps delivery, most of the participants felt “not” or “not at all” confident (see attachment 3 ). Considering the years of specialty training, the feeling of security among frequently performed interventions such as the curettage, hysteroscopy and Caesarean section increased over time. However, divided into groups with and without a simulation training in obstetrics (44% with simulation training) or gynaecology (20%) offered in their hospital there was a noticeable increase of safety among the trainees who could use simulation training for their further education. Particularly, in these interventions (simple laparoscopy, management of postpartum bleeding or shoulder dystocia and vacuum extraction) that can arise in a hospital at any time, the trainees with simulation training feel up to 12% more confident than those participants without (see figure 4 ; box with broken line: noticeable differences between “with” and “without” simulation training). A total of 422 trainees took part in the survey. Of these 209 (49.5%) were currently trained in Germany, 116 (27.5%) in Switzerland and 97 (23%) in Austria. 88.9% were female and the median age was 32 years. Around three-fourths of the study participants were in the third year of training or more. 42% worked at a large hospital and training facility with at least 500 beds. 77% of participants worked full-time. Part-time work and work sharing Although the largest number of participants worked full time, the majority (70%) rated part-time workloads between 80 and 95 % as the most attractive employment obligation. However, this did not seem to reflect a general desire to work less, as part-time work with less than 40% pensum was classified as “not attractive or not at all attractive” by 82% and 40-55% pensum were perceived as considerably less attractive by 45% of participants. Overall, part-time working models seemed to be widespread already, as 94% of the participants confirmed that their hospital offered some form of reduced pensum (see table 1 ). Workload Notably, we found distinct country-specific differences in the performance of medical procedures by non-medical healthcare professionals. Certain invasive interventions no longer need to be carried out by doctors only (see figure 1 ). For example, 98% of the trainees in Switzerland stated that they “never or rarely” insert an intravenous line, whereas 85% of German trainees “always to often” performed this intervention. Generally, the trainees described to spend the majority of their daily working hours on documentation. 27% perceived this as “not or not at all efficient” and only a quarter of the trainees indicated to have an assistant available (e.g. ward assistant), supporting them in documentation and organization tasks. Training of sub-specialties in OBGYN Regarding the training of specific sub-areas in OBGYN, 76% of participants rated obstetrics as “well to very well” represented in all participating countries (see figure 2 ). Major differences between the countries were observed in gynaecological oncology and senology as well as in prenatal care. Interestingly, only 5% of participants stated Endocrinology being “well or very well” covered within their curricula. Paediatric Gynaecology, as well as Sexual Health and Reproductive Medicine were seen as underrepresented sub-areas in OBGYN training by a vast majority of participants (see attachment 2 ). Assessment and implementation of current training regulations Interestingly, only 22% of participants felt “well to very well” prepared for their work as a specialist in the hospital setting and only 11% felt “well to very well” prepared for working in private practice. The majority of trainees assessed themselves only “moderately prepared” for the further work in a hospital (66%) or private practice (57%). In addition, this was associated with the fact that only 47% of trainees stated that they regularly fulfilled the required obligatory numbers of self-performed interventions (see figure 3 ). 53% of participants stated that they were actually faced with notable to serious difficulties to fulfil the required obligatory numbers of self-performed interventions being documented. As a result, two-thirds of the participants agreed that an electronically kept logbook is useful when documenting the obligatory interventions. However, the survey showed substantial, country-specific differences in the actual implementation of an electronically kept logbook. Whereas 86% of the participants from Switzerland answered that their logbook is kept electronically, only 5% from Germany and 1% from Austria currently kept their logbook electronically. Furthermore, an annual evaluation interview being documented in written form was offered to only 48% of participants. 54% of the trainees stated that they had a supervisor giving advice on questions with medical content or about career planning. Confidence during intervention To determine if the lack of self-performed interventions and the corresponding documentation have an impact on the feeling of security among the trainees, participants were asked how confident they feel in standard situations and interventions in OBGYN. Thereby, the intrinsic feeling of safety when they perform standard surgery is an important parameter for determining the quality of training. Interestingly, around two-thirds of trainees felt “confident to very confident” during standard interventions like curettage, Caesarean section and hysteroscopy. Among other interventions like simple laparoscopy and vacuum extraction as well as the management of emergencies in obstetrics like postpartum bleeding, or shoulder dystocia more than half of participants felt only "moderately" confident. When it comes to rare situations like breech birth or forceps delivery, most of the participants felt “not” or “not at all” confident (see attachment 3 ). Considering the years of specialty training, the feeling of security among frequently performed interventions such as the curettage, hysteroscopy and Caesarean section increased over time. However, divided into groups with and without a simulation training in obstetrics (44% with simulation training) or gynaecology (20%) offered in their hospital there was a noticeable increase of safety among the trainees who could use simulation training for their further education. Particularly, in these interventions (simple laparoscopy, management of postpartum bleeding or shoulder dystocia and vacuum extraction) that can arise in a hospital at any time, the trainees with simulation training feel up to 12% more confident than those participants without (see figure 4 ; box with broken line: noticeable differences between “with” and “without” simulation training). Although the largest number of participants worked full time, the majority (70%) rated part-time workloads between 80 and 95 % as the most attractive employment obligation. However, this did not seem to reflect a general desire to work less, as part-time work with less than 40% pensum was classified as “not attractive or not at all attractive” by 82% and 40-55% pensum were perceived as considerably less attractive by 45% of participants. Overall, part-time working models seemed to be widespread already, as 94% of the participants confirmed that their hospital offered some form of reduced pensum (see table 1 ). Notably, we found distinct country-specific differences in the performance of medical procedures by non-medical healthcare professionals. Certain invasive interventions no longer need to be carried out by doctors only (see figure 1 ). For example, 98% of the trainees in Switzerland stated that they “never or rarely” insert an intravenous line, whereas 85% of German trainees “always to often” performed this intervention. Generally, the trainees described to spend the majority of their daily working hours on documentation. 27% perceived this as “not or not at all efficient” and only a quarter of the trainees indicated to have an assistant available (e.g. ward assistant), supporting them in documentation and organization tasks. Regarding the training of specific sub-areas in OBGYN, 76% of participants rated obstetrics as “well to very well” represented in all participating countries (see figure 2 ). Major differences between the countries were observed in gynaecological oncology and senology as well as in prenatal care. Interestingly, only 5% of participants stated Endocrinology being “well or very well” covered within their curricula. Paediatric Gynaecology, as well as Sexual Health and Reproductive Medicine were seen as underrepresented sub-areas in OBGYN training by a vast majority of participants (see attachment 2 ). Interestingly, only 22% of participants felt “well to very well” prepared for their work as a specialist in the hospital setting and only 11% felt “well to very well” prepared for working in private practice. The majority of trainees assessed themselves only “moderately prepared” for the further work in a hospital (66%) or private practice (57%). In addition, this was associated with the fact that only 47% of trainees stated that they regularly fulfilled the required obligatory numbers of self-performed interventions (see figure 3 ). 53% of participants stated that they were actually faced with notable to serious difficulties to fulfil the required obligatory numbers of self-performed interventions being documented. As a result, two-thirds of the participants agreed that an electronically kept logbook is useful when documenting the obligatory interventions. However, the survey showed substantial, country-specific differences in the actual implementation of an electronically kept logbook. Whereas 86% of the participants from Switzerland answered that their logbook is kept electronically, only 5% from Germany and 1% from Austria currently kept their logbook electronically. Furthermore, an annual evaluation interview being documented in written form was offered to only 48% of participants. 54% of the trainees stated that they had a supervisor giving advice on questions with medical content or about career planning. To determine if the lack of self-performed interventions and the corresponding documentation have an impact on the feeling of security among the trainees, participants were asked how confident they feel in standard situations and interventions in OBGYN. Thereby, the intrinsic feeling of safety when they perform standard surgery is an important parameter for determining the quality of training. Interestingly, around two-thirds of trainees felt “confident to very confident” during standard interventions like curettage, Caesarean section and hysteroscopy. Among other interventions like simple laparoscopy and vacuum extraction as well as the management of emergencies in obstetrics like postpartum bleeding, or shoulder dystocia more than half of participants felt only "moderately" confident. When it comes to rare situations like breech birth or forceps delivery, most of the participants felt “not” or “not at all” confident (see attachment 3 ). Considering the years of specialty training, the feeling of security among frequently performed interventions such as the curettage, hysteroscopy and Caesarean section increased over time. However, divided into groups with and without a simulation training in obstetrics (44% with simulation training) or gynaecology (20%) offered in their hospital there was a noticeable increase of safety among the trainees who could use simulation training for their further education. Particularly, in these interventions (simple laparoscopy, management of postpartum bleeding or shoulder dystocia and vacuum extraction) that can arise in a hospital at any time, the trainees with simulation training feel up to 12% more confident than those participants without (see figure 4 ; box with broken line: noticeable differences between “with” and “without” simulation training). This study aimed to identify the current situation of OBGYN training as well as transferable advantages of the different training systems in Germany, Austria and Switzerland. Anonymous surveys on “customer satisfaction” with their training and work situation, but also the assessment of the heads of the facilities are centrally recorded in some countries. Since 1996, the Swiss Institute for Medical Training (SIWF) has been carried out an annual survey among Swiss trainees which has served as a model for the current international survey. For decades, the results of the SIWF survey have served as feedback in order to recognize and promote successful concepts or to promptly uncover weak issues. Annually, the results of the survey are published online, and thus, offer young doctors an assessment basis for choosing an attractive employment and training position. At the same time, the data provide an annual benchmarking of the institutional training quality . Cross-border cooperation offers a great opportunity to learn and benefit from other training systems. PGME is teamwork that requires shared commitment to innovation, shared responsibility, supportive frameworks, and a teaching culture . Besides a few regional and interdisciplinary evaluation projects , , to our best knowledge this survey is the first cross-border project with special focus on OBGYN as specific subject area. This data serves as a valuable basis for further research and development in the field of supranational PGME in OBGYN. To overcome country-specific differences in training there has been great effort to harmonize training standards in OBGYN. The tendency of pan-European harmonization is the result of the increasing mobility of medical specialists and patients and the need for quality assurance of training throughout Europe , , . However, within the European Union, all countries have mutually recognized training qualifications for graduate and postgraduate medical education. This mutual recognition mostly is not content-related but based on minimum requirements, including training sites (recognized teaching hospitals) and duration of training , , . A push in the direction of a common, harmonized, European curriculum for advanced training in OBGYN is the Project for Achieving Consensus in Training (PACT) of the European Board and College of Obstetrics and Gynaecology . The curriculum defines content and competencies during a three-year basic training course (so-called “core”), which is the same for every gynaecologist. This is followed by a two-year advanced training phase with elective modules that can be chosen depending on the desired profile (so-called “electives”). Therefore, the EBCOG PACT offers sufficient opportunities to overcome country-specific burdens of PGME. Within our study group, Austria has already implemented the EBCOG PACT structure in OBGYN PGME. An important step towards improvement and maintenance of high quality PGME is to distribute the limited time resources as best as possible and to restructure non-medical tasks or to evaluate bureaucratic processes . According to a recent study by Trezzini et al., a trainee spends 167 minutes per day on documenting patient records. This corresponds to 27% of their working time. Instead of reduction, the medical bureaucracy has substantially increased in recent years . Also, within the present survey, the participants complained that bureaucracy takes up a large part of their everyday working time. Medical documentation and organization of standard procedures are seen to be inefficient and do drastically reduce the satisfaction of trainees . If additional tasks as venipuncture or insertion of an intravenous line are performed by trainees, this has major impact on patient care and quality of PGME. Simple tools such as digitized Dictaphones with voice recognition, but also major structural changes such as clinical nurses or physician assistants assuming tasks of clinical routine could considerably relieve the workload and increase time dedicated to PGME. In light of increasing workload and bureaucracy, it is not surprising that required obligatory numbers of interventions can barely be fulfilled during the standard length of training. However, instead of adjusting numbers or structure, the results suggest that missing interventions were subsequently attested and documented. Are the required numbers of different interventions too high for the existing number of cases? Can certain interventions and diagnostic measures only be carried out in specialized centres? Is the current routine clinical workload and bureaucratic effort incomparable to former generations of trainees? Comparing the logbooks of the three countries regarding the required number of interventions and diagnostic measures, serious differences are detectable. Whereas in Austria and Switzerland a total of 85 and 80 obstetric interventions performed by the trainee are required, respectively, German trainees only need 25 Caesarean sections and “contribution” in further obstetric interventions. 50 colposcopies are required in Germany and Switzerland, however, only 20 in Austria. A total of 275 gynaecological surgeries are required in Austria, 255 in Switzerland, and only 200 in Germany. We have to strike out new paths in order to ensure that not numbers, but practical verifications attest the level of training in OBGYN. So-called “Entrustable Professional Activities” (EPAs) can support the relationship between trainer and trainee being part of competence-based medical education. An EPA is a detailed description of a medical activity, e.g. a Caesarean section, which combines the knowledge, skills and attitudes required for this procedure. In countries such as the Netherlands and Canada, EPAs have already found their way into continuing medical education in various specialties ( Einleitung, . They support the change in PGME, away from an “on-off knowledge-based” examination at the end of training to a modern, practice-adapted and competence-oriented training concept. Trainees receive timely feedback on their activities and annual goals can be defined and evaluated. EPAs thus also form a valuable basis for annual evaluation meetings. An electronically kept logbook is also indispensable for recording such advanced training competencies and target-oriented evaluation discussions. In addition to written documentation, the digital form also enables a timely evaluation of the level of training and should be an essential part of a modern training program. Simulation training covers a wide range of training opportunities from high-tech team simulation training and skill drills to low fidelity training units. Each of these methods has its justification, as they train completely different abilities. While the team simulation training, which is increasingly established in obstetrics, is primarily about consolidation and training of treatment coordination, low fidelity models help to understand concepts based on simple technical repetitions. Although, simulation units are often accompanied by high costs, the present survey illustrates the positive impact of this additional training. In our survey, trainees who confirmed participation in any type of simulation training felt more confident especially in situations and interventions that are part of the basic training of OBGYN such as simple laparoscopy or postpartum bleeding. Simulation training offers a great benefit for modern PGME by increasing the efficiency in gaining experience and thus, improving the patients' safety. Like the SIWF survey, this study can not cover all aspects of the current training situation in the three countries. The study group chose to focus on certain topics that shape the daily worklife of trainees. The selection of topics and data was done to our best knowledge but is a limitation to the study. Certainly, further studies that cover more aspects of the basic training in OBGYN are needed to create a broader picture of the current situation of training in OBGYN in Germany, Austria and Switzerland. The current postgraduate training for OBGYN is already at a very high level in Germany, Austria and Switzerland. The aim is to jointly further develop this advanced training to be future-oriented. With the help of this survey, current weak points can be identified. Projects and ideas such as EBCOG PACT, EPAs, the reduction of bureaucracy through digitization and deepening skills through simulation training make a valuable contribution to compensate for these deficits and to adapt to future requirements. In this way, it is possible to secure the high level of European postgraduate training in OBGYN for future generations. We thank Larissa Luchsinger and Jeanine Ammann, research assistant at ETH Zurich, for the helpful evaluations, discussions and additions to this survey. We thank everyone who participated in the study. The study was funded with support of the DGGG, OEGGG and SGGG, however, the societies had no further involvement except for the financial support. The authors declare that they have no competing interests. Supplementary material – original ouestionnaire Sub-specialties represented in PGME in OBGYN in Germany, Austria and Switzerland Intrinsic feeling of safety among trainees during standard situations and interventions |
Molecular diagnostics tailoring personalized cancer therapy—an oncologist’s view | 7761f3d8-fdcd-46ce-a0d9-fb599918d7ed | 10948510 | Pathology[mh] | Precision oncology is currently revolutionizing the treatment of patients with cancer. Deciphering the underlying molecular alterations as drivers of cancer development and progression has led to an improved ability to understand and potentially interfere with pro-oncogenic pathophysiologic pathways . This development has been enabled by several key factors. First, analyzing genetic information has technically evolved in the recent years via methods of next-generation sequencing (NGS), enabling inexpensive and fast molecular diagnostics in routine clinical practice . Further, decades of experimental research have led to the precise characterization of distinct oncogenic pathways including pro-oncogenic driver mutations, tumor-suppressive mechanisms, and components of the cancer-host interaction including mechanisms of impaired anti-cancer immunity . Third, advances in drug development have led to the ability to directly interfere with these distinct molecular pathways (e.g., via tyrosine kinase inhibitors or monoclonal antibodies), which allows the development of personalized treatment approaches based on distinct patterns of genetic alterations . In the present review, we aim to define the concept of personalized cancer therapy specifically in the context of molecular diagnostics, to discuss the application of molecular profiling in tumor-agnostic therapeutic decision-making, and to point out current challenges and potential future directions of the precision oncology approach from an oncologist´s view. By incorporating comprehensive information on an individual patient’s level, the precision medicine approach aims to effectively guide disease prevention, diagnosis, and personalized treatment selection . Since cancer has been long recognized as a disease that is driven by an accumulation of genetic aberrations, the field of oncology has taken a pioneering role in the precision medicine paradigm . Historically, medical cancer therapy comprised a small number of cytotoxic chemotherapies that were selected depending on the respective tumor histology and disease location. Prompted by a steadily improved understanding of carcinogenesis and genetics, which was mainly enabled by the development of novel DNA analysis techniques such as polymerase chain reactions, finally in the late 1990s the first molecular targeted drug therapies were developed. The successful clinical implementation of the monoclonal HER2-antibody trastuzumab and the BCR-ABL tyrosine kinase inhibitor imatinib marked the earliest milestones in precision oncology and heralded a new treatment era of molecular stratified cancer therapy . In parallel, fundamental technological advances that have resulted in the development of NGS have revolutionized molecular profiling by dramatically decreasing analytic costs and turnaround time. In contrast to conventional sequencing techniques, NGS enables simultaneous analysis of multiple genes with high accuracy . Due to its high efficiency, the development of NGS thus paved the way for large-scale sequencing efforts including the cancer genome atlas projects that enabled a comprehensive genomic characterization of various tumors and thereby further transformed our understanding of oncogenesis and cancer evolution . Importantly, several recurrent genetic alterations were detected across different cancer types and subsequently characterized as potential therapeutic targets. Consequently, over recent years, an extensive and rapidly growing arsenal of drug therapies targeting numerous genetic alterations including gene mutations, rearrangements, and amplifications have been developed and effectively implemented in clinical practice . This was accompanied by a steadily rising application of NGS technologies in routine clinical practice for tailoring molecular stratified cancer treatment decisions in various tumor types such as non-small cell lung cancer , colorectal cancer , and biliary tract cancer . Finally, a significant step towards a personalized cancer treatment strategy has been taken with the recent approval of the first tumor-agnostic therapies, which are administered based merely on the discovery of a specific molecular mutation regardless of cancer histology and tissue of origin . Although the concept of cancer-agnostic personalized treatment guided by molecular profiling is highly promising, its successful clinical implementation is accompanied by key challenges that require careful consideration. One major practical hurdle is the complex and multistep workflow of matching targeted therapies to detected molecular alterations . This process starts by defining the questions whether the overall health status of the patient allows for implementation of molecular profiling, at what time point during the patient’s journey molecular profiling is initiated, whether a re-biopsy of the tumor lesion or, alternatively, a liquid biopsy is needed, and, last but not least, which diagnostic genetic analysis should be conducted. Further steps include NGS-analysis and bioinformatic data processing, variant calling, and the functional assessment of identified genetic alterations, parameters which are coordinated by pathologists and geneticists . Each individual step in this multi-layered process poses specific challenges and pitfalls that are discussed in more detail in other articles of this series. Yet, from an oncologist’s view, we consider the final step of the precision oncology workflow, namely the clinical annotation and clinical actionability assessment of detected molecular alterations, as the most critical and least defined step in the implementation of precision oncology. In this review, we will focus on the clinical remits of personalizing cancer therapies based on individual genetic compositions, pointing out key aspects including the purposeful selection of patients suitable for extended profiling, decision criteria for the choice of diagnostics, and finally the process of actionability assessment and biomarker guided therapeutic decision-making. Figure illustrates the workflow of a highly standardized and outcome-centered molecular tumor board (MTB) at a major Austrian academic center, which might serve as a potential template for integrating genomic cancer sequencing in clinical care among others. Currently, the overall efficacy of identifying actionable targets by molecular profiling in unselected patients with cancer remains low . Therefore, the European Society for Medical Oncology (ESMO) restricts its recommendations on the routine clinical use of multigene NGS testing to advanced non-small-cell lung cancer, prostate cancer, ovarian cancer, and cholangiocarcinoma. In advanced colorectal cancer, multigene testing can be considered an alternative to single gene polymerase chain reaction testing if additional costs are acceptable . Apart from that, multigene sequencing to tailor genome-guided individualized therapies is not routinely recommended and should only be performed within the framework of an academic program and restricted to patients in whom the testing results might have a direct impact on the clinical management . In contrast, major academic centeres in the USA and elsewhere opt for early and comprehensive tumor and germline genetic testing in largely all cancer patients , a view that remains not uncontested . The first critical step of genomic cancer sequencing remains the careful clinical evaluation whether molecular profiling is even indicated. In general, at the current level of understanding of tumor biology and available targeted therapies, it is unreasonable and ineffective to perform comprehensive genomic profiling in every early-stage cancer, as highly efficacious established treatments might be available in this setting . On the other hand, patients with advanced cancers considered for molecular profiling must be candidates to receive further antineoplastic treatment based on performance status, comorbidities, organ functions, and patient preference. Genomic profiling in patients with a reduced performance status or significant comorbidity burden are less likely to benefit from targeted therapies. Therefore, pursuing precision oncology approaches in such patients might even be detrimental as they might raise false hopes and even delay adequate palliative care interventions . In any case molecular profiling is considered, the patient’s autonomy must be carefully preserved. Therefore, prior molecular profiling is initiated; it is inevitable to inform the patient accurately and comprehensively about the realistic chance of finding a potential target, as well as potential implications of somatic mutational tumor testing such as the detection of molecular alterations highly suspicious for inherited cancer syndromes. The most robust clinical data for the implementation of extended molecular profiling to tailor targeted therapy exists for patients who have exhausted all established and clinically efficacious treatment options and retain an adequate performance status. In this setting, several precision oncology trials evaluating different concepts of NGS testing to guide targeted cancer treatment have reported promising results . Furthermore, patients with rare cancers with limited evidence-based treatment options and patients with exceptional treatment response patterns are potentially favorable candidates for extended genetic profiling . However, it remains an open question whether early initiation of precision oncology might provide more benefit to cancer patients . This is based on the hypothesis that targeted drug exert better effects in cancer cells prior to their exposure to several lines of chemotherapy, radiotherapy, or alternative treatment modalities. In current clinical practice, targeted cancer gene hotspot panels covering 20–500 genes are mostly used for genomic profiling. In this regard, various NGS platforms are available each offering a slightly different spectrum of DNA and RNA coverage . The choice of panel sequencing must be made individually depending on several factors including the type and stage of disease, the treatment history, the availability of previous sequencing results, the accessibility of targeted treatments, and of course the financial resources . At present, the use of comprehensive genomic profiling including whole genome, whole exome, and transcriptome sequencing is mainly restricted to scientific purposes. The benefit for patient outcomes when comparing large comprehensive genomic sequencing efforts to targeted cancer gene panels is likely, but so far, the data suggesting a robustly improved identification of clinically relevant somatic alterations remains inconclusive. Several trials evaluating the clinical utility of comprehensive genomic profiling are currently ongoing . Another remaining controversial aspect of molecular profiling in clinical practice is whether archival tissue should be used for profiling. Considering that most patients who are candidates for extended genomic profiling have advanced disease in which clonal evolution is known to take place under therapeutic selective pressure, most precision oncology trials involve obligatory fresh biopsies to screen for molecular targets . However, in clinical routine, obtaining invasive procedures of tissue re-biopsies poses significant challenges based on patient preference, procedural risk, and issues of time delay. Since genomic profiling of matched tissue and circulating tumor DNA (ctDNA) samples have shown high concordance rates of detected molecular aberrations, minimally invasive liquid biopsy of circulating tumor DNA offers an increasingly attractive alternative in this setting . Beyond that, as ctDNA is thought to be released into the bloodstream from different tumor lesions at the same time, liquid biopsy might even provide a more comprehensive capture of the molecular composition of a patient’s tumor than a single tissue biopsy . However, the reliability and accuracy of ctDNA sequencing has to be further improved and clinically validated before a widespread implementation for tailoring genome guided treatment decisions can be performed in routine clinical practice . The work process of actionability assessment heavily relies on the accuracy and validity of the molecular report provided by the pathologist. Hence, a state of the art functional annotation and appropriate reporting of detected alterations represents a key prerequisite for all further steps of actionability assessment .. This underlines the crucial role of pathologists and the necessity of close interdisciplinary interaction between clinicians and pathologists for the successful implementation of precision oncology in clinical care. Since the functional role of variants of unknown significance is unclear, only variants classified as pathogenic or likely pathogenic are usually considered for the actionability assessment. The identification of predictive biomarkers for antineoplastic therapy represents the cornerstone of the clinical annotation process. So far, a broad and steadily increasing spectrum of molecular predictive biomarkers could be identified and clinically validated in specific cancer types. These include gene amplifications and protein overexpression (e.g., HER2) , gene mutations (e.g., BRAFV600E) ., gene fusions (e.g., EML4-ALK rearrangement), and compound biomarkers such as the microsatellite status and tumor mutational burden .. Many of these genetic drivers can be found across multiple cancer types with varying frequencies, which prompted the idea of genome guided treatment selection irrespective of the cancer histology and origin. Although this concept is highly promising and has been recently reinforced by the paramount example of highly efficacious tumor agnostic NTRK fusion targeting , we have been taught that the efficacy of targeted therapies in one cancer type cannot be automatically extrapolated to others. This is perfectly illustrated by the BRAF V600E mutation that can be effectively targeted by single agent or combined BRAF plus MEK inhibition in metastatic melanoma . and NSCLC , but not in colorectal cancer due to a feedback upregulation of the epidermal growth factor receptor (EGFR) that necessitates further EGFR blockade . Hence, the key challenge of the actionability assessment is to interpret the detected molecular alteration in the context of the present cancer histology and the co-mutational tumor profile. For this purpose, a comprehensive literature research of preclinical and clinical evidence is critical. Fortunately, various partly publicly available precision oncology knowledge databases such as OncoKB, the Jackson Laboratory Clinical Knowledge Base, and My Cancer Genome that provide regularly updated curated data on cancer associated molecular alterations including reference to their clinical actionability have been established . Furthermore, commercially available decision support platforms that utilize different algorithms of molecular profiling guided therapy matching are offered. Importantly, the concordance of the actionability assessment by these platforms has been shown to be rather low , which underlines that algorithm generated treatment suggestions should only be used for decision support and must always be critically scrutinized by a clinician in the context of the individual patient case. Accordingly, it will be necessary that specialized oncological centers with sufficient infrastructural support provide the clinical annotation of molecular profiles for individual patients because treatment decisions based solely on decision support software fail to yield a comprehensive overview of therapeutic options . To harmonize the clinical interpretation and actionability assessment of molecular alterations for personalized cancer treatment, the European Society for Medical Oncology Translational Research and Precision Medicine Working Group has proposed a framework that shall enable a more precise classification and prioritization of molecular targets. With due regard to the available clinical evidence supporting a biomarker drug interaction and its consequent clinical implications, the ESMO Scale for Clinical Actionability of molecular Targets (ESCAT) defines six levels of evidence for molecular targets . Tier I comprises alteration drug matches that have been shown to result in improved clinical outcome in prospective clinical trials. According to the underlying trial design, further subclassifications of the evidence level in Tier Ia (randomized), Tier Ib (non-randomized), and Tier Ic (basket trial) are feasible. Tier I targets should be considered standard of care. Tier II defines alteration drug matches that are associated with clinical activity; however, the magnitude of benefit is not clear yet. These are considered investigational targets, which should be primarily matched within the framework of a clinical trial or registry study. Tier III describes hypothetical alteration drug matches, which are suspected to result in a potential clinical benefit based on prospective trial data on the same target in another cancer type (Tier IIIa) or on the detection of an alteration functionally closely related to a known Tier I alteration (Tier IIIb). Tier III targets should be ideally investigated as part of innovative precision oncology trial concepts such as N-of-1 trials. Tier IV targets are supported exclusively by preclinical evidence and should thus not be considered targetable in clinical practice. Tier V alteration drug matches have been shown to be associated with antitumor activity that however did not translate into improved survival. In this regard, combinational therapy approaches can be considered within the framework of a clinical trial if functionally plausible. Tier X alterations have no preclinical or clinical evidence of actionability. Owing to the rapidly increasing number of already established and newly identified molecular biomarkers and corresponding approved targeted therapies, a pertinent clinical interpretation of genomic sequencing results has lately become increasingly complex and time consuming. Genomic knowledge databases and decision support platforms outlined previously can assist in the clinical actionability assessment of detected alterations; most clinicians are neither aware of these tools nor do they have the genetic knowledge and timely resources for an accurate interpretation of the literature. Therefore, an expert evaluation of sequencing results is crucial to optimize the efficient clinical application of NGS testing for therapeutic target identification. For this purpose, molecular tumor boards are increasingly established in cancer centers, which provide a multidisciplinary platform to enable the successful integration of the precision oncology approach in patient care. So far, universal recommendations on the composition and workflows of molecular tumor boards (MTBs) are missing. However, most MTBs are constituted by experts from different medical fields including clinicians, pathologists, geneticists, bioinformaticians, and molecular biologists. The main tasks of the MTB encompass the initiation of appropriate genetic testing, assessment of molecular profiling results for target identification and personalized treatment recommendation, assistance of diagnosis in cases with indeterminate histology aberrations, and the detection of inherited cancer susceptibilities . To ensure optimal decision-making, a comprehensive review and evaluation of the medical history, the duration and response of previous antineoplastic therapies, and the availability of archival tumor samples, as well as previous molecular testing results of each individual patient is inevitable. In addition to the abovementioned areas of responsibility, the MTB has a key educational role to deepen the understanding of molecular oncology and spread the knowledge how to adequately utilize cancer genome diagnostics for tailoring patient care. Further, MTBs shall serve as a venue to foster innovative translational research projects with the ultimate goal of identifying novel predictive biomarkers and resistance mechanisms and thereby fully incorporating the research concept of bedside-to-bench and back . Since the appropriate implementation of MTBs requires a high level of expertise from different medical specialties that are usually only provided by selected academic institutions, so far only a small minority of cancer patients can benefit from MTB facilities . This poses one key challenge, which might be overcome by the implementation of centrally coordinated precision oncology initiatives that provides a virtually accessible platform for patient case discussion, knowledge exchange, and translation research design across multiple cancer institutions . The identification of tumor-agnostic unifying molecular compositions that enable personalized therapies is the main goal in precision oncology. In recent years, multiple genetic alterations have been identified that serve as therapeutic targets irrespective of underlying cancer types (Table ). In the following section, we provide a brief overview of two established examples of tumor-agnostic genomic targets as prime examples for the high therapeutic potential of personalized oncology approaches. Microsatellite instability and genetic hypermutability have lately received increasing clinical attention as tumor-agnostic predictive biomarkers for immune checkpoint inhibitor response. Microsatellite instability is caused by DNA mismatch repair deficiency that evokes an accumulation of genetic alterations in short non-coding repetitive DNA segments distributed throughout the genome and referred to as microsatellites. Since the DNA mismatch repair system plays a key role in maintaining genomic stability, its deficiency is further associated with an increased number of somatic tumor mutations . The phenomenon of genetic hypermutability specified by the tumor mutational burden (TMB) strongly correlates with the abundance of tumor neoantigen formation which has been proposed to be critical for immune checkpoint inhibitor (ICI)–mediated T cell response . These findings prompted the clinical investigation of ICI therapy in patients with DNA mismatch repair deficiency and/or high TMB. DNA mismatch repair deficiency that can be either assessed on the protein expression level by immunohistochemistry or indirectly by the genomic detection of microsatellite instability (MSI) can be found across various cancer types with an overall prevalence of approximately 4%. The highest disease-specific prevalence of MSI is observed in Lynch syndrome–associated cancers including endometrial, colorectal, and gastric adenocarcinoma . Importantly, in a seminal study by Le et al., PD-1 blockade with the ICI inhibitor pembrolizumab resulted in a remarkable response rate of 52% and a high proportion of durable remissions in heavily pre-treated patients with different types of MSI high advanced carcinomas . These findings prompted the first tumor-agnostic therapy approval by the Food and Drug Administration (FDA), which was recently further justified by several cancer type specific trials confirming the exceptional efficacy of ICI therapy in patients with MSI high tumors . The tumor-agnostic predictive validity of the TMB is less clear. One basket phase II trial enrolling patients with selected advanced solid tumors, demonstrated a significantly higher response rate of the PD-1 antibody pembrolizumab in patients with TMB high tumors, defined as ≥ 10 tumor specific mutations/megabase detected by the targeted FoundationOne CDx assay . While data on major tumor types such as breast, colorectal, and prostate cancer not included in this trial were missing, the FDA approved pembrolizumab for treatment of TMB high tumors irrespective of cancer histology. This approval was challenged by a comprehensive retrospective cohort study of more than 1500 patients treated with ICI therapy by McGrail et al. indicating that the TMB only discriminates ICI efficacy in the subset of tumor types in which CD8 cells correlate with the neoantigen load, whereas in other cancer types such as breast and prostate cancer the TMB was not associated with ICI response. Importantly, tumor specific subgroups in this study were small, which alleviates its general validity . Thereby, further research is warranted to clarify the tumor-agnostic predictive role of the TMB for ICI efficacy. Neurotrophic tropomyosin-receptor kinase (NTRK) genes are physiologically involved in neural development and encode a family of receptor tyrosine kinases . Fusions of NTRK genes have been identified as oncogenic drivers in different solid and hematologic malignancies in adults and children, with heterogeneous gene fusion partners that constitutively activate tyrosine kinase signalling . In general, NTRK fusions represent a very rare type of genetic alterations in general oncologic populations. For example, two large genetic screening studies suggest an overall prevalence of NTRK-fusions of 0.3% among cancer patients . However, in several rare cancers including secretory breast carcinoma, infantile fibrosarcoma, secretory salivary gland cancer, or pediatric thyroid carcinomas NTRK fusions represent a common genetic alteration . Further, albeit rare, NTRK fusions are detected across various more frequent cancers, broadening their clinical relevancy . Importantly, the development of tyrosine kinase inhibitors (TKIs) targeting NTRK represents a promising therapeutic approach for patients. In detail, larotrectinib and entrectinib have demonstrated profound clinical activity in the treatment of patients with cancers and underlying NTRK fusions. In a pooled analysis of basket trials enrolling patients with NTRK-fusion positive advanced cancers, larotrectinib treatment led to an ORR of 78% and median PFS was 37 months . Further, ORR with entrectinib was 61%, with a median PFS of 14 months . Synoptically, the availability of highly active NTRK inhibitors highlights the potential of broadly assessing NTRK fusion positivity in patients with advanced cancers beyond established therapeutic options, especially in those with specific rare cancer types . Despite remarkable advances in precision oncology, important limitations and challenges of genome guided therapy remain to be solved to enable a broader and more efficient clinical implementation and thereby maximize patient benefit . First, over the process of clonal evolution in tumorigenesis and cancer progression, cancers acquire a variety of pro-oncogenic molecular aberrations. Thereby, over the course of disease, cancers evolve towards a higher level of heterogeneity and subclonality . Consequently, therapeutic efficacy in very advanced cancer settings is limited a priori due to the high probability of underlying genetic properties of cancers to circumvent single-target personalized therapies. Conceptually, targeting of specific cancer driver genes in earlier treatment settings might therefore enable a more pronounced anti-cancer effect. In the future, earlier integration of personalized treatment approaches in clinical care might therefore yield more promising therapeutic efficacy. Secondly, precision oncology is currently limited in our current understanding of attributing the degree of pathogenicity of identified genetic alterations. Specifically, cancers frequently attain a number of passenger co-mutations which are not vital for cancer progression. Furthermore, healthy tissues have been demonstrated to acquire somatic mutations with varying degrees of pathologic significance. For example, somatic mutations in hematopoiesis obtained with increasing frequency with higher age are frequently detected in diagnostic evaluation of circulating tumor DNA and thereby decrease the specificity of observed mutational patterns. Moreover, mutations in various classic prooncogenic driver genes have been detected in different non-malignant diseases . These limitations might be overcome in the future with the implementation of personalized modeling of the functional impact of identified genetic alterations and respective therapeutic targeting on RNA, protein, or cellular levels . In addition, with the advent of artificial intelligence–based technologies, pathologic and clinical annotation of molecular diagnostics might be further facilitated. Thirdly, from a practical point of view, precision oncology is currently limited in several aspects of structural and technical restraints. The current timeframe from initiation of molecular diagnostics until the actual implementation of personalized therapies might take up to several weeks. Thereby, in a mostly advanced oncologic therapeutic setting, a considerable proportion of patients are lost during the process. In addition, availability of recent tissue samples is frequently necessary to enable reliable genetic information on the current molecular makeup of a cancer, due to the process of clonal evolution and genetic mechanisms of treatment resistance that might accumulate during previous anti-cancer therapies. Therefore, personalized oncology frequently depends on the performance of novel tissue sampling and biopsy, which might affect the risk–benefit ratio of this treatment approach. However, in the future, advances in the field of liquid biopsies via analyzing circulating tumor DNA hold the potential to potentially replace the need for additional tissue-based testing . Finally, one major hurdle for a widespread global adoption of the precision oncology approach that must be considered is the financial burden coming along with comprehensive genomic sequencing and even more the cost of targeted therapy itself. Unfortunately, access to molecular profiling and personalized cancer therapy is currently restricted to a small minority of patients with cancer in high-income countries. However, on the long term, more precise and efficacious treatment selection of targeted cancer therapies by improved prediction of treatment benefit might even reduce costs compared to unguided conventional treatment by enabling an ambulatory management and avoid hospitalizations associated with disease complications . Studies that specifically include cost-effectiveness evaluations of the precision oncology approach are therefore urgently needed. Treatment of patients with cancer is currently undergoing a dramatic shift towards personalized therapy using molecular diagnostics. However, important limitations remain to be solved in order to maximize patient benefit, including levels of cancer-specific genetic heterogeneity, interpretation and clinical annotation of identified genetic alterations, and current technical constraints in molecular diagnostics. In the future, with an improved understanding of the complex underlying molecular mechanisms via integrating various layers of genetic and functional analyses in a refined process of personalized clinical decision-making, alongside with an enhanced ability to dynamically detect and monitor individual cancer-driving molecular aberrations via liquid biopsies, personalized oncology will dramatically change our current concept of cancer therapy. |
Addressing the Human Experience of Chronic Kidney Disease: A Call to Transform Kidney Care | 030cd8b7-819c-4f0e-a7d3-1ab5b5aedcbd | 11787170 | Patient Education as Topic[mh] | This article is published with digital features, including a graphical abstract and patient perspective video, to facilitate understanding of the article. To view digital features for this article, go to 10.6084/m9.figshare.27242076. Chronic kidney disease (CKD) is a growing healthcare concern that affects approximately 850 million people worldwide and is projected to be the fifth most common cause of death by 2040 . It is a progressive condition characterized by abnormalities of kidney structure or function that are present for > 3 months, with implications for health . Early-stage CKD is often asymptomatic; however, kidney function typically declines over time and patients with later stages of CKD may experience a range of symptoms, such as poor sleep, pain, poor mobility, fatigue, and taste disturbances . CKD is also associated with an increased risk of cardiovascular morbidity and mortality . Early intervention is essential to limit disease progression to kidney failure and avoid the need for kidney replacement therapies, such as dialysis and transplant . Living with CKD negatively impacts a patient’s quality of life and mental health, with one in four patients experiencing depression . The disease burden, especially if dialysis is required, can lead to unemployment, and treatment costs are a major concern for patients and their families, even in countries with national health services . CKD can also negatively affect the quality of life of caregivers . To provide appropriate person-centered care, patients and their caregivers need education about the disease and its management, healthy living advice, and mental health support. The Global Kidney Health Atlas 2023 report describes the barriers to optimal kidney care that affect patients (knowledge, attitude, capacity to pay), healthcare professionals (HCPs) (availability, access, knowledge, specialties), and healthcare systems (availability, access, capability) . Patients’ primary concerns included the ability to work, their mobility, and the financial impact of kidney disease and treatment . This commentary builds on the report to highlight the real-life experiences of people living with CKD. The diverse author perspectives in this commentary include patients and patient advocacy representatives, as well as primary care and specialist HCPs who support the transformation of CKD care. When discussing the impact of living with CKD, authors’ concerns comprised relationships and support; work and finances; and awareness, intervention, and prevention of disease progression to preserve quality of life. This article is based on author experience and previously published research, and does not describe any new studies performed by any of the authors. To support your reading of this commentary, please see the associated video available from the online/HTML version of the manuscript or follow the digital features link under the abstract. Support networks help patients to meet their basic needs (e.g., food preparation, medication management, transport to medical appointments). These networks also provide emotional support, which is vital given the substantial mental burden associated with CKD. The level of care that patients require evolves as the disease progresses, with more assistance necessary for advanced CKD. A patient’s support network is multilayered. Their inner circles comprise partners, family members, and/or close friends who play an essential role as caregivers, followed by extended family, work colleagues, and fellow patients or social communities. Formal networks such as patient organizations and support groups form the outer layer. For example, in the USA, the National Kidney Foundation helps to connect patients to their peers and helps to act as a bridge between the perspectives of patients and HCPs . However, these formal support networks may not be accessible to all patients or may be absent in some countries. A patient’s relationship with their healthcare team also forms a vital part of their outer support network; this is facilitated by continuity of care and patient confidence in their HCPs. Primary care HCPs, who are the likely first point of contact for patients with early-stage disease, need to be appropriately trained to identify and care for patients with early-stage CKD. Support networks are essential for patients, but the burden of CKD can make it challenging to maintain personal connections (e.g., missing social events or family occasions because of poor health or diet restrictions), and some relationships may suffer over time. Partners, family, or close friends often become caregivers, a role which may intensify as the patient’s care requirements or energy levels change. Caregiving for patients with moderate or advanced CKD can be particularly exhausting and time-consuming. The resulting relationship strain can negatively affect a patient’s quality of life . Isolated patients may not have sufficient support, necessitating a move to assisted living facilities; this loss of independence could further affect their well-being. Regarding intermediate support networks, people with CKD may not receive the same help as people with other serious conditions because of these patients’ ability to present as if they are well and the lack of public awareness of CKD. Strong patient support networks are associated with meaningful outcomes, such as reduced frailty or mortality, as well as positive psychosocial outcomes for patients . Kidney Research UK recommends assessing patients' psychosocial needs and providing kidney-specific psychosocial support irrespective of their clinical severity . However, patients often report shortages of counselors and psychologists . Caring for caregivers is another important aspect of care for chronic conditions, and interventions should consider caregivers and patients both as a group and as individuals . Funding research on how CKD intersects with mental health and investing in psychosocial support will help patients and caregivers. Ultimately, preventing CKD progression will preserve patient quality of life and alleviate the need for caregiving support. Moderate to advanced CKD may also affect a patient’s ability to study, work, or progress in their career . Time management is key to balancing education, training, or work commitments with treatment, and patients may have to change their contracted hours or role to accommodate living with CKD. Loss of income or potential earnings can affect patient mental health and can have repercussions on their family life. Access to clinical care is a concern for patients, and their employment status may alter their health insurance status . This may be particularly concerning in countries without public health funding for treatment or with substantial public waiting lists for medical care. The ramifications of CKD also extend to a patient’s close support network: for example, a primary caregiver may not be able to work while providing practical support for their loved one. Additionally, family finances can be compromised for extended periods of time to support kidney replacement therapy costs when no public health funding is available. Employers can create a supportive working environment by identifying and addressing barriers to sustainable employment (e.g., flexible working hours, working from home) . Limiting disease progression circumvents many of the challenges associated with presenteeism and absenteeism, thereby benefiting patients, caregivers, and employers by enabling people to contribute to the workforce for longer. Quality of life decreases as CKD progresses , so the timely diagnosis and treatment of CKD is paramount to preserving patient well-being. Although CKD has no cure, and reversing kidney damage is generally not possible, current treatments can slow kidney function decline and can help to avoid progression to kidney failure . However, early-stage CKD is a silent disease because it is usually asymptomatic, and patients are unaware of their condition unless appropriate testing is carried out. Even after testing, a formal diagnosis is often not made . Detection of early-stage CKD therefore relies on screening at-risk populations (i.e., people with diabetes, hypertension, cardiovascular disease, obesity, or a family history of kidney disease). International clinical guidelines recommend two tests for CKD: (1) estimated glomerular filtration rate (eGFR), a blood test indicating kidney function; and (2) urine albumin-to-creatinine ratio (UACR), a test of kidney damage . Both tests act as independent predictors of risk to inform CKD stage, prognosis, and treatment; however, UACR testing is not standardized or widely performed , possibly due to a misperception that it does not inform treatment decisions . In some circumstances, HCPs may not be appropriately trained or have the time or resources to recognize CKD risk factors and implement subsequent detection measures. Although HCPs understand that type 2 diabetes is an important risk factor, the risk factors for CKD are not always communicated effectively to patients. More patient and HCP education is needed about the risks that may make people more vulnerable to CKD. A new treatment vision encourages discussion between patients and HCPs about maintaining kidney health, not just avoiding kidney failure. Patients with comorbidities associated with CKD would benefit from understanding their risk of developing CKD and the advantages of specific preventive measures and self-care strategies to empower them to proactively seek kidney health evaluations. Patients with newly diagnosed CKD are often poorly informed about their disease, progression, or treatment options, and may not feel empowered or know how to manage their lifestyle risks. Indeed, a study evaluating the effectiveness of home screening found that many people do not seek treatment or medical advice after home tests of kidney function reveal possible evidence of CKD . Participants who did not seek help were more likely to have a lower socioeconomic status, higher body mass index, diabetes, and poorer kidney function than those who did visit their primary care HCP . Understanding the behaviors that support health and encourage patients to be proactive in their care should inform educational materials on CKD. For example, the albuminuria, blood pressure, cholesterol, diabetes, and eGFR (ABCDE) approach provides patients with a framework for their conversations with HCPs . Improving patient understanding of the eGFR and UACR measurements used to diagnose CKD could help to normalize discussions of kidney health and allow patients to participate in their care . Shared decision-making, which integrates the best available evidence, clinical experience, and patient preferences, can increase patient engagement and improve outcomes . Patient and community education can help to overcome low awareness of CKD, enhance understanding of the disease, and improve outcomes . Kidney health-related education could facilitate informed decision-making for first-contact HCPs and provide techniques and tools to help patients navigate their healthcare. Appropriate intervention and treatment should follow diagnosis. However, there is a substantial shortage of HCPs in low- and middle-income regions, and many countries across all income brackets lack sufficient CKD HCPs, with patients identifying HCP shortages as a major barrier to care . Although generics for some medicines are available, patients may struggle to access and pay for treatments if their healthcare is not fully funded. In these situations, health authorities need to communicate the benefits of prevention for both individual and public health, as well as identify what can be done with available resources (e.g., preserving kidney health through lifestyle changes, such as improving diet, managing weight, and increasing physical activity, and providing access to affordable medication) . Healthcare systems and decision-makers need to be convinced that transforming CKD care through early diagnosis and intervention is a priority. The need for strategies that prevent the occurrence, as well as the progression, of CKD was highlighted by a 2024 report from Japan, where excess healthcare spending was associated with early-stage CKD, increasing further for advanced CKD . Although there are clear clinical and financial benefits to timely intervention, we have focused on the potential improvements in well-being for patients and their caregivers in this commentary. Patients consider the lack of government policies dealing with CKD to be a primary obstacle to care, with CKD-specific policies in place in only 37% of countries . There is no current consensus about whether healthcare systems and governments should employ CKD screening programs, or about which specific groups should be targeted. This is despite affordable, accessible, and effective methods to diagnose and treat CKD, as well as international clinical practice guidelines and endorsement by clinical associations . Overcoming this inertia will require the engagement of multiple stakeholders, and the communication of a unified message by patients and caregivers, as well as primary care and specialist HCPs. Conclusively, we need to convey the value of early diagnosis and guideline-directed medical therapies to policymakers, budget holders, and primary care HCPs. By preventing disease progression, patients can live more years in better health, avoid invasive treatments, and maintain their well-being. Living with CKD can test a patient’s resilience and coping mechanisms. Person-centered care considers how to support a patient’s quality of life, as well as providing clinical interventions. Early identification and holistic intervention could slow disease progression and could protect quality of life, financial well-being, and work activity for patients with CKD. Engaging with patients, providers, and policymakers to raise awareness of the importance of kidney health and early diagnosis could improve patient outcomes and satisfaction with care, as well as facilitate more effective use of healthcare resources. Patients, caregivers, and HCPs need to communicate this message with a unified voice to convince decision-makers to prioritize the early diagnosis and treatment of CKD, thereby minimizing the long-term implications for healthcare systems. |
Improving clinical reasoning and communication during handover: An intervention study of the BRIEF-C tool | 229cd018-7023-4071-8ed9-4d2a9a452791 | 11086570 | Internal Medicine[mh] | Miscommunication is known to occur during handovers but improves when a tool is used to structure the information shared. Most structured communication tools during handover are atheoretical, omit communication strategies, have limited evidence of validity and reliability, and have been tested only in simulated environments. The BRIEF-C (Background, Responsible diagnosis, Included differential diagnosis, Excluded differential diagnosis, Follow-up, Communication) tool was developed based on cognitive bias theory, addresses communication strategies, has strong emerging psychometric evidence, and improves communication during handover in a clinical environment, potentially contributing to improved patient care and safety. Our study can guide further research into optimising handover processes, promoting patient safety, and improving healthcare team collaboration. In practice, healthcare professionals can use the BRIEF-C to enhance communication during handovers, potentially reducing errors and improving patient outcomes. Policy-wise, this study emphasises the importance of standardised handover tools and practices, which could lead to implementing guidelines and protocols that prioritise effective communication in healthcare settings. Transition periods along the continuum of clinical care are recognised for their inherent vulnerability and high risk to patient outcomes. Clinical handover (also known as handover or handoffs) involves transferring ‘professional responsibility and accountability for some or all aspects of care for a patient, or group of patients, to another person or professional group on a temporary or permanent basis’. One estimate of handovers in Australian hospitals was reported at over 7 million annually. Despite their frequency and necessity, the quality of clinical handovers remains a concern. For instance, in a survey of junior doctors in the UK, 32% reported that the handover process was poor, 50% adequate, 17% good, and only 1% stated that it was excellent, with only 6% receiving written handovers. More recently, a survey of resident trainees in the USA indicated that 15% of adverse events, errors, or near misses could be attributed to poor handoffs (compared with 19% for long work hours, 20% for limited supervision, 5% for caring for other patients, 12% for caring for their own patients, and 57% for other reasons). Poor handovers have been linked to several detrimental outcomes, including care discontinuity, adverse events, and legal malpractice claims. Intriguingly, malpractice insurance data in the USA identified clinical handover as a prominent cause of claims, especially among trainees (accounting for 20% of cases). Given these concerns, this study seeks to enhance care transition during handovers by introducing and training the utilisation of a theoretically grounded tool named the Background, Responsible diagnosis, Included differential diagnosis, Excluded differential diagnosis, Follow-up, and Communication (BRIEF-C). Handover tools Several tools, such as the ISBAR (Introduction, Situation, Background, Assessment, Recommendation) and I-PASS (Illness severity, Patient information, Action list, Situational awareness and contingency plans, and Synthesis) exist to aid in organising communication during handovers. However, evidence of reliability and validity is limited, and many studies were conducted in simulated, not clinical, settings. Moreover, the development of these tools was not informed by theory. Our project is unique in that the handover tool we developed includes a structured framework intended to mitigate cognitive biases that can contribute to communication errors, an often overlooked aspect in the domain of handover communication. Cognitive bias theory Healthcare professionals often use heuristics or ‘shortcuts’ to formulate initial diagnoses. Although this approach is practical and adequate in most situations, there can be instances of failed heuristics. The failed heuristics are often seen and labelled as cognitive biases, which are consistent deviations in judgement due to information processing limits, decision-making shortcuts and emotional, moral or social influences. The process of diagnostic clinical reasoning and reducing diagnostic cognitive errors has been identified as one of the blind spots in patient safety and healthcare. In fact, cognitive errors have been demonstrated to contribute to up to 74% of system-related factors and up to 65% of diagnostic errors in Internal Medicine. Furthermore, Emergency Medicine is considered to be an ill-structured and chaotic environment at high risk of producing cognitive errors. In such conditions, the combination of poor communication strategies combined with a failed heuristic that might be carried on during the transition of care will create the perfect storm for adverse patient outcomes. The development of the BRIEF-C was based on several premises. First, clinical handover is a high-risk, high-frequency procedure that occurs in healthcare where the majority of communication failures could be averted by using a handoff tool. Second, improving the diagnostic process has emerged as a crucial benchmark for patient safety. Third, the heuristics and biases research suggests that quality decision-making might be influenced by judgement blind spots. Fourth, the approach to handovers was reframed as an opportunity for enhancing diagnostic care through system and process improvements. Such interventions should address the two dimensions of the linguistic dialogue during handovers: the informational structure embedded within an active collaborative format. Fifth, translation of communication strategies from high-reliability organisations. Here, tailoring some of the learnt lessons (e.g., face-to-face, interactive questioning, readback, limiting interruptions, delaying the transfer of responsibility during critical actions, overhearing others’ updates) will support grounding in communication to establish that what has been said is understood and interpredictability of actions and behaviours during the joint activity of handover exchange. Enhancing effective transitions across the patient care continuum is expected to result in more effective and timely care, foster improved communication among diverse healthcare professionals and patients, and improve planning for discharge and follow-up care in the community. There is a need for improved tools and standardised processes to narrow, and ultimately eliminate, the gap between transmitting and receiving patient information, in addition to considering the diagnostic safety at times of handover and transition to care. For example, surgical residents in Canada have reported their involvement in preventable patient injuries, underlining the importance of refining communication while integrating the diagnostic reasoning process for reducing patient harm and addressing our psychological tendencies to accept diagnostic suggestions. In summary, this study examines psychometric evidence of the BRIEF-C tool that was developed based on cognitive bias theory to see if its use improves communication during handovers. The study includes residents and physicians from two diverse clinical disciplines: Internal Medicine and Orthopaedic Surgery. While diverse in context, healthcare providers in both disciplines have to communicate and make judgements under parallel conditions of clinical uncertainty. By focusing on Internal Medicine and Orthopaedic Surgery specialties, the study captures scenarios with potential high-stakes outcomes. This choice enhances the practical applicability of the study’s findings, as effective communication in these settings is crucial to avoid errors, ensure proper diagnosis and treatment, and ultimately enhance patient well-being. Several tools, such as the ISBAR (Introduction, Situation, Background, Assessment, Recommendation) and I-PASS (Illness severity, Patient information, Action list, Situational awareness and contingency plans, and Synthesis) exist to aid in organising communication during handovers. However, evidence of reliability and validity is limited, and many studies were conducted in simulated, not clinical, settings. Moreover, the development of these tools was not informed by theory. Our project is unique in that the handover tool we developed includes a structured framework intended to mitigate cognitive biases that can contribute to communication errors, an often overlooked aspect in the domain of handover communication. Healthcare professionals often use heuristics or ‘shortcuts’ to formulate initial diagnoses. Although this approach is practical and adequate in most situations, there can be instances of failed heuristics. The failed heuristics are often seen and labelled as cognitive biases, which are consistent deviations in judgement due to information processing limits, decision-making shortcuts and emotional, moral or social influences. The process of diagnostic clinical reasoning and reducing diagnostic cognitive errors has been identified as one of the blind spots in patient safety and healthcare. In fact, cognitive errors have been demonstrated to contribute to up to 74% of system-related factors and up to 65% of diagnostic errors in Internal Medicine. Furthermore, Emergency Medicine is considered to be an ill-structured and chaotic environment at high risk of producing cognitive errors. In such conditions, the combination of poor communication strategies combined with a failed heuristic that might be carried on during the transition of care will create the perfect storm for adverse patient outcomes. The development of the BRIEF-C was based on several premises. First, clinical handover is a high-risk, high-frequency procedure that occurs in healthcare where the majority of communication failures could be averted by using a handoff tool. Second, improving the diagnostic process has emerged as a crucial benchmark for patient safety. Third, the heuristics and biases research suggests that quality decision-making might be influenced by judgement blind spots. Fourth, the approach to handovers was reframed as an opportunity for enhancing diagnostic care through system and process improvements. Such interventions should address the two dimensions of the linguistic dialogue during handovers: the informational structure embedded within an active collaborative format. Fifth, translation of communication strategies from high-reliability organisations. Here, tailoring some of the learnt lessons (e.g., face-to-face, interactive questioning, readback, limiting interruptions, delaying the transfer of responsibility during critical actions, overhearing others’ updates) will support grounding in communication to establish that what has been said is understood and interpredictability of actions and behaviours during the joint activity of handover exchange. Enhancing effective transitions across the patient care continuum is expected to result in more effective and timely care, foster improved communication among diverse healthcare professionals and patients, and improve planning for discharge and follow-up care in the community. There is a need for improved tools and standardised processes to narrow, and ultimately eliminate, the gap between transmitting and receiving patient information, in addition to considering the diagnostic safety at times of handover and transition to care. For example, surgical residents in Canada have reported their involvement in preventable patient injuries, underlining the importance of refining communication while integrating the diagnostic reasoning process for reducing patient harm and addressing our psychological tendencies to accept diagnostic suggestions. In summary, this study examines psychometric evidence of the BRIEF-C tool that was developed based on cognitive bias theory to see if its use improves communication during handovers. The study includes residents and physicians from two diverse clinical disciplines: Internal Medicine and Orthopaedic Surgery. While diverse in context, healthcare providers in both disciplines have to communicate and make judgements under parallel conditions of clinical uncertainty. By focusing on Internal Medicine and Orthopaedic Surgery specialties, the study captures scenarios with potential high-stakes outcomes. This choice enhances the practical applicability of the study’s findings, as effective communication in these settings is crucial to avoid errors, ensure proper diagnosis and treatment, and ultimately enhance patient well-being. Theoretical framework for tool development To address the need for effective communication frameworks that mitigate the potential for failed heuristics at times of transition of care, a member of the research team member (GA) initiated the tool development process by designing the core clinical data items. Feedback and redesign refinement were then completed in consultation with an expert in high-reliability organisations expert (RB), who provided insights drawn from strategies in high-risk industries like aviation. The result is a clinical handover tool—the BRIEF-C (see ). While the research identifies a plethora of biases, the design of our handover tool focuses on three biases that seem most prevalent in our clinical experience. The first bias, order effect, leads to better recall of information shared at the beginning and end of a conversation, often forgetting details in the middle. The BRIEF-C addresses this bias by systematically organising clinical data flow. Each item focuses on crucial clinical data required for effective information exchange. Confirmation bias, the second bias, involves seeking evidence to support a hypothesis, neglecting contradictory evidence. Our BRIEF-C tool incorporates two items: one for including and another for excluding differential diagnoses during handover. These items spotlight how discussions should include all considered possibilities and the presentation of evidence to rule-out tentative hypotheses. The third bias, momentum bias, occurs when a diagnostic label becomes entrenched through intermediaries when it might have just started as a possibility. This bias is represented in the last two items in regarding communication where receivers of the information summarise their understanding (read back) and senders actively listen (hear back). These two actions may lead to further discussion to resolve misunderstandings, challenge assumptions and reduce diagnostic errors. Notably, BRIEF-C triggers practitioners involved in the handover to share their understanding, or mental models, identified as crucial in communication and often missed in other handover tools. Examination of validity Evidence of content validity was obtained by a literature review on common influences on decision-making in healthcare ; experience and observations from clinical simulation environments ; and three specific cognitive biases. To obtain evidence of construct validity, a principal component analysis (PCA) was conducted using varimax rotation to obtain the most interpretable factors. Prior to conducting the PCA, the suitability of the data was evaluated using Kaiser-Meyer-Olkin and Bartlett’s test of sphericity scores. For each component (or factor), items were considered based on eigenvalues greater than or equal to 1.0, while the highest factor loadings were used to determine the number and meaning of the components. Initially, the measurement included an eighth item related to the patient’s status (e.g., current vital signs). However, including this item created several split loadings, uninterpretable factors, and less explained variance than with this item excluded. Thus, this item was excluded from further analyses and a subsequent PCA iteration. A second PCA was conducted on the seven items from the BRIEF-C tool, shown in . The KMO indicator was 0.74 and Bartlett’s test of sphericity was p<0.05, both suggesting an adequate degree of correlation among the tool’s items—which is an indication that the data meet the statistical assumptions for running a PCA. Across three iterations (i.e., number of attempts to obtain a solution with one or more factors), a total of two factors explained 71.72% of the variance (a large majority is considered desirable), as indicated in the table with respective factor (or correlation) loadings. This analysis revealed two factors, with the first five BRIEF-C items measuring diagnostic clinical reasoning and the last two items identifying communication. Thus, two subscales were obtained. In other words, the BRIEF-C appears to measure two types of information shared in the handover. The first is diagnostic clinical reasoning measured by the first five BRIEF-C items shown in . The second type of handover information is communication, which is measured by the two BRIEF-C items of communication read and hear back. As shown in the second last row of the table, the consistency of the five items measuring diagnostic clinical reasoning was high (0.82) as was the consistency of the two items measuring communication (0.99). To demonstrate consistency between raters, inter-rater reliability was calculated with Cohen’s kappa. As seen in the last row, adequate to high consistency was found. Sample We conducted an intervention pre–post-test design involving two assessment times (pre-test and post-test) within the Departments of Surgery and Internal Medicine of a large urban academic hospital in Midwestern Canada. Observations of internists and surgeons took place while they conducted handovers for current hospital patients from September 2017 to March 2018. A total of seven groups participated in the pre-test and seven in the post-test. Those groups consisted of rounding residents on the clinical services of inpatient Internal Medicine and inpatient Orthopaedic Surgery. Residents rotate monthly, and standard operation for handover consists of an in-person meeting between the night-time on-call team and the daytime incoming team. The Orthopaedic Surgery night-time on-call team is usually composed of one faculty and 1 to 2 junior and senior residents who provide a handover at 07:00 hours to 1 daytime faculty and 3 to 4 junior and senior residents. The inpatient Internal Medicine night-time on-call team is usually composed of 2 junior and one senior resident who provide a handover at 08:00 hours to 2 daytime faculty and 6 to 8 junior residents and medical students. All residents attending the handovers were the same between the pre-intervention and post-intervention weeks for each of the different monthly clinical rotations and participated in providing or receiving the handovers on different days depending on their call schedule. Each group discussed 1 to 10 clinical handovers of new admissions (Mean=3.28; SD=2.20) summing up a total of 176 patients discussed during 144 handover meetings. An a priori power calculation using G*Power V.3.1.9.2 with a conventional small effect size of 0.25, alpha of 0.05 and power of 0.85, indicates a required sample size of 146 handovers. This effect size was selected as it is similar to results obtained by Dory et al . It is noted that only 144 handovers were included in this study due to technical problems. Procedure Handovers that took place during the first week of clinical rotations for Internal Medicine and Orthopaedic Surgery at a tertiary care hospital were audio recorded. They occurred during the daily in-person handovers at time of transition of care between the night-time on-call team and the daytime incoming team. The recording device was set up on the side of the room to make it less intrusive, but the research assistant was also present. After a week of pre-test data collection, an educational intervention on the use of the BRIEF-C was started. Both faculty and residents in Internal Medicine and Oorthopaedic Surgery were invited to, individually or in small groups, complete a 30-minute face-to-face PowerPoint tutorial at different times depending on availability over a 1-week period. The training presentation was developed and delivered by a research team member (GA), and it included an explanation of the BRIEF-C items along with applied examples of patient handoffs. Additional time was provided for practice applications on newly admitted patients with the on-call team. Beginning the following week, residents and faculty were encouraged to use the BRIEF-C tool, and feedback was provided by GA or rounding faculty. As a reminder, laminated and coloured cards were taped on computer screens in the meeting rooms, showing the components of the BRIEF-C tool. In the final and fourth week of each of the monthly clinical rotations, post-test data collection was completed with audio recordings of handovers. Handovers in the fourth week were conducted by the same group of residents and medical students who were rounding in the first week of pre-data collection, and they all participated in providing or receiving handovers on different days depending on their call schedule (i.e, give a handover if part of the night-time on-call team, or receive a handover if part of the daytime patient care team alternatively on different days). All transcripts of the audio recordings were evaluated independently by two BRIEF-C trained raters, who were blinded to the residency group (i.e., Internal Medicine or Orthopaedic Surgery), as well as time of assessment (i.e., pre-test, post-test). One rater was an internist (GA) who also developed the BRIEF-C tool and the educational module. The second rater was an orthopaedic surgeon (MC) who was trained in the use of and scoring with the BRIEF-C tool. Neither rater was aware of whether the audio recordings occurred during pre-test or post-test. Both raters participated in a practice scoring session with the tool until they achieved 90% consensus on sample audio recordings. Consent and patient data were managed as follows. First, the study was approved by the University of Calgary Conjoint Health Research Ethics Board (REB16-0499). Second, a research assistant (rather than a preceptor) obtained informed consent from the medical and surgical staff. Third, all participant learners were informed that their participation would not affect their rotation assessment and that they could withdraw without penalty. Fourth, the audio recordings and transcriptions of all handovers were assigned a number rather than by name, and patient data were de-identified. Fifth, the recording device was set up on the side of the room to make it less intrusive, and the research assistant monitored the device to ensure its proper use and safe keeping. Sixth, all transcripts of the audio recordings were evaluated independently by two BRIEF-C trained raters, who were blinded to the residency group (i.e., Internal Medicine or Orthopaedic Surgery), as well as the time of assessment (i.e., pre-test and post-test). Finally, the research assistant signed the health authority’s confidentiality form, which is a standard procedure for all researchers. Intervention analysis Descriptive statistics in the form of mean, SD, and frequency counts were used to describe the data. To determine if the BRIEF-C scores changed after the educational intervention, one-sample t-tests were conducted. Specifically, the mean of the pre-test scores served as the designated value against which to compare the post-test mean score. This test was selected because the pre-samples and post-samples were not matched. This test was also used to examine differences in the time required to complete handovers before and after the intervention. Furthermore, to compare differences in handover scores between surgical and internal medicine residents, an independent samples t-test was conducted. A p-value of 0.05 served as the criterion to judge statistical significance and effect sizes were reported where results were significant to provide a comprehensive understanding of the outcomes. To address the need for effective communication frameworks that mitigate the potential for failed heuristics at times of transition of care, a member of the research team member (GA) initiated the tool development process by designing the core clinical data items. Feedback and redesign refinement were then completed in consultation with an expert in high-reliability organisations expert (RB), who provided insights drawn from strategies in high-risk industries like aviation. The result is a clinical handover tool—the BRIEF-C (see ). While the research identifies a plethora of biases, the design of our handover tool focuses on three biases that seem most prevalent in our clinical experience. The first bias, order effect, leads to better recall of information shared at the beginning and end of a conversation, often forgetting details in the middle. The BRIEF-C addresses this bias by systematically organising clinical data flow. Each item focuses on crucial clinical data required for effective information exchange. Confirmation bias, the second bias, involves seeking evidence to support a hypothesis, neglecting contradictory evidence. Our BRIEF-C tool incorporates two items: one for including and another for excluding differential diagnoses during handover. These items spotlight how discussions should include all considered possibilities and the presentation of evidence to rule-out tentative hypotheses. The third bias, momentum bias, occurs when a diagnostic label becomes entrenched through intermediaries when it might have just started as a possibility. This bias is represented in the last two items in regarding communication where receivers of the information summarise their understanding (read back) and senders actively listen (hear back). These two actions may lead to further discussion to resolve misunderstandings, challenge assumptions and reduce diagnostic errors. Notably, BRIEF-C triggers practitioners involved in the handover to share their understanding, or mental models, identified as crucial in communication and often missed in other handover tools. Evidence of content validity was obtained by a literature review on common influences on decision-making in healthcare ; experience and observations from clinical simulation environments ; and three specific cognitive biases. To obtain evidence of construct validity, a principal component analysis (PCA) was conducted using varimax rotation to obtain the most interpretable factors. Prior to conducting the PCA, the suitability of the data was evaluated using Kaiser-Meyer-Olkin and Bartlett’s test of sphericity scores. For each component (or factor), items were considered based on eigenvalues greater than or equal to 1.0, while the highest factor loadings were used to determine the number and meaning of the components. Initially, the measurement included an eighth item related to the patient’s status (e.g., current vital signs). However, including this item created several split loadings, uninterpretable factors, and less explained variance than with this item excluded. Thus, this item was excluded from further analyses and a subsequent PCA iteration. A second PCA was conducted on the seven items from the BRIEF-C tool, shown in . The KMO indicator was 0.74 and Bartlett’s test of sphericity was p<0.05, both suggesting an adequate degree of correlation among the tool’s items—which is an indication that the data meet the statistical assumptions for running a PCA. Across three iterations (i.e., number of attempts to obtain a solution with one or more factors), a total of two factors explained 71.72% of the variance (a large majority is considered desirable), as indicated in the table with respective factor (or correlation) loadings. This analysis revealed two factors, with the first five BRIEF-C items measuring diagnostic clinical reasoning and the last two items identifying communication. Thus, two subscales were obtained. In other words, the BRIEF-C appears to measure two types of information shared in the handover. The first is diagnostic clinical reasoning measured by the first five BRIEF-C items shown in . The second type of handover information is communication, which is measured by the two BRIEF-C items of communication read and hear back. As shown in the second last row of the table, the consistency of the five items measuring diagnostic clinical reasoning was high (0.82) as was the consistency of the two items measuring communication (0.99). To demonstrate consistency between raters, inter-rater reliability was calculated with Cohen’s kappa. As seen in the last row, adequate to high consistency was found. We conducted an intervention pre–post-test design involving two assessment times (pre-test and post-test) within the Departments of Surgery and Internal Medicine of a large urban academic hospital in Midwestern Canada. Observations of internists and surgeons took place while they conducted handovers for current hospital patients from September 2017 to March 2018. A total of seven groups participated in the pre-test and seven in the post-test. Those groups consisted of rounding residents on the clinical services of inpatient Internal Medicine and inpatient Orthopaedic Surgery. Residents rotate monthly, and standard operation for handover consists of an in-person meeting between the night-time on-call team and the daytime incoming team. The Orthopaedic Surgery night-time on-call team is usually composed of one faculty and 1 to 2 junior and senior residents who provide a handover at 07:00 hours to 1 daytime faculty and 3 to 4 junior and senior residents. The inpatient Internal Medicine night-time on-call team is usually composed of 2 junior and one senior resident who provide a handover at 08:00 hours to 2 daytime faculty and 6 to 8 junior residents and medical students. All residents attending the handovers were the same between the pre-intervention and post-intervention weeks for each of the different monthly clinical rotations and participated in providing or receiving the handovers on different days depending on their call schedule. Each group discussed 1 to 10 clinical handovers of new admissions (Mean=3.28; SD=2.20) summing up a total of 176 patients discussed during 144 handover meetings. An a priori power calculation using G*Power V.3.1.9.2 with a conventional small effect size of 0.25, alpha of 0.05 and power of 0.85, indicates a required sample size of 146 handovers. This effect size was selected as it is similar to results obtained by Dory et al . It is noted that only 144 handovers were included in this study due to technical problems. Handovers that took place during the first week of clinical rotations for Internal Medicine and Orthopaedic Surgery at a tertiary care hospital were audio recorded. They occurred during the daily in-person handovers at time of transition of care between the night-time on-call team and the daytime incoming team. The recording device was set up on the side of the room to make it less intrusive, but the research assistant was also present. After a week of pre-test data collection, an educational intervention on the use of the BRIEF-C was started. Both faculty and residents in Internal Medicine and Oorthopaedic Surgery were invited to, individually or in small groups, complete a 30-minute face-to-face PowerPoint tutorial at different times depending on availability over a 1-week period. The training presentation was developed and delivered by a research team member (GA), and it included an explanation of the BRIEF-C items along with applied examples of patient handoffs. Additional time was provided for practice applications on newly admitted patients with the on-call team. Beginning the following week, residents and faculty were encouraged to use the BRIEF-C tool, and feedback was provided by GA or rounding faculty. As a reminder, laminated and coloured cards were taped on computer screens in the meeting rooms, showing the components of the BRIEF-C tool. In the final and fourth week of each of the monthly clinical rotations, post-test data collection was completed with audio recordings of handovers. Handovers in the fourth week were conducted by the same group of residents and medical students who were rounding in the first week of pre-data collection, and they all participated in providing or receiving handovers on different days depending on their call schedule (i.e, give a handover if part of the night-time on-call team, or receive a handover if part of the daytime patient care team alternatively on different days). All transcripts of the audio recordings were evaluated independently by two BRIEF-C trained raters, who were blinded to the residency group (i.e., Internal Medicine or Orthopaedic Surgery), as well as time of assessment (i.e., pre-test, post-test). One rater was an internist (GA) who also developed the BRIEF-C tool and the educational module. The second rater was an orthopaedic surgeon (MC) who was trained in the use of and scoring with the BRIEF-C tool. Neither rater was aware of whether the audio recordings occurred during pre-test or post-test. Both raters participated in a practice scoring session with the tool until they achieved 90% consensus on sample audio recordings. Consent and patient data were managed as follows. First, the study was approved by the University of Calgary Conjoint Health Research Ethics Board (REB16-0499). Second, a research assistant (rather than a preceptor) obtained informed consent from the medical and surgical staff. Third, all participant learners were informed that their participation would not affect their rotation assessment and that they could withdraw without penalty. Fourth, the audio recordings and transcriptions of all handovers were assigned a number rather than by name, and patient data were de-identified. Fifth, the recording device was set up on the side of the room to make it less intrusive, and the research assistant monitored the device to ensure its proper use and safe keeping. Sixth, all transcripts of the audio recordings were evaluated independently by two BRIEF-C trained raters, who were blinded to the residency group (i.e., Internal Medicine or Orthopaedic Surgery), as well as the time of assessment (i.e., pre-test and post-test). Finally, the research assistant signed the health authority’s confidentiality form, which is a standard procedure for all researchers. Descriptive statistics in the form of mean, SD, and frequency counts were used to describe the data. To determine if the BRIEF-C scores changed after the educational intervention, one-sample t-tests were conducted. Specifically, the mean of the pre-test scores served as the designated value against which to compare the post-test mean score. This test was selected because the pre-samples and post-samples were not matched. This test was also used to examine differences in the time required to complete handovers before and after the intervention. Furthermore, to compare differences in handover scores between surgical and internal medicine residents, an independent samples t-test was conducted. A p-value of 0.05 served as the criterion to judge statistical significance and effect sizes were reported where results were significant to provide a comprehensive understanding of the outcomes. By using the demographic data available for the sample, we examined if the duration of handovers differed between internal medicine and surgical residents. Residents in Internal Medicine took longer (in seconds) to complete handovers (M=270.82 s, SD=253.19) than did residents in Surgery (M=208.47 s, SD=208.13), t (140)=2.06, p<0.05, Cohen’s d=0.35 (small-medium). There was no difference, however, in the duration of clinical handovers at pre-test (M=223.86 s, SD=149.47) compared with post-test (M=271.78 s, SD=207.32), t(64)=1.86, p>0.05. Pre-test versus post-test The scores on the BRIEF-C items are shown in . An increase was observed in the number of residents who completed various types of information sharing after the educational intervention. In particular, no one was rated as completing read-back and hear-back communication strategies in the pre-test, whereas almost a quarter of participants did in post-test. The mean of each subscale was then calculated by summing the respective items and dividing by the number of items within that subscale (five diagnostic clinical reasoning items and two communication items). These reasoning and communication variable scores were then compared between pre-test and post-test. Results showed a significant improvement from pre-measures of the diagnostic clinical reasoning items (M=0.97, SD=0.50) to post-test (M=1.31, SD=0.64), t(64)=4.26, p<0.05, Cohen’s d=0.63 (medium-large). The pre-communication scores (M=0.02, SD=0.16) also significantly improved at post-test (M=0.48, SD=0.83), t(64)=4.52, p<0.05, Cohen’s d=0.83 (large). The scores on the BRIEF-C items are shown in . An increase was observed in the number of residents who completed various types of information sharing after the educational intervention. In particular, no one was rated as completing read-back and hear-back communication strategies in the pre-test, whereas almost a quarter of participants did in post-test. The mean of each subscale was then calculated by summing the respective items and dividing by the number of items within that subscale (five diagnostic clinical reasoning items and two communication items). These reasoning and communication variable scores were then compared between pre-test and post-test. Results showed a significant improvement from pre-measures of the diagnostic clinical reasoning items (M=0.97, SD=0.50) to post-test (M=1.31, SD=0.64), t(64)=4.26, p<0.05, Cohen’s d=0.63 (medium-large). The pre-communication scores (M=0.02, SD=0.16) also significantly improved at post-test (M=0.48, SD=0.83), t(64)=4.52, p<0.05, Cohen’s d=0.83 (large). The results of the study indicate that there was a notable (medium to large size) improvement in the quality of communication that occurred during handovers, following the adoption of structured communication using the BRIEF-C. In addition, evidence of several types of validity and reliability was obtained. Communication improvement may have occurred for several reasons. The structured format of the BRIEF-C outlines all the areas of pertinent patient clinical information and actions on the part of healthcare professionals. By ensuring that all areas are addressed, there is less possibility of omitting important information. Second, engaging in a standardised approach ensures critical information is consistently communicated to minimise variations in practices and improve overall continuity of care. Third, the inclusion of different opportunities for open dialogue safeguards against three common human tendencies for bias (order, confirmation, momentum), which supports the mandate for diagnostic safety and adaptive decision-making. Fourth, the BRIEF-C’s applicability across different specialties, as evident from the diverse backgrounds of experts involved, may promote interdisciplinary team collaboration and facilitate smoother transitions when patients are handed over between different healthcare teams. Our results, therefore, seem to be relevant to a variety of handover conversations. It is noted, however, that the rate of completion at post-test remained at under 50% for most of the BRIEF-C items. Thus, there is room for further improvement that can be promoted through ongoing education, practice, monitoring, and feedback. A particular strength of this study was the use of this tool within a clinical setting, which represents authentic applications in patient handovers. There was no standardisation or other control of patient data. The advantage of limited standardisation is that the BRIEF-C was applied to heterogeneous presentations of clinical cases, akin to the diverse scenarios encountered in typical hospital environments. Several factors could explain the observation that residents specialising in Internal Medicine took more time for handovers compared with their Surgical counterparts. One factor may involve the complexity of cases and information depth; Internal Medicine cases often involve complex medical histories and comorbidities. This complexity may have required more detailed communication during handovers, leading to extended discussions. Another factor may be that Internal Medicine physicians develop comprehensive treatment plans that include interdisciplinary considerations, leading to more extensive handover discussions. Moreover, we cannot disregard the possibility of communication norms, where different specialties might have varying norms for the level of detail and thoroughness expected during handovers, influencing the duration of these discussions. Interestingly, there was no difference in the duration of clinical handovers when comparing pre-test and post-test phases, suggesting that the tool does not demand additional time resources. Instead of burdening healthcare providers at times of transition of care, it seems that the tool facilitates the provision of effective communication and the process of diagnostic reasoning. Various forms of evidence of validity and reliability converge for the BRIEF-C’s scores as an indicator of quality communication during patient handovers. In terms of content validity, diverse background experts (high-reliability organisation, Medicine, Surgery) participated in the tool evaluation. Their collective knowledge and experience enable the BRIEF-C items to assess communication during handover accurately. This diversity also broadened the scope of content validity. The BRIEF-C tool comprehensively addresses healthcare information needs across various specialties while integrating safeguards to promote effective communication and mitigate cognitive biases. Specifically, the explicit listing of the ruled-in and ruled-out diagnoses offers an opportunity to mitigate confirmation bias. At the same time, the read/hear-back communication component invites open inquiry and discussion. Such discussion extends beyond traditional summary statements with a focus on pinning down decision-making based on intuition, logic or both while bringing clarity to the interaction of reasoning strategies (e.g., deductive vs inductive). Furthermore, this read/hear-back phase is situated at a critical time during the handover dialogue where the attention of participants is heightened (order effect bias), offering a barrier to our tendency to accept offered conclusions (i.e., momentum bias) rather than refuting such confirmations. Factorial validity was confirmed through an analysis of the scale’s underlying factor structure, encompassing diagnostic clinical reasoning and communication factors. Both of these factors represent quality communication about a patient during handover. Regarding construct validity, measurement experts explain that a change in scores on the construct being measured after an intervention provides evidence that the defined factors (e.g., diagnostic clinical reasoning and communication) were represented in the scores. In the present study, the intervention’s impact on scores (i.e., BRIEF-C scores improved after participants learnt about how all the items represent quality information that needs to be shared during handover) provided evidence of construct validity. Evidence of reliability was obtained through good inter-rater reliability between two raters from differing specialisations (Internal Medicine and Surgery). This result underscores the robust nature of the BRIEF-C tool in producing consistent ratings across healthcare professionals with distinct training and clinical backgrounds. Moreover, the raters in our study are clinicians from within the evaluated disciplines, who are experts and most experienced on the specific performance objectives that were assessed in the measure, and, thus, best suited to evaluate them. Also, high inter-item reliability for each factor indicates that each factor’s items are similarly rated. This finding suggests that all the items are measuring the construct that each factor (diagnostic reasoning and communication) represents. The consistency of ratings between raters was in the acceptable to high range, suggesting that perhaps more training on using the BRIEF-C is warranted. The results of this study satisfy the first step in the validation pathway, offering robust evidence of validity and reliability. Future studies should examine other forms of accuracy and consistency. For example, consequential validity can be explored by measuring cognitive biases that may be reduced due to education and administration of the BRIEF-C. Also, pre-test and post-test evaluations can measure improved patient care and outcomes. Regarding strengths and limitations, since data were collected in a clinical environment involving residents and healthcare professionals, results may be more generalisable than if the data had been collected in a simulation environment. Indeed, if the emotions, thoughts and real system pressures/expectations that drive us to perform can only be approximated in simulation—according to its definition, then the ability to complete a study in an actual clinical environment is certainly noteworthy. However, it is important to note that the specific clinical site was not chosen randomly. It is, thus, advisable to extend the application of the BRIEF-C to various clinical settings to gather broader evidence of its generalisability. The study occurred in a busy clinical environment where healthcare providers were actively engaged in clinical duties so it was not feasible to randomly select participants or raters. Also, we were able to access a large number of handovers in a busy clinical environment, with only two handovers remaining inaccessible. The absence of a comparison group (a group that did not have the education intervention and was administered the BRIEF-C twice), means that alternative explanations for the observed improvement in the BRIEF-C scores cannot be completely ruled out. For example, it is possible that improvement was due to learning through experience gained during the month-long rotation rather than due solely to the use of the BRIEF-C. It is also possible that handover communication improved as a result of awareness of participating in a study and being audio recorded, but this explanation does not account for lower scores at pre-test than at post-test. Future studies should include data collection at 6 months to evaluate follow-up changes in handover practices and confirmation of continued use of the BRIEF-C tool. Another primary source of bias in the present study is that the education facilitator, a member of the research team, developed the education intervention and completed the BRIEF-C items at pre-test and post-test. This bias might have resulted in lower scores during the former, and higher scores during the latter administration. On the other hand, we implemented blinding strategies to ensure the education facilitator was unaware of the participant groups to help minimise conscious or subconscious influence when scoring the transcripts. Nevertheless, it is possible that the content of the clinical presentation suggested whether it was a surgical or medical patient. On the other hand, the two raters were following a standardised and objective framework with a focus on the presence or absence of clinical elements applicable in both contexts, in addition to the fact that it was not a comparison between medicine and surgery. Additionally, we applied two different randomisation orders of the pretest and post-test transcripts for both raters to reduce the likelihood of systematic bias and ensure an even distribution of potential influence across both data collection points. While these methods attempted to reduce bias, we recommend that future evaluation of this handover tool include reviewers outside of the research team. In conclusion, our study’s results contribute to the growing evidence on effective handover practices. Our study encourages a culture of continuous improvement in handover practices. Regular use of the BRIEF-C and ongoing assessment of communication quality can drive awareness, accountability and refinement of handovers. Future direction should focus on the integration of such tools within electronic health records systems to facilitate and enhance the handover processes. It is also recommended that the BRIEF-C be studied across additional clinical roles and settings. |
Comparison of the overall fit of three-unit posterior fixed dental prostheses fabricated with laser sintering and conventional casting methods | fa2bade1-5028-473c-8792-c8ef592fd829 | 11847754 | Dentistry[mh] | Marginal and internal adaptation between the restoration and abutment tooth is crucial for the biomechanical strength of the restoration–tooth complex. Successful marginal and internal adaptation enhance the fracture resistance of the restoration . Marginal adaptation can be affected by tooth preparation, and it is important for the structural strength of fixed dental prostheses (FDPs) . Similarly, the internal fit is important for achieving the clinical success of FDPs . Stress concentrations may be reduced with improvements in the internal fit, which may increase the strength of dentures . The internal fit is evaluated by measuring the vertical difference between the internal surface of the framework and the axial wall of the prepared abutment . The marginal adaptation of a restoration can be measured by evaluating the fit between the margins of the restoration and the prepared abutment teeth . Marginal discrepancies between the abutment teeth and restoration behave as a reservoir for microorganisms, and eventually, caries and periodontal problems can be seen on abutment teeth . Although it was reported that marginal discrepancy values (MDVs) less than 120 μm should be considered clinically acceptable, there is no consensus on the maximum clinically acceptable MDV . Recently, the use of cobalt-chromium (Co-Cr) alloys to produce the metal framework of FDPs has increased . Co-Cr alloys have ideal mechanical properties, and the material is inexpensive when compared with other precious alloys . Co-Cr frameworks can be fabricated with both conventional casting (lost-wax) and laser sintering methods. The fabrication technique may affect the marginal and internal adaptation of the metal framework . To produce a metal framework with the conventional casting method, a wax pattern is prepared on the abutment teeth. Thereafter, investing, wax elimination, and casting are achieved, respectively . However, the conventional method is time-consuming and requires a technician’s skills to achieve an eligible wax pattern . Additionally, wax material has some disadvantages, such as thermal sensitivity, fragility, and a high coefficient of thermal expansion . Recently, computer-aided design and computer-aided manufacturing (CAD-CAM) systems have been used in combination with conventional casting methods . 3D-printed polymer (3DP) or CAD-CAM wax blocks may be preferred over manual carving of the wax pattern. After the design in the CAD program, wax patterns are achieved with CAM systems by subtractive manufacturing from wax blocks, or polymer patterns are produced with 3D-printing systems, and the wax or polymer patterns are cast with the conventional casting method . After the metal framework is designed in the CAD program, it can be produced directly with selective laser melting (SLM) or selective laser sintering (SLS) . SLM and SLS systems are additive manufacturing methods, and a high-energy carbon-dioxide laser beam is used to fuse or sinter Co-Cr metal powder particles in layers until the designed framework is obtained . Furthermore, direct metal laser sintering (DMLS) was introduced as an SLS method. One or multiple laser beams are used to construct a metal framework from metal powders by sintering approximately 20- to 60-µm-thick layers at each shooting . The general advantages of laser sintering systems are saving time and eliminating technician-dependent errors and material-dependent limitations in wax carving, investing, and casting procedures. On the other hand, the marginal and internal fit of the restorations may be different depending on the abutment teeth, even if the fabrication method was the same . Limited studies have comparatively evaluated three-unit Co-Cr dental restoration substructure fabrication methods for different abutment teeth . Although the preparation principles are general, morphological differences still exist after the preparations, and these may impact the fit of FDPs . There are no definite conclusions regarding the effect of different fabrication methods on the marginal and internal fit in planning FDPs in posterior teeth . Previous studies that comparatively evaluated the effect of various fabrication methods on the adaptation of FDPs prepared in different posterior abutment teeth are not sufficient to determine a consensus . The purpose of this study was to compare different fabrication methods for posterior FDPs and provide clinical insights into the effect of fabrication methods on the fit of FDPs prepared for different types of abutment teeth. The first null hypothesis of the study was that the Co-Cr metal framework fabrication method would have no effect on the marginal, occlusal, and internal fit of the three-unit posterior FDP framework. The second null hypothesis was that there would be no differences between the marginal, occlusal, and internal fit of three-unit posterior FDPs created with the same fabrication method. Sample size and prior power analysis were achieved with G*Power v3.1 for “Means – Many groups ANOVA: main effects and interactions.” A medium effect size f of 0.25 with alpha 0.05, power 0.80, numerator df 4 (five manufacturing methods and molar-premolar abutment types), and number of groups 10 resulted in the required number of specimens, 196. Therefore, 200 abutment teeth specimens and 20 FPD framework specimens were prepared per fabrication group. With a laser plastic sintering machine (Formiga P 110, EOS GmbH, Krailling, Germany), a maxillary master model was prepared from fine polyamide powder (Polyamide 12, PA 2105; EOS GmbH). Three-unit FDPs were prepared for posterior regions on the segmented master models. For posterior FDPs, right first premolar and first molar teeth were prepared (Fig. ). Preparations of the abutment teeth were made according to the principles of general tooth preparation: 1.5 to 2 mm on the occlusal surfaces, 1 to 1.5 mm on the axial surfaces, and a convergence angle of approximately 6 degrees with chamfer finish lines. Thereafter, prepared teeth in the master models were evaluated with a parallelometer (Dental Surveyor; Unident Instrument Pvt., Ltd., New Delhi, India) to prevent undercut formation. After final adjustments were made, digital or conventional impressions were made from the master model according to the fabrication method used for each group. Five different fabrication methods were utilized to prepare a total of 100 Co-Cr three-unit FDP metal framework specimens ( n = 20) (Fig. ). DMLS Three-dimensional models were made by scanning the master model with the prepared teeth with a desktop scanner (DentalWings 7 Series; DentalWings, Montreal, Canada). The design program (DWOS crown and bridge CAD Software, DentalWings) was used to virtually design the metal frameworks. The copings of Co-Cr FDP frameworks were set as 0.5 mm in thickness. The virtual design was adjusted to obtain a 70-µm cement space on the entire inner surface 0.5 mm from the margin levels. The connector areas of the frameworks were set to 7 mm 2 for FDP specimens. A total of 20 Co-Cr three-unit FPD frameworks were produced by the DMLS machine (EOS M 270; EOS GmbH), using its recommended powder (SP2: EOS SP2; Turku, Finland), according to the manufacturer’s instructions. SLM The same design procedure as in the DMLS group was followed, and a total of 20 Co-Cr three-unit FDP frameworks were produced by the SLM machine (ConceptLaser Mlab, ConceptLaser GmbH, Lichtenfels, Germany) according to the manufacturer’s instructions using its recommended powder (Remanium ® star; Dentaurum, Ispringen, Germany). Manual wax carving (cast) Impressions of the master model were made with vinyl polysiloxane (VPS) (Elite HD + Putty soft; Zhermack SpA, Badia Polesine, Italy), and plaster (Hera Moldastone CN; Heraeus Kulzer GmbH, Hanau, Germany) models were obtained. Thereafter, removable die models were prepared with the Pindex system (ED-Laser; Dentalfarm Srl, Turin, Italy). Die spacers were utilized to standardize the cement space (Durolan; DFS Diamon, Riedenburg, Germany). A total of 70 μm of cement space was created by applying four layers of 15-µm-thick (Durolangold; DFS Diamon) die spacer and one layer of 10-µm-thick die spacer (Durolanblue; DFS Diamon) to abutment teeth, except 0.5-mm marginal areas. Wax patterns were prepared using modeling wax (GEO Dip gelb; Renfert GmbH, Hilzingen, Germany) melted in a melting unit (Hotty; Renfert GmbH). To standardize the dimensions of the pontics and connector parts of the cast FDP specimens, the pontics and connector parts of wax FDP patterns were milled from wax blocks (Alliance Wax Disc; Almadent, Izmir, Turkiye) using the standard CAD design and were connected to wax copings using the electric waxing unit (Hot 2 touch; Dentalfarm, Turin, Italy). After spruing, specimens were invested with an investment material (Castorit-Super C; Dentaurum, Ispringen, Germany). After the preheating and wax-elimination process at 950 °C, Co-Cr ingots (Kera C; Eisenbacher Dentalwaren ED GmbH, Main, Germany) were melted at approximately 1400 °C in an induction casting machine (INF 2010; Mikrotek Dental, Ankara, Turkiye) using a ceramic crucible (Bego Fornax; Bego, Bremen, Germany). After divestments, the specimens were examined under a 5× laboratory microscope. Specimens with unacceptable casting defects were excluded from the study, and new specimens were fabricated. A total of 20 Co-Cr FDP frameworks were prepared with the conventional casting method. CAD/CAM (3DP) In this group, the specimens were prepared according to CAD designs using translucent polymeric material (Fusia RF080; DWS Srl, Thiene, Italy) utilizing a laser stereolithography machine (Digitalwax 028D; DWS Srl). The specimens were stored in a light-resistant box for 48 h and tried on the abutment teeth before casting. The polymeric FDP patterns were sprued, invested, and cast from a Co-Cr alloy (Kera C; Eisenbacher Dentalwaren ED GmbH) according to the conventional casting procedures. A total of 20 Co-Cr FDP frameworks were prepared. CAD/CAM Wax (CAW) The CAW specimens were milled from CAD/CAM wax blocks (Alliance Wax; Almadent), and the specimens were prepared according to the conventional casting method described for the Cast and 3DP groups. A total of 20 Co-Cr FDP frameworks were fabricated. The marginal and internal fit of the frameworks were examined with the silicone replica technique (SRT). VPS materials of different colors and viscosities were used to detect discrepancies between the three-unit FDP metal frameworks and prepared abutment teeth. VPS material (Fit Checker Advanced White; GC Corp., Tokyo, Japan) was applied to the copings of FDP frameworks. FDP frameworks were seated onto abutment teeth with finger pressure. To apply an equal load during the setting of silicone material, a vertical pressure of 40 N was applied onto the pontics of the specimens with a digital dynamometer device (Digital Force Gauge SH-500, Geratech Ltd., Hong Kong, Taiwan) for one minute. After the specimens were removed, the inside of the specimens was filled up with light-body VPS impression material (Betasil Light Vario, Müller-Omicron GmbH&Co., KG., Lindlar, Germany). Silicone indexes were cut by a razor in both the mesiodistal and buccolingual directions (Fig. ). Marginal and internal discrepancy measurements were made under a stereomicroscope (AZ100M motorized multipurpose zoom microscope; Nikon Corp., Kanagawa, Japan) at 100× magnification (Fig. ). Nine measurement points can be seen from the buccal view in the mesiodistal direction measured for each abutment tooth (Fig. ). Furthermore, 10 measurement points can be seen from the distal view in the bucco-palatinal direction measured for each abutment tooth (Fig. ). A total of 19 measurement points (seven occlusal points, four axial points, four chamfer points, and four margin points) were evaluated for each abutment tooth, and a total of 38 measurement points were evaluated for each FPD framework specimen . At each measurement point, three measurements were made, and the mean discrepancy value was recorded . The mean marginal (four margin points), internal (four axial points and four chamfer points) and occlusal (seven occlusal points) discrepancy values (ODVs) for each abutment tooth were obtained by calculating the mean values of all measurement points. Statistical analysis was done with the SPSS program (Version 22). Considering the distribution of the variables, skewness, and kurtosis values, the variables revealed a normal distribution, as the values were between + 1.5 and − 1.5. The effect of abutment type and fabrication method was evaluated with two-way ANOVA (α = 0.05). Pairwise comparisons of groups were evaluated with post hoc Bonferroni adjustment ( P adj . = 0.05/10, α = 0.005). Three-dimensional models were made by scanning the master model with the prepared teeth with a desktop scanner (DentalWings 7 Series; DentalWings, Montreal, Canada). The design program (DWOS crown and bridge CAD Software, DentalWings) was used to virtually design the metal frameworks. The copings of Co-Cr FDP frameworks were set as 0.5 mm in thickness. The virtual design was adjusted to obtain a 70-µm cement space on the entire inner surface 0.5 mm from the margin levels. The connector areas of the frameworks were set to 7 mm 2 for FDP specimens. A total of 20 Co-Cr three-unit FPD frameworks were produced by the DMLS machine (EOS M 270; EOS GmbH), using its recommended powder (SP2: EOS SP2; Turku, Finland), according to the manufacturer’s instructions. The same design procedure as in the DMLS group was followed, and a total of 20 Co-Cr three-unit FDP frameworks were produced by the SLM machine (ConceptLaser Mlab, ConceptLaser GmbH, Lichtenfels, Germany) according to the manufacturer’s instructions using its recommended powder (Remanium ® star; Dentaurum, Ispringen, Germany). Impressions of the master model were made with vinyl polysiloxane (VPS) (Elite HD + Putty soft; Zhermack SpA, Badia Polesine, Italy), and plaster (Hera Moldastone CN; Heraeus Kulzer GmbH, Hanau, Germany) models were obtained. Thereafter, removable die models were prepared with the Pindex system (ED-Laser; Dentalfarm Srl, Turin, Italy). Die spacers were utilized to standardize the cement space (Durolan; DFS Diamon, Riedenburg, Germany). A total of 70 μm of cement space was created by applying four layers of 15-µm-thick (Durolangold; DFS Diamon) die spacer and one layer of 10-µm-thick die spacer (Durolanblue; DFS Diamon) to abutment teeth, except 0.5-mm marginal areas. Wax patterns were prepared using modeling wax (GEO Dip gelb; Renfert GmbH, Hilzingen, Germany) melted in a melting unit (Hotty; Renfert GmbH). To standardize the dimensions of the pontics and connector parts of the cast FDP specimens, the pontics and connector parts of wax FDP patterns were milled from wax blocks (Alliance Wax Disc; Almadent, Izmir, Turkiye) using the standard CAD design and were connected to wax copings using the electric waxing unit (Hot 2 touch; Dentalfarm, Turin, Italy). After spruing, specimens were invested with an investment material (Castorit-Super C; Dentaurum, Ispringen, Germany). After the preheating and wax-elimination process at 950 °C, Co-Cr ingots (Kera C; Eisenbacher Dentalwaren ED GmbH, Main, Germany) were melted at approximately 1400 °C in an induction casting machine (INF 2010; Mikrotek Dental, Ankara, Turkiye) using a ceramic crucible (Bego Fornax; Bego, Bremen, Germany). After divestments, the specimens were examined under a 5× laboratory microscope. Specimens with unacceptable casting defects were excluded from the study, and new specimens were fabricated. A total of 20 Co-Cr FDP frameworks were prepared with the conventional casting method. In this group, the specimens were prepared according to CAD designs using translucent polymeric material (Fusia RF080; DWS Srl, Thiene, Italy) utilizing a laser stereolithography machine (Digitalwax 028D; DWS Srl). The specimens were stored in a light-resistant box for 48 h and tried on the abutment teeth before casting. The polymeric FDP patterns were sprued, invested, and cast from a Co-Cr alloy (Kera C; Eisenbacher Dentalwaren ED GmbH) according to the conventional casting procedures. A total of 20 Co-Cr FDP frameworks were prepared. The CAW specimens were milled from CAD/CAM wax blocks (Alliance Wax; Almadent), and the specimens were prepared according to the conventional casting method described for the Cast and 3DP groups. A total of 20 Co-Cr FDP frameworks were fabricated. The marginal and internal fit of the frameworks were examined with the silicone replica technique (SRT). VPS materials of different colors and viscosities were used to detect discrepancies between the three-unit FDP metal frameworks and prepared abutment teeth. VPS material (Fit Checker Advanced White; GC Corp., Tokyo, Japan) was applied to the copings of FDP frameworks. FDP frameworks were seated onto abutment teeth with finger pressure. To apply an equal load during the setting of silicone material, a vertical pressure of 40 N was applied onto the pontics of the specimens with a digital dynamometer device (Digital Force Gauge SH-500, Geratech Ltd., Hong Kong, Taiwan) for one minute. After the specimens were removed, the inside of the specimens was filled up with light-body VPS impression material (Betasil Light Vario, Müller-Omicron GmbH&Co., KG., Lindlar, Germany). Silicone indexes were cut by a razor in both the mesiodistal and buccolingual directions (Fig. ). Marginal and internal discrepancy measurements were made under a stereomicroscope (AZ100M motorized multipurpose zoom microscope; Nikon Corp., Kanagawa, Japan) at 100× magnification (Fig. ). Nine measurement points can be seen from the buccal view in the mesiodistal direction measured for each abutment tooth (Fig. ). Furthermore, 10 measurement points can be seen from the distal view in the bucco-palatinal direction measured for each abutment tooth (Fig. ). A total of 19 measurement points (seven occlusal points, four axial points, four chamfer points, and four margin points) were evaluated for each abutment tooth, and a total of 38 measurement points were evaluated for each FPD framework specimen . At each measurement point, three measurements were made, and the mean discrepancy value was recorded . The mean marginal (four margin points), internal (four axial points and four chamfer points) and occlusal (seven occlusal points) discrepancy values (ODVs) for each abutment tooth were obtained by calculating the mean values of all measurement points. Statistical analysis was done with the SPSS program (Version 22). Considering the distribution of the variables, skewness, and kurtosis values, the variables revealed a normal distribution, as the values were between + 1.5 and − 1.5. The effect of abutment type and fabrication method was evaluated with two-way ANOVA (α = 0.05). Pairwise comparisons of groups were evaluated with post hoc Bonferroni adjustment ( P adj . = 0.05/10, α = 0.005). There were significant differences between the fabricating methods and the abutment types in the measurement regions for the MDV ( P < 0.001), internal discrepancy value (IDV) ( P < 0.001), and ODV ( P < 0.001) (Table ). The discrepancy results of FDPs for each abutment tooth according to the fabrication methods are graphically shown in Fig. . Table indicates vertically the comparison of measurements obtained in the marginal, internal, and occlusal regions between abutment teeth for the same fabrication method. In addition, the comparison of measurements on the same abutment tooth between different fabrication methods are horizontally shown in Table . For both the Cast and DMLS methods, there were significant differences between the MDVs of premolar and molar abutments ( P < 0.05). When the fabrication methods were compared for molar abutment, the largest MDVs were determined for the Cast method ( P adj . < 0.005), and the smallest MDVs were determined for the DMLS method ( P adj. < 0.005). When IDVs were examined, significant differences were seen between premolar and molar abutments for the 3DP, SLM, and Cast fabrication methods ( P < 0.05). There was no significant difference between the 3DP and Cast fabrication methods for the IDVs of the same abutment tooth ( P adj . > 0.005). Similarly, there was no significant difference between the IDVs of the SLM, CAW, and DMLS fabrication methods for the same abutment ( P adj . > 0.005). For the 3DP and Cast fabrication methods, there were significant differences between the ODVs of premolar and molar abutments ( P < 0.05). However, for the ODVs of premolar abutments, there were no significant differences between the fabrication methods ( P adj . > 0.005). There were no significant differences between the ODVs of molar abutments when the 3DP and Cast fabrication methods ( P adj . > 0.005) were used. Similarly, significant differences were not defined between the ODVs of molar abutments when the SLM, CAW, and DMLS fabrication methods were used ( P adj . > 0.005). This study was conducted to compare the fit of three-unit posterior FDP frameworks fabricated with different techniques. Although there were no significant differences between the fabrication methods in terms of the ODVs of FDPs for premolar abutment teeth, significant differences were detected between the fabrication methods for molar abutment teeth. There were no significant differences between the discrepancy values of different abutments when certain fabrication methods were used, but there were significant differences between the different abutment teeth for the same fabrication method according to the measured region, such as marginal, internal, and occlusal. Therefore, both the first and second null hypotheses of the study were partially accepted. The insignificant difference in ODVs between fabrication methods for premolar abutments may be explained by the difference between the surface area of the abutment teeth. The premolar abutment has less angular morphology than the molar abutment, and the crown part of the premolar shows a more rounded form than the molar . Angular areas increase the probability of framework misfit due to the technical limitations of fabrication methods . The differences between the fabrication methods should be considered when preparing and working on more angular tooth surfaces. Accordingly, the highest ODVs in fabrication methods were obtained for molar abutments because the teeth inherently have a series of angles and curves on the occlusal surface . Therefore, chamfer finish line prefered as the finish line because, it has been reported that chamfer finish line presented less discrepancy values and more accuracy for impressions in previous studies . The discrepancy values of laser groups were lower than the values of 3DP, CAW, and Cast groups. This might be related to the possibility of the formation of an oxide layer in the casting process and gas porosity on irregular surfaces . Furthermore, the effect of the interaction between the investment material and Co-Cr could be a reason for the higher discrepancy values . Other parameters that might affect the fit of the FDPs before or during casting are the technician’s skills, contraction of wax material, and shrinkage of polymeric 3DP material . On the other hand, the diameter of the milling bur used in the CAD/CAM system may affect the accuracy of restorations . The milling bur may not sufficiently prepare sharp internal angles because of the size discrepancies between the preparation edges and the bur . Additionally, horizontal planes (x, y) and vertical planes (z) may affect the discrepancy values in CAM systems. Particularly errors in the z-plane may contribute to the shrinkage of the material . In line with this information, in this study, the IDVs and ODVs were higher than the MDVs of the 3DP and CAW groups that the CAD/CAM wax blocks were used. Consistent with previous studies, the ODVs were higher than the MDVs in this study, and the mean ODVs measured in the posterior teeth ranged between 88.61 μm and 141.49 μm and were lower than those of the previous studies . Whereas flat occlusal surfaces were prepared in other studies , in the present study, the occlusal side of the premolar and molar teeth were prepared according to tooth preparation principles, and the teeth morphologies and frameworks were not cemented before discrepancy measurements were performed. Additionally, when the marginal, internal, and occlusal discrepancies between the abutment teeth and FDPs were evaluated, a general increase in the discrepancy values from the marginal to the occlusal regions was observed. This difference might be because the more detailed the occlusal or internal structure is, the more complete fit is prevented due to the increase in surface area . The MDVs of the FDP substructures are affected by the prepared tooth morphology and cement thickness . In the present study, the aim was to standardize the cement space. Considering previous studies, the cement space prepared for all abutment teeth was set to be 70 μm . Thus, the narrowing in the axial regions could be hindered, and the passive fit of the FDPs to the abutment teeth was ensured . Moreover, in this study, there were differences between wax pattern preparation methods. This difference could be a reason for the inconsistent discrepancy values of the substructures obtained in the Cast, CAW, and 3DP groups, although the same alloy, Co-Cr, was cast . Although the lost-wax technique was used for each group, some steps, such as scanning and designing software conditions, might be responsible for discrepancy differences between the Cast, CAW, and 3DP groups . Moreover, the polymerization shrinkage of 3DP polymeric material until the casting time and the light sensitivity may have affected the fit values of the 3DP group. Marginal accuracy is effective in the structural longevity of the restoration cement. Considering the work of the researchers McLean and von Fraunhofer , the MDVs in all fabrication technique groups were below the clinically acceptable limit (< 120 μm). The results of this study were not in line with some of the previous studies . In this study, the discrepancy values for the SLS group were not lower than those of the conventional casting methods. This might be explained by the morphology of the abutment and different parameters of laser sintering systems. Örtorp et al. . reported the mean internal gap as 144 ± 67 μm for the conventional Cast group and 151 ± 58 for the DMLS group. In this study, the mean IDVs were 114.36 ± 10.15 for the Cast group and 83.44 ± 13.18 for the DMLS group. This difference might be due to the measurement points and the evaluation methods used. Conventional fabrication methods showed higher discrepancy values than the other groups for all abutments. Modeling was done manually in the Cast group and the coping margins were manually adjusted, and the distortions or ununiform thickness may have affected the adaptation of the wax specimens. Although the substructure was obtained with the conventional cast method in the Cast, CAW, and 3DP groups, the difference in the pattern preparation methods and material-dependent variations might be considered reasons for the differences between the groups . The DMLS and SLM groups showed better and more consistent fit values than those of the other methods, which is in accordance with previous studies . However, some studies concluded that conventional cast methods showed better results than the DMLS system . Abutment morphology, laser beam, laser intensity, and the material used for the replica technique may have led to these results . In this study, SRT was selected for discrepancy measurements because it is a convenient and inexpensive method and has been performed in previous studies . The SRT method is nondestructive, and its reliability is high . In a previous study, micro-computed tomography and the SRT method were comparatively evaluated . It was reported that the values in all the regions measured with micro-computed tomography were not significantly different from those measured using SRT . However, the material used in SRT may impact the measured discrepancy values. The rupture strength of the material, the ability to imitate cement material, and thixotropy are important . In light of the findings from this in-vitro study, laser metal sintering methods might be a more suitable alternative to conventional methods, especially in scenarios where the use of FDPs involves abutments with complex morphological features. However, this study did not consider the effects of veneer material, different framework materials, various cement types, and cementation techniques. Future studies should therefore take these factors and the dynamic oral conditions into account. Within the limitations of this in-vitro study, abutment type played a significant role in discrepancy values in the Cast method compared to that of the CAW method, in which the discrepancy values did not significantly affected from abutment type. Furthermore, the SLM and 3DP methods did not appear to be affected by abutment type with respect to marginal discrepancies. No effect of abutment type on occlusal discrepancies was observed in laser sinterization methods (DMLS and SLM). However, internal discrepancies in the SLM method might be influenced by abutment type. Despite the significant differences between methods, the internal and occlusal discrepancies values might be expected similar when 3DP or Cast fabrication methods were preferred. Moreover, the effects of SLM, CAW, and DMLS fabrications methods on discrepancy values might be similarly. However, DMLS can be considered as a fabrication method for achieving optimal fit in terms of marginal, internal, and occlusal accuracy in three-unit FDPs, particularly in maxillary posterior regions. |
Sleep in Residents: A Comparison between Anesthesiology and Occupational Medicine Interns | c2e27884-3893-4564-88af-ce9c3d1ca244 | 9915358 | Preventive Medicine[mh] | Sleep has a significant impact on mental and physical health . Workers with sleep deprivation are exposed to disorders such as anxiety, depression , cognitive impairment and metabolic syndrome . They also have a lower resilience to stress and a greater risk of occupational injuries and errors . Sleep problems are particularly common in healthcare workers (HCWs) . Resident physicians are among the most exposed to sleep problems . The sleep deprivation of resident physicians is a problem that can negatively impact their health as well as that of their patients, as sleepiness is known to increase the risk of errors . In the past few years, several countermeasures have been taken to prevent this risk . Sleep deprivation affects many categories of doctors differently, depending on the type of activity that is required at night. Anesthesiologists are known to be particularly prone to sleep problems and this is especially true in conditions of understaffing and excessive workload. During the COVID-19 pandemic, resident anesthetists in a hub hospital for COVID-19 patients experienced a high rate of sleep disturbance . The main factor in sleep loss was to be attributed both to the tension linked to the responsibility of facing a new disease with unprecedented procedures and techniques and to the excessive load of night work . Mostly during the first wave, the anxiety of facing the new disease with new safety procedures and new diagnostic and therapeutic methods and the fear of contracting the disease were the main reasons for sleep problems . Afterwards, when the frontline staff gained full control of the new safety measures, sleep problems appeared to be mainly linked to excessive workload and lack of time for meditation and relaxation . Since the health and alertness of anesthesiologists are critical to the quality of their assistance, the prevention of sleep problems is a priority for healthcare companies . In Italy, national health authorities adopted an emergency solution to solve this problem by increasing the number of anesthesiologists, which allowed a better distribution of night work and thus reduced the workload . According to this policy, in some European countries, such as Italy, university hospitals increased the number of available scholarships, and some residents were hired with fixed-term contracts to deal with the emergency . As a result of these measures, the average number of night shifts that each resident has had to cover has decreased compared to what it was before the pandemic. The number of night shifts covered is of particular importance in defining health risk. In Italy, in accordance with the European Directives, a “night worker” is considered to be someone who covers more than 80 night shifts a year. The discussion is open on the possibility that any health effects may also be evident for workers who work fewer than 80 nights a year. For example, does the health of resident anesthetists who do not reach 80 night shifts in a year have any differences with the health of other residents who perform the same hours of service but during the day? Nowadays, two years after the start of the COVID-19 pandemic, the need has arisen to evaluate if the current working conditions of anesthesiology residents who work night shifts during their training may have a negative impact on their health. We decided to carry out this study with the aims of evaluating (1) the quantity and quality of sleep in a sample of anesthesiology residents (ARs, who work night shifts) in comparison with that of occupational medicine residents (OMRs, who do not perform night work); and (2) the association between sleep and cardiac frequency, footsteps, work-related distress, fatigue, anxiety, depression and happiness. 2.1. Study Design and Population This cross-sectional study was conducted between April 2022 and July 2022. Participants included 37 residents: anesthesiology residents (ARs, who work night shifts, n = 21, 11 males and 10 females, aged 29.3 ± 3.2) and occupational medicine residents (OMRs, who do not work night shifts, n = 16, 7 males and 9 females, aged 31.3 ± 2.8). Residents in Italy work the same number of hours as hospital physicians, 38 h per week . Night-shift work is defined as a period of at least 3 consecutive hours in the interval between 0 a.m. and 5 a.m. . ARs were working between 4 and 6 night shifts per month, so they did not reach 80 night shifts per year, which are considered in Italy the minimum level to be defined as “night workers”. All residents were asked to wear an activity tracker (see below) and complete a questionnaire with a series of validated scales. The research was conducted according to the principles of the Helsinki Convention on unpaid, healthy adult volunteers who signed an informed consent form. The research was approved by the Ethics Committee of the Catholic University of the Sacred Heart (Project 1226, 4 November 2016). 2.2. Activity Tracker All residents wore an activity tracker (Fitbit ® Inspire HR, Fitbit Inc., San Francisco, CA, USA) on their non-dominant wrist for one week. This device objectively measured the amount of nighttime sleep (in minutes), cardiac frequency and number of footsteps per day. Data were extracted from the device, considering the mean daily values for each participant. To compare the two groups, we excluded night sleep minutes from night shifts. 2.3. Questionnaire At the end of the week, each participant completed a questionnaire including sociodemographic data (e.g., age, sex) and validated scales exploring sleep (e.g., sleep quality, daytime sleepiness), fatigue, mental health (e.g., anxiety, depression), work-related distress and happiness. Sleep quality was evaluated using the Pittsburgh Sleep Quality Index (PSQI) , Italian version , a self-report on subjective sleep quality over the previous 4 weeks with 18 questions. The first four questions were analyzed at the following times: bedtime, minutes to fall asleep, get-up time and hours of sleep per night. The next 10 questions investigated the frequency of sleeping troubles caused by different reasons (e.g., not being able to fall asleep, waking up during the night, needing to go to the bathroom, breathing problems, coughing, feeling too cold or too hot, having bad dreams and experiencing pain). The following two questions inquired about the use of sleep medication and trouble staying awake during daily activities. Each of these questions had to be answered on a 4-point scale ranging from “never” to “three times or more a week”. The last two questions asked if it had been a problem for the participant to keep up enough enthusiasm to get things done (with a 4-point scale ranging from “no problem at all” to “a very big problem”) and a subjective rating of the participants’ sleep quality on a 4-point scale from “very good” to “very bad”. The 18 items of the PSQI were summed up to a general score. Higher scores represented worse sleep quality: according to the cut-off score suggested by the authors of the questionnaire , subjects with scores higher than 5 were considered “poor sleepers”. In this study, Cronbach’s alpha showed good internal consistency reliability (0.792). Sleepiness was measured with the Italian version of the Epworth Sleepiness Scale (ESS) . Participants rated their chances of sleeping in eight situations on a 4-point scale, scoring from 0 (“would never doze”) to 3 (“high chance of dozing”). The questions investigated the following daily activities: sitting and reading, watching TV, sitting inactive in a public place, as a passenger in a car for an hour without a break, lying down to rest in the afternoon, sitting and talking to someone, sitting quietly after lunch without alcohol and sitting in a car while stopped for a few minutes in the traffic. The results could have a minimum score of 0 and a maximum score of 24, with the normal range going between 0 and 10. A score above 10 indicated high sleepiness during the daily activities. Cronbach’s alpha for the ESS in this study showed acceptable internal consistency (0.704). Fatigue was measured with the Fatigue Assessment Scale (FAS) , Italian version. The FAS consists of 10 questions, including two subscales: mental fatigue and physical fatigue. Each response was graded on a 5-point Likert scale from 1 (“never”) to 5 (“always”). Scores on questions 4 and 10 have been inversely recoded. By adding the scores of all the answers, the total FAS score was obtained, with a range of 10 to 50. A total FAS score less than 22 indicates that there is no fatigue; a score greater than or equal to 22 indicates that there is fatigue . Cronbach’s alpha in this study showed acceptable reliability (0.620). Anxiety and depression were assessed using the Italian version of “Goldberg’s Anxiety and Depression Scale” (GADS) , referring to the previous 10-day period. GADS is composed of two scales of nine binary questions each; one point is awarded for each positive answer. A score of 5 or more on the anxiety subscale, or 2 or more on the depression subscale, indicates suspected clinically evident anxiety or depression . The reliability of the GADS subscales in this study was high (Cronbach’s alpha was 0.839 for anxiety and 0.789 for depression). The perception of work-related stress was measured using the Italian version of Siegrist’s short “Effort–Reward Imbalance” (ERI) scale , which contains three questions for the effort variable and seven for the reward variable. All items had graded responses on a 4-point Likert scale from 1 (“totally agree”) to 4 (“totally disagree”). The resulting subscales were between 3 and 12 (effort) and between 7 and 28 (reward), respectively. The weighted effort/reward imbalance (ERI) ratio indicates the level of occupational distress. According to the literature , ERI values greater than 1 indicate distress . Cronbach’s alphas were 0.720 for the effort subscale and 0.751 for the reward subscale. Happiness was measured using the Ab-del-Khalek single item (“Do you feel happy in general?”) answered on an 11-point scale (0–10) . 2.4. Statistical Analyses IBM Corp. software released in 2019 was used for statistical analysis (IBM SPSS Statistics for Windows, Version 26.0. Armonk, NY, USA: IBM Corp., release 15.0). All data were first analyzed for normality of distribution using the Kolmogorov–Smirnov test of normality. Descriptive statistics were performed for questionnaire scores, continuous variables were expressed as mean ± SD, categorical variables were displayed as frequencies, and the appropriate parametric (Student’s t) or non-parametric (Mann–Whitney U) test was used to assess the significance of the differences between subgroups. Correlations were calculated with the Pearson or Spearman correlation coefficients, as appropriate. Due to the small sample size, a multivariate approach to statistical analysis was not recommended; however, the effect size (according to Cohen 1988) of the r metric for the Pearson correlation coefficient was reported in order to estimate the magnitude and clinical relevance of the relationship between variables independently of the number of enrolled subjects. The effect of sleep and stress on daytime sleepiness was studied using multiple linear regression analysis. Sleep duration, sleep quality and perceived stress were entered as predictors, and sleepiness measured with the Epworth scale was entered as a dependent variable. After adjustment for age and gender, a p -value <0.05 was considered statistically significant. This cross-sectional study was conducted between April 2022 and July 2022. Participants included 37 residents: anesthesiology residents (ARs, who work night shifts, n = 21, 11 males and 10 females, aged 29.3 ± 3.2) and occupational medicine residents (OMRs, who do not work night shifts, n = 16, 7 males and 9 females, aged 31.3 ± 2.8). Residents in Italy work the same number of hours as hospital physicians, 38 h per week . Night-shift work is defined as a period of at least 3 consecutive hours in the interval between 0 a.m. and 5 a.m. . ARs were working between 4 and 6 night shifts per month, so they did not reach 80 night shifts per year, which are considered in Italy the minimum level to be defined as “night workers”. All residents were asked to wear an activity tracker (see below) and complete a questionnaire with a series of validated scales. The research was conducted according to the principles of the Helsinki Convention on unpaid, healthy adult volunteers who signed an informed consent form. The research was approved by the Ethics Committee of the Catholic University of the Sacred Heart (Project 1226, 4 November 2016). All residents wore an activity tracker (Fitbit ® Inspire HR, Fitbit Inc., San Francisco, CA, USA) on their non-dominant wrist for one week. This device objectively measured the amount of nighttime sleep (in minutes), cardiac frequency and number of footsteps per day. Data were extracted from the device, considering the mean daily values for each participant. To compare the two groups, we excluded night sleep minutes from night shifts. At the end of the week, each participant completed a questionnaire including sociodemographic data (e.g., age, sex) and validated scales exploring sleep (e.g., sleep quality, daytime sleepiness), fatigue, mental health (e.g., anxiety, depression), work-related distress and happiness. Sleep quality was evaluated using the Pittsburgh Sleep Quality Index (PSQI) , Italian version , a self-report on subjective sleep quality over the previous 4 weeks with 18 questions. The first four questions were analyzed at the following times: bedtime, minutes to fall asleep, get-up time and hours of sleep per night. The next 10 questions investigated the frequency of sleeping troubles caused by different reasons (e.g., not being able to fall asleep, waking up during the night, needing to go to the bathroom, breathing problems, coughing, feeling too cold or too hot, having bad dreams and experiencing pain). The following two questions inquired about the use of sleep medication and trouble staying awake during daily activities. Each of these questions had to be answered on a 4-point scale ranging from “never” to “three times or more a week”. The last two questions asked if it had been a problem for the participant to keep up enough enthusiasm to get things done (with a 4-point scale ranging from “no problem at all” to “a very big problem”) and a subjective rating of the participants’ sleep quality on a 4-point scale from “very good” to “very bad”. The 18 items of the PSQI were summed up to a general score. Higher scores represented worse sleep quality: according to the cut-off score suggested by the authors of the questionnaire , subjects with scores higher than 5 were considered “poor sleepers”. In this study, Cronbach’s alpha showed good internal consistency reliability (0.792). Sleepiness was measured with the Italian version of the Epworth Sleepiness Scale (ESS) . Participants rated their chances of sleeping in eight situations on a 4-point scale, scoring from 0 (“would never doze”) to 3 (“high chance of dozing”). The questions investigated the following daily activities: sitting and reading, watching TV, sitting inactive in a public place, as a passenger in a car for an hour without a break, lying down to rest in the afternoon, sitting and talking to someone, sitting quietly after lunch without alcohol and sitting in a car while stopped for a few minutes in the traffic. The results could have a minimum score of 0 and a maximum score of 24, with the normal range going between 0 and 10. A score above 10 indicated high sleepiness during the daily activities. Cronbach’s alpha for the ESS in this study showed acceptable internal consistency (0.704). Fatigue was measured with the Fatigue Assessment Scale (FAS) , Italian version. The FAS consists of 10 questions, including two subscales: mental fatigue and physical fatigue. Each response was graded on a 5-point Likert scale from 1 (“never”) to 5 (“always”). Scores on questions 4 and 10 have been inversely recoded. By adding the scores of all the answers, the total FAS score was obtained, with a range of 10 to 50. A total FAS score less than 22 indicates that there is no fatigue; a score greater than or equal to 22 indicates that there is fatigue . Cronbach’s alpha in this study showed acceptable reliability (0.620). Anxiety and depression were assessed using the Italian version of “Goldberg’s Anxiety and Depression Scale” (GADS) , referring to the previous 10-day period. GADS is composed of two scales of nine binary questions each; one point is awarded for each positive answer. A score of 5 or more on the anxiety subscale, or 2 or more on the depression subscale, indicates suspected clinically evident anxiety or depression . The reliability of the GADS subscales in this study was high (Cronbach’s alpha was 0.839 for anxiety and 0.789 for depression). The perception of work-related stress was measured using the Italian version of Siegrist’s short “Effort–Reward Imbalance” (ERI) scale , which contains three questions for the effort variable and seven for the reward variable. All items had graded responses on a 4-point Likert scale from 1 (“totally agree”) to 4 (“totally disagree”). The resulting subscales were between 3 and 12 (effort) and between 7 and 28 (reward), respectively. The weighted effort/reward imbalance (ERI) ratio indicates the level of occupational distress. According to the literature , ERI values greater than 1 indicate distress . Cronbach’s alphas were 0.720 for the effort subscale and 0.751 for the reward subscale. Happiness was measured using the Ab-del-Khalek single item (“Do you feel happy in general?”) answered on an 11-point scale (0–10) . IBM Corp. software released in 2019 was used for statistical analysis (IBM SPSS Statistics for Windows, Version 26.0. Armonk, NY, USA: IBM Corp., release 15.0). All data were first analyzed for normality of distribution using the Kolmogorov–Smirnov test of normality. Descriptive statistics were performed for questionnaire scores, continuous variables were expressed as mean ± SD, categorical variables were displayed as frequencies, and the appropriate parametric (Student’s t) or non-parametric (Mann–Whitney U) test was used to assess the significance of the differences between subgroups. Correlations were calculated with the Pearson or Spearman correlation coefficients, as appropriate. Due to the small sample size, a multivariate approach to statistical analysis was not recommended; however, the effect size (according to Cohen 1988) of the r metric for the Pearson correlation coefficient was reported in order to estimate the magnitude and clinical relevance of the relationship between variables independently of the number of enrolled subjects. The effect of sleep and stress on daytime sleepiness was studied using multiple linear regression analysis. Sleep duration, sleep quality and perceived stress were entered as predictors, and sleepiness measured with the Epworth scale was entered as a dependent variable. After adjustment for age and gender, a p -value <0.05 was considered statistically significant. ARs had a shorter sleep duration than OMRs, according to sleep recordings. On average, OMRs slept one hour and twenty minutes more than ARs every night ( p < 0.001). The heart rate of ARs was also significantly higher than that of OMRs ( p < 0.001). Conversely, the physical activity of the two groups, evaluated as the average number of footsteps taken, showed no significant difference . The analysis of subjective data collected through questionnaires revealed that in both groups, perceived stress was on average high, above all due to the effect of effort, which has an average score of 8.2 + 2.2, which is 68.3% of the theoretical maximum of the scale (12 points), while the rewards obtained from work have an average score of 13.7 + 3.7 (48.9% of the scale maximum). Work-related fatigue was not different in the two groups, while sleepiness was much greater in ARs than in OMRs: the difference was highly significant ( p < 0.001). Residents who do not work night shifts also reported a higher level of happiness in life than ARs, while anxiety and depression levels were not significantly different between the two groups . We investigated the correlations between the variables collected in the study in the set of residents and in the two sub-groups of ARs and OMRs . Age had no direct correlation with any of the variables, while gender was correlated with anxiety; in fact, female doctors had higher levels of anxiety than males . Sleeping time was significantly related to happiness, whereas heart rate was negatively related to happiness. Poor sleep quality was positively associated with anxiety, depression, daytime sleepiness and fatigue and negatively associated with happiness. As expected, anxiety and depression were strongly correlated with each other and negatively correlated with happiness. Perceived fatigue appeared to be related to stress, anxiety, depression and poor-quality sleep . Multiple regression analysis indicated that the major determinant of daytime sleepiness in the whole group was sleep quality; sleep duration, in fact, did not show a significant predictive value in the multivariate model, and occupational stress was also not significant . Only in the subgroup of Ars is sleep duration a predictor of sleepiness, as is age. This study demonstrated that resident anesthesiologists who work night shifts, even if the number of such shifts is smaller than what is considered likely to alter biorhythms, sleep significantly less than their colleagues who are not engaged in night shifts and report a higher level of daytime sleepiness. Sleep quality was also correlated with daytime sleepiness in residents. ARs also had a higher mean heart rate than OMRs, while walking the same number of footsteps in their daily routine. This is probably to be interpreted as evidence of the greater criticality of the clinical tasks required of anesthesiologists compared to those performed by OMRs. The immediate transition from rest to full attention that is typical of ARs shifts is not frequent in OMRs activities, which are generally planned in advance. The reported levels of occupational stress, fatigue, anxiety and depression did not show differences between ARs and OMRs. However, ARs were significantly less happy than OMRs, and this finding deserves further investigation. The results went in the direction suggested by the literature. It is not surprising that anesthesiologists sleep less than non-night-shift residents. It is well known that night workers can develop a syndrome called Shift Work Sleep–Wake Disorder (SWSWD) . A meta-analysis of longitudinal studies confirmed that SWSWD is linked to a higher overall risk of negative mental health outcomes, specifically for depressive symptoms . Shift work has a negative impact on cognitive functions and increases the incidence of metabolic disorders . However, the resident anesthesiologists recruited for this study were not night workers but covered only a limited number of night shifts and therefore should not have had any health problems. The shorter sleep duration in Ars compared to that of other residents could indicate insufficient recovery after night shifts, or it could be the effect of other factors, such as workload and responsibilities, which may interfere with sleep. The effects of workload, anxiety and emotional factors have been invoked to explain sleep disturbances reported by healthcare workers during the COVID-19 pandemic . The female gender seems more prone to such disorders . Healthcare students, on the other hand, are at risk of sleep disturbances from stressful academic and clinical workloads, even if they do not work night shifts . The higher average heart rate measured in the ARs could indicate that they have a higher mental workload than the OMRs, and this may partially justify the lower sleep time. An aspect that can significantly differentiate the work of ARs from that of OMRs is that the former have been professionally exposed to calls for emergency interventions, while the latter have not. Emergency and first aid at night have a greater mental weight than the planned prevention work carried out by the OMRs during the day. Healthcare personnel have reported hypomanic symptoms brought on by on-calls with inadequate sleep, demonstrating the impact of working conditions on their wellness and raising questions about their ability to make decisions after long work shifts . According to an observational study, after completing a night call duty, anesthesiologists’ reaction times significantly rose; performance decline was linked to a larger subjective reliance on avoidance as a coping mechanism . Sleep deprivation affects residents’ ability to handle crisis situations in anesthesia. The main mistakes that were made were incorrect drug administration and dosage, a failure to recognize hypotension in time and a failure to inform the surgical team of the situation . A study on fatigue in emergency medicine showed that residents were significantly less alert at the completion of the night shift . In terms of mental health, partial sleep deprivation impairs anesthesiologists’ overall mood state and cognitive abilities, causing them to become tenser, angrier, more exhausted, confused and irritable, and sleepier . Additionally, prolonged duty shifts or quick returns to work with insufficient recovery time can disturb the circadian rhythm and cause acute and chronic sleep deprivation, which may lead to detrimental effects on chronic outcome measures (e.g., functional ability and work–life balance) . Moreover, a bidirectional relationship between sleep and mood has been found, and substantial shifts in sleep timing have been shown to lead to shorter sleep and poorer mood . In some studies, anesthesiologists presented a poorer quality of sleep, with excessive daytime somnolence, perceived stress and higher sedative use compared with other healthcare populations . Stress brought on by inadequate sleep and overwork may be a factor in sudden spikes in blood pressure and sympathetic nervous system activity . A study examining whether heart rate variability (HRV) varied amongst different physician specialties after recovery from day work and night call duty found that anesthesiologists had less dynamic HRV following day work and during night call duty, indicating higher levels of physiological stress . The alterations that have been found in the literature on anesthesiologists in hospital service are in the same direction as the effects observed in this small group of resident anesthetists, who work night shifts less frequently and for a shorter period of time than doctors hired on permanent contracts. The study presented here, which was conducted on a small group of residents, has the merit of highlighting a problem that deserves further study. If, in fact, young ARs already show a reduction in sleep and an increase in daytime sleepiness due to the few night shifts worked, it will be necessary to prepare adequate support measures. Organizational measures may help to balance the emotional involvement of residents in their work, which has been attributed to the gap between high professional demand and trainees’ lack of experience and knowledge . A well-designed shift organization may affect the frequency of sleep disturbances, but the evidence about the best type of schedule is still conflicting. For example, night-float rotations, which were designed to alleviate the workload of residents on night call, induced disturbances of sleep and mood and decreased alertness in residents . There is undoubtedly a need for well-designed longitudinal studies that compare the different options in order to find the organizational methods most respectful of workers’ well-being and patient safety. The observation that night-shift healthcare professionals are at risk for sleep deprivation and/or circadian rhythm abnormalities has led many health agencies to plan health promotion interventions . Programs should endeavor to take the necessary safeguards because sleep loss affects workers’ capacity for complex rational decision making. This can be accomplished by keeping nighttime complexity and decision-making speed as low as possible. Additionally, some organizational and individual actions can hasten the body’s adaptation to a new shift and lower the risk of circadian rhythm disorders. The most Important limitation of this pilot study was the small sample size. Studies conducted on a larger number of residents may provide more reliable results. A second important limitation is the cross-sectional nature of the study, which does not allow for asserting the causality of the observed associations. Only a longitudinal study could indicate whether the lower level of happiness observed in anesthesiologists was a consequence of occupational exposure, or whether it was rather an accidental finding. Finally, the technical characteristics of the device used must also be considered. Even if a comparative study of the products available on the market indicates that the used device is reliable , we have observed that, since it is very sensitive, connection could be lost when detaching from the skin (e.g., during the night due to accidental movements that cannot be controlled or limited), As a result, micro-awakenings were frequently recorded and a lower number of hours of actual sleep were registered. We found that this problem occurred in approximately 10% of registrations, with equal frequency in ARs and OMRs. In two cases, the frequency of interruptions was high enough to require a repetition of the registration. The observation that covering only one night shift per week is associated with a very significant reduction in sleep duration in ARs and is accompanied by increased daytime sleepiness should stimulate further longitudinal research on more numerous case series than this one pilot study. If confirmed, the results should lead to changes in work organization in order to favor the recovery of correct sleeping conditions and alertness. Mitigating measures should aim at implementing a sleep management system, i.e., a coordinated sequence of interventions at various levels (structural, organizational and individual) for the improvement of sleep hygiene and the prevention of drowsiness . Healthcare companies have a vested interest in carrying out this type of intervention since daytime sleepiness is particularly hazardous for physicians performing first aid and emergency intervention in critically ill patients. |
Subsets and Splits